Carousel review, 12 October

A hundred and twenty custom websites cost me about 44 dollars, which is about 37 cents each (read 2026-10-10). Here is the part I did not say yet. On 10 October 2026, 106 built fine, 9 came out thin, 3 were the wrong trade and 2 failed. None are sent. Nobody has seen one. I do not know if it gets a single reply. So the test is simple: send a handful, count the replies, and post the number whatever it is. Comment FLEET and I'll send the cost table.

#AIagents #localSEO #buildinpublic

We ran a first batch of 120 AI-built demo sites on 10 October 2026: 60 movers and 60 roofers. 106 built fine, 9 thin, 3 the wrong trade, 2 failed. Total cost about 44 dollars, about 37 cents a lead (read 2026-10-10). Built is not sent. None are live and we have no reply data yet. The lesson for agency owners: cost the failures next to the successes, and test with a handful before you scale. Reply below with what you would pay for a demo made from your own real details.

#agencylife #localSEO

Batch one: 120 leads, 106 built, 9 thin, 3 wrong trade, 2 failed, about 44 dollars all in (read 2026-10-10). The cost is mostly the AI copy pass, then Google data. Nothing deployed yet, and the next step is a small send and an honest reply count. vespio.ai

#AIagents #buildinpublic

My AI agent kept saying "done". It was rounding up. Done meant it wrote the thing, not that it checked the thing.

So I built a gate. Every plan has one line: what must be true when this is finished. A script runs that line, and if it fails the word done is refused.

If you're 18 to 27 and building with AI, steal the checklist on slide 6.

#AIagents #Claude

Comment FLEET

My AI agent kept saying "done". It was rounding up. Done meant it wrote the thing, not that it checked the thing.

So I built a gate. Every plan has one line: what must be true when this is finished. A script runs that line, and if it fails the word done is refused.

If you're 18 to 27 and building with AI, steal the checklist on slide 6.

#AIagents #Claude

Comment FLEET

My AI agent kept saying "done". It was rounding up. Done meant it wrote the thing, not that it checked the thing.

So I built a gate. Every plan has one line: what must be true when this is finished. A script runs that line, and if it fails the word done is refused.

If you're 18 to 27 and building with AI, steal the checklist on slide 6.

#AIagents #Claude

Comment FLEET

My AI agent kept saying "done". It was rounding up. Done meant it wrote the thing, not that it checked the thing.

So I built a gate. Every plan has one line: what must be true when this is finished. A script runs that line, and if it fails the word done is refused.

If you're 18 to 27 and building with AI, steal the checklist on slide 6.

#AIagents #Claude

Comment FLEET

My AI agent kept saying "done". It was rounding up. Done meant it wrote the thing, not that it checked the thing.

So I built a gate. Every plan has one line: what must be true when this is finished. A script runs that line, and if it fails the word done is refused.

If you're 18 to 27 and building with AI, steal the checklist on slide 6.

#AIagents #Claude

Comment FLEET

My AI agent kept saying "done". It was rounding up. Done meant it wrote the thing, not that it checked the thing.

So I built a gate. Every plan has one line: what must be true when this is finished. A script runs that line, and if it fails the word done is refused.

If you're 18 to 27 and building with AI, steal the checklist on slide 6.

#AIagents #Claude

Comment FLEET

The most expensive word in an agency is "done" said without a check.

We attach one line to every plan: what must be true when it is finished. A script runs it at wrap-up, and the word done is refused if it fails. Live means we opened the live page. Tested means we ran it.

Honest limit: the gate checks only what we told it to check.

What is the last thing called finished in your team that was not?

#agencylife

Reply below

Agents round up, so we made the rounding impossible. One outcome line per plan, one script that runs it, and no done until it passes.

The proof pairs are simple: live means the live page opened, tested means it ran, fixed means the old failing case now passes.

#AIagents #Claude

vespio.ai

In May 2026 I got an AI bill for 3,357 dollars. I had read the spend cap as 200. It was 200,000.

If you're 18 to 27 and learning to build online with AI agents, this is the mistake I want you to skip. No cost check before the script ran. No caching. A cap I misread.

The fix was not "be careful". It was a hook that blocks any call with no cost check first.

Save this for your next agent project.

#AIagents #Claude

Comment FLEET

In May 2026 I got an AI bill for 3,357 dollars. I had read the spend cap as 200. It was 200,000.

If you're 18 to 27 and learning to build online with AI agents, this is the mistake I want you to skip. No cost check before the script ran. No caching. A cap I misread.

The fix was not "be careful". It was a hook that blocks any call with no cost check first.

Save this for your next agent project.

#AIagents #Claude

Comment FLEET

In May 2026 I got an AI bill for 3,357 dollars. I had read the spend cap as 200. It was 200,000.

If you're 18 to 27 and learning to build online with AI agents, this is the mistake I want you to skip. No cost check before the script ran. No caching. A cap I misread.

The fix was not "be careful". It was a hook that blocks any call with no cost check first.

Save this for your next agent project.

#AIagents #Claude

Comment FLEET

In May 2026 I got an AI bill for 3,357 dollars. I had read the spend cap as 200. It was 200,000.

If you're 18 to 27 and learning to build online with AI agents, this is the mistake I want you to skip. No cost check before the script ran. No caching. A cap I misread.

The fix was not "be careful". It was a hook that blocks any call with no cost check first.

Save this for your next agent project.

#AIagents #Claude

Comment FLEET

In May 2026 I got an AI bill for 3,357 dollars. I had read the spend cap as 200. It was 200,000.

If you're 18 to 27 and learning to build online with AI agents, this is the mistake I want you to skip. No cost check before the script ran. No caching. A cap I misread.

The fix was not "be careful". It was a hook that blocks any call with no cost check first.

Save this for your next agent project.

#AIagents #Claude

Comment FLEET

In May 2026 I got an AI bill for 3,357 dollars. I had read the spend cap as 200. It was 200,000.

If you're 18 to 27 and learning to build online with AI agents, this is the mistake I want you to skip. No cost check before the script ran. No caching. A cap I misread.

The fix was not "be careful". It was a hook that blocks any call with no cost check first.

Save this for your next agent project.

#AIagents #Claude

Comment FLEET

In May 2026 I got an AI bill for 3,357 dollars. I had read the spend cap as 200. It was 200,000.

If you're 18 to 27 and learning to build online with AI agents, this is the mistake I want you to skip. No cost check before the script ran. No caching. A cap I misread.

The fix was not "be careful". It was a hook that blocks any call with no cost check first.

Save this for your next agent project.

#AIagents #Claude

Comment FLEET

In May 2026 I got an AI bill for 3,357 dollars. I had read the spend cap as 200. It was 200,000.

If you're 18 to 27 and learning to build online with AI agents, this is the mistake I want you to skip. No cost check before the script ran. No caching. A cap I misread.

The fix was not "be careful". It was a hook that blocks any call with no cost check first.

Save this for your next agent project.

#AIagents #Claude

Comment FLEET

We run client work with AI agents, and in May 2026 one unchecked script and a misread spend cap became a 3,357 dollar bill.

Three causes: no cost check before the call, no caching, and a cap that read 200 when it was 200,000. Our fix is mechanical, not a reminder. A hook blocks any API call that has no cost check first.

If your team uses agents, who owns the cap number, and when did someone last read it?

#AIagents

Reply below

Building on agents means the bill is part of the architecture. Ours taught us that in May 2026: 3,357 dollars from a script with no cost check, no caching and a misread cap.

The design rule we kept: guards live in hooks, not in good intentions. Estimate first, cap the run, cache the static prompt, block the rest.

#AIagents #Claude

vespio.ai

The most common way to check a redirect lies to you. It follows the redirect all the way to a 200 and says everything is fine, even when the page is gone.

If you're 18 to 27 and building sites for people, learn the one-line check on slide 5. It asks for the redirect itself and does not follow it.

#SEO #websitemigration

Comment AUDIT and I'll run a free AI-visibility check on your business.

Moving a site comes down to one question: does every old URL that ranks go somewhere useful?

The common check (curl -I -L) follows to a 200 and hides the problem. We ask for the redirect itself: status code plus target, no following. A 301 to a topical page is good. A 404 is a lost ranking. Cutovers fail silently, and the bleed shows up weeks later as lost calls.

How does your team verify redirects?

#SEO

Comment AUDIT for a free AI-visibility check.

Verification should fail loudly. A check that follows redirects turns a 404 into a 200 and reports success.

One request per old URL, print the status and the target, never follow it. That is the shape of a check you can trust.

#SEO #websitemigration

vespio.ai

Before you move a website that ranks, check one thing: does every old URL still land somewhere?

If you're 18 to 27 and building sites for local businesses, this one check protects you from the worst kind of mistake, the silent one.

Checklist in the slides. Save it.

#AIagents #Claude

Comment AUDIT and I'll run a free AI-visibility check on your business.

Before you move a website that ranks, check one thing: does every old URL still land somewhere?

Our gate looks at three killers: redirects drafted but not deployed, page count compressed, and junk links pointing in. It also proves the form and the tracking before the switch.

Who signs off the move on your team?

#localSEO

Comment AUDIT for a free AI-visibility check.

A migration gate is a list of deterministic checks: status codes per old URL, page parity, form delivery, tag firing. If any fails, the flip waits.

No opinions, just exit codes.

#buildinpublic #Claude

vespio.ai

Do this tonight: fill in your own website form and start a timer.

If you're 18 to 27 and selling local lead gen, this is the 5 minute test I'd run on every site before I told anyone it works. Write down what arrives and when.

Steps on the slides. Try it on your own site first.

#AIagents #Claude

Comment AUDIT and I'll run a free AI-visibility check on your business.

Do this tonight: fill in your own website form and start a timer.

If nothing happens in five minutes, that is the finding. Note where it stopped: form, inbox, phone or nowhere. Fix the first gap and run it again.

When did you last test your own form?

#localSEO

Comment AUDIT for a free AI-visibility check.

The cheapest monitor is a human with a timer. Submit, watch the phone and the inbox, log the gap.

That test is the spec for any lead-response automation we build.

#buildinpublic

vespio.ai

Most leads that go quiet are not lost. They are waiting for the second message nobody sent.

If you're 18 to 27 and want to be useful to local businesses, learn this ladder: day 1 answer, day 3 add something useful, day 7 close kindly.

Save it for your first client.

#AIagents #Claude

Comment AUDIT and I'll run a free AI-visibility check on your business.

Most leads that go quiet are not lost. They are waiting for the second message nobody sent.

Our ladder: day 1 a fast answer, day 3 something useful, day 7 a kind close. Three templates, two reminders, stop when they reply.

What does your team send after the first reply?

#localSEO

Comment AUDIT for a free AI-visibility check.

A follow-up ladder is a tiny state machine: three templates, two timers, one stop condition. Easy to automate, easy to get wrong without the stop.

#buildinpublic

vespio.ai

I have not written a line of code today and Claude Code has run my whole morning. Four commands: /d-session, /do, /pbp, /wrap. Here is what each one does, in the order I type them. Comment FLEET

#claudecode #aiagents #buildinpublic

Our working day in Claude Code is four commands: one shows the day, one works a plan, one says where we are, one closes and commits. Each is a saved checklist, so John and I run the same loop. Reply below

#claudecode #aiagents #agencyowner

Claude Code is not only a coding tool. A skill is a checklist with a name, and four of them make a day: /d-session, /do, /pbp and /wrap. This is the loop we build and run Vespio with. vespio.ai

#claudecode #aiagents #buildinpublic

I built a guard to block risky commands. Then I typed the override into a comment and it let me through.

If you're 18 to 27 and learning to build with AI agents, test your guards by trying to fool them. The fix was small: the override only counts as a leading assignment, never as text inside the command.

Save this for your next agent project.

#AIagents #Claude

Comment FLEET

A safety guard that can be talked past is only a suggestion. Ours accepted an override typed into a comment, so we changed it to accept the override only as a leading assignment that names the exact change.

For anyone running agents with access to live systems: has anyone tried to get past your guards on purpose?

#AIagents

Reply below

Guards should match on what runs, not on the text around it. Ours once accepted an override placed in a comment.

The fix: leading assignment only, the override must name the thing it changes, and that name must appear in the command itself. Then a red case that tries to fool it.

#AIagents #Claude

vespio.ai

I let AI write my demo websites. I do not let it make up a single number. Every claim needs a source on the page: years in business from their own site, a rating from their Google listing. On 10 October 2026 I checked 6 years-in-business claims against the live sites and all 6 matched (read 2026-10-10). The copy can still be wrong, so I read every one before it goes anywhere. Using AI is easy. Using it honestly is the skill. Comment FLEET and I'll send the evidence rule.

#AIcopy #AIagents #buildinpublic

A demo website that says 20 years in business when the owner has never said it is worse than no demo at all. Our rule for AI-drafted copy: every claim has a source on the page or it does not ship. Years in business come from their own site, ratings from their Google listing. On 10 October 2026 we checked 6 such claims against live sites and 6 of 6 matched word for word (read 2026-10-10). AI drafts, a person reads. Reply below with the worst thing AI has made up for you.

#agencylife #AIcopy

The tool takes a business's own web address and drafts a ten-page demo from their real photos, reviews and words. The constraint that matters: no claim ships without a source. Each demo carries a Demo, not the live site ribbon and is not indexed. On 10 October 2026, 6 of 6 years-in-business claims matched the live sites (read 2026-10-10). vespio.ai

#AIagents #buildinpublic

I kept adding skills to my AI agent. About 300 skill names now load on every turn, and that is not free.

If you're 18 to 27 and building with Claude Code, more is not better. I merged 30 skills into others and archived 13, taking 302 folders down to 266 (14 Sep 2026).

Save this for your next agent project.

#AIagents #Claude

Comment FLEET

We added a skill every time a workflow repeated. Soon about 300 skill names were loading on every turn. On 14 Sep 2026 we cut 302 folders to 266 by merging near-duplicates and archiving the unused.

For anyone running agents in an agency: what did you add last quarter that you never use?

#AIagents

Reply below

Every skill name is context spent before work begins. We went from 302 skill folders to 266 on 14 Sep 2026 by merging near-duplicates behind flags and archiving what nothing called.

Design rule: one clear trigger per skill, and a picker for the rest.

#AIagents #Claude

vespio.ai

A few years ago I was broke and building sites that never ranked. Now 11 paying clients (read 2026-10-11) are run day to day by John, with AI agents we built. The agents draft, humans send. It broke three times and the rules came from the breaks. I put the rules and the fails in one place.

#aiagents #agencylife #buildinpublic

Comment AUDIT and I'll run a free AI-visibility check on your business.

A few years ago I was broke and building sites that never ranked. Now 11 paying clients (read 2026-10-11) are run day to day by John, with AI agents we built. The agents draft, humans send. It broke three times and the rules came from the breaks. I put the rules and the fails in one place.

#aiagents #agencylife #buildinpublic

Comment AUDIT and I'll run a free AI-visibility check on your business.

A few years ago I was broke and building sites that never ranked. Now 11 paying clients (read 2026-10-11) are run day to day by John, with AI agents we built. The agents draft, humans send. It broke three times and the rules came from the breaks. I put the rules and the fails in one place.

#aiagents #agencylife #buildinpublic

Comment AUDIT and I'll run a free AI-visibility check on your business.

Eleven paying clients (read 2026-10-11), and the person running them day to day is not the owner. Delivery sits with one person; the agents draft reports, audits and logs; every client email is a draft a human sends. The lesson from three breaks: the rules are the product, not the agents. Reply below

#agencyowner #aiagents #localseo

Comment AUDIT for a free AI-visibility check.

We run an agency on agents, and the org chart is three lines: a delivery lead, an owner on oversight, agents that draft. Hooks enforce the rules because memory is what failed. That is the pattern we are building into the platform. vespio.ai

#aiagents #claudecode #buildinpublic

A success message on your form is not proof it works.

For anyone 18 to 27 learning to build for local businesses, here is the check that matters: submit a real test to your own address and see if the email arrives. I learned that a rendered success box proves nothing.

Steps are on the slides. Save this before you build your next form.

#AIagents #Claude

Comment AUDIT and I'll run a free AI-visibility check on your business.

Most agency form bugs are invisible: the page says thanks and no lead is delivered.

Our rule: a form is proven by a received email, not a rendered success message. Test to an internal address, check every field, check the ad click tags came through.

Who on your team owns the test submit before a site goes live?

#localSEO

Comment AUDIT for a free AI-visibility check.

Forms are a pipeline, not a component. Page, worker, mail API, inbox, with a spam check on the front and attribution fields kept all the way through.

The only honest test is the end of the pipe: did the email arrive with every field?

#buildinpublic #Claude

vespio.ai

I ran the same QA on ten client sites in eight days. It found 293 faults and 201 got fixed and verified.

If you're 18 to 27 and learning to build sites, the useful part is that the same few faults showed up almost everywhere: slow fonts, late hero images, oversized pictures. Learn those and you can fix most sites.

Swipe for the five checks you can run on your own site today.

#webdev #AIagents

Comment AUDIT and I'll run a free AI-visibility check on your business.

Ten client sites, one QA process, eight days: 293 faults found, 201 fixed and verified (July 2026).

The pattern matters more than the total. Render-blocking fonts hit five sites. Late-loading hero images were the biggest speed lever on 6 of 10. Security headers were missing on sites where Lighthouse still showed 100, because it does not score them.

Which of those is still sitting on your own site?

#localSEO #sitespeed

Comment AUDIT for a free AI-visibility check.

We run one QA pass over every site and log each fault as a row, found and fixed, so the totals are countable. Ten sites, eight days, 293 found, 201 fixed (July 2026).

The design lesson: the same five faults repeat, so the fixes belong in a shared skill and not in each site.

#AIagents #buildinpublic

vespio.ai

My biggest cold email campaign sent 17,212 emails and got 137 unique replies. That is about 0.8 percent, read from an audit dated 21 September 2026 (read 2026-10-11). It was the biggest, not the best. A campaign of 2,251 sends got 196 unique replies, 8.7 percent. Plan your list from the reply rate, not from hope. Counts only, no names, no reply text.

#coldemail #outreach #buildinpublic

Comment AUDIT and I'll run a free AI-visibility check on your business.

My biggest cold email campaign sent 17,212 emails and got 137 unique replies. That is about 0.8 percent, read from an audit dated 21 September 2026 (read 2026-10-11). It was the biggest, not the best. A campaign of 2,251 sends got 196 unique replies, 8.7 percent. Plan your list from the reply rate, not from hope. Counts only, no names, no reply text.

#coldemail #outreach #buildinpublic

Comment AUDIT and I'll run a free AI-visibility check on your business.

My biggest cold email campaign sent 17,212 emails and got 137 unique replies. That is about 0.8 percent, read from an audit dated 21 September 2026 (read 2026-10-11). It was the biggest, not the best. A campaign of 2,251 sends got 196 unique replies, 8.7 percent. Plan your list from the reply rate, not from hope. Counts only, no names, no reply text.

#coldemail #outreach #buildinpublic

Comment AUDIT and I'll run a free AI-visibility check on your business.

My biggest cold email campaign sent 17,212 emails and got 137 unique replies. That is about 0.8 percent, read from an audit dated 21 September 2026 (read 2026-10-11). It was the biggest, not the best. A campaign of 2,251 sends got 196 unique replies, 8.7 percent. Plan your list from the reply rate, not from hope. Counts only, no names, no reply text.

#coldemail #outreach #buildinpublic

Comment AUDIT and I'll run a free AI-visibility check on your business.

My biggest cold email campaign sent 17,212 emails and got 137 unique replies. That is about 0.8 percent, read from an audit dated 21 September 2026 (read 2026-10-11). It was the biggest, not the best. A campaign of 2,251 sends got 196 unique replies, 8.7 percent. Plan your list from the reply rate, not from hope. Counts only, no names, no reply text.

#coldemail #outreach #buildinpublic

Comment AUDIT and I'll run a free AI-visibility check on your business.

My biggest cold email campaign sent 17,212 emails and got 137 unique replies. That is about 0.8 percent, read from an audit dated 21 September 2026 (read 2026-10-11). It was the biggest, not the best. A campaign of 2,251 sends got 196 unique replies, 8.7 percent. Plan your list from the reply rate, not from hope. Counts only, no names, no reply text.

#coldemail #outreach #buildinpublic

Comment AUDIT and I'll run a free AI-visibility check on your business.

My biggest cold email campaign sent 17,212 emails and got 137 unique replies. That is about 0.8 percent, read from an audit dated 21 September 2026 (read 2026-10-11). It was the biggest, not the best. A campaign of 2,251 sends got 196 unique replies, 8.7 percent. Plan your list from the reply rate, not from hope. Counts only, no names, no reply text.

#coldemail #outreach #buildinpublic

Comment AUDIT and I'll run a free AI-visibility check on your business.

My biggest cold email campaign sent 17,212 emails and got 137 unique replies. That is about 0.8 percent, read from an audit dated 21 September 2026 (read 2026-10-11). It was the biggest, not the best. A campaign of 2,251 sends got 196 unique replies, 8.7 percent. Plan your list from the reply rate, not from hope. Counts only, no names, no reply text.

#coldemail #outreach #buildinpublic

Comment AUDIT and I'll run a free AI-visibility check on your business.

Our biggest old outreach campaign: 17,212 sent, 137 unique replies, about 0.8 percent. A 2,251-send campaign in the same audit got 196 unique replies, 8.7 percent. Across 522 campaigns and 324,901 sends, 2,115 unique replies, an upper bound because it is summed per campaign (audit dated 21 September 2026, read 2026-10-11). Volume is not the lever. List quality and size planned from the reply rate is. Reply below with the reply rate you actually see.

#agencylife #coldemail

Comment AUDIT for a free AI-visibility check.

The honest baseline before we change outreach: 522 campaigns, 324,901 sends, 6.51 unique replies per 1,000 sends (audit dated 21 September 2026, read 2026-10-11). The biggest campaign reached 0.8 percent. The best at 1,000+ sends reached 8.7 percent. We measure first, then build. vespio.ai

#coldemail #buildinpublic

Nine of my scheduled jobs were failing, and nothing told me. My own health report did, on 5 Oct 2026.

If you're 18 to 27 and building online with AI agents, set up the automation, then set up the thing that checks the automation. A job that dies quietly looks exactly like one that works.

Save this for your next agent project.

#AIagents #Claude

Comment FLEET

Our health report showed 9 scheduled jobs failing on 5 Oct 2026. None of them had alerted anyone.

The lesson for an agency owner: every automated job needs an owner and a check that does not depend on the job reporting itself. Silence is not a green light.

Who checks your automations, and when did they last look?

#AIagents

Reply below

Scheduled agents fail in two ways: loudly with a non-zero exit, or quietly by going silent. Our report on 5 Oct 2026 showed 9 FAIL and 1 STALE.

The design rule: a watcher compares last run to expected schedule, and reads the log before labelling a failure.

#AIagents #Claude

vespio.ai

I ran five AI coding windows at once on one laptop. By morning six things had broken.

If you're 18 to 27 and learning to build with agents, this is what nobody tells you about running more than one: they share a git index, a rate limit and a bill. One commit swept up other windows' files. One limit killed 3 of 11 sub-agents.

Swipe for all six and the rules I now run by.

#AIagents #Claude

Comment FLEET

I ran five AI coding windows at once on one laptop. By morning six things had broken.

If you're 18 to 27 and learning to build with agents, this is what nobody tells you about running more than one: they share a git index, a rate limit and a bill. One commit swept up other windows' files. One limit killed 3 of 11 sub-agents.

Swipe for all six and the rules I now run by.

#AIagents #Claude

Comment FLEET

I ran five AI coding windows at once on one laptop. By morning six things had broken.

If you're 18 to 27 and learning to build with agents, this is what nobody tells you about running more than one: they share a git index, a rate limit and a bill. One commit swept up other windows' files. One limit killed 3 of 11 sub-agents.

Swipe for all six and the rules I now run by.

#AIagents #Claude

Comment FLEET

One night, five AI windows, one shared checkout. Six things broke, and none of them were about the quality of the work. They were about what the windows shared.

A commit swept other windows' files. A rate limit killed 3 of 11 sub-agents. API credits ran out four times between July and September 2026 and every job still printed exit 0.

If your team runs agents in parallel, which of those is already true for you?

#AIagents

Reply below

Parallel agents fail on shared state, not on intelligence. Our five-window night (11 to 14 Sep 2026) showed it six ways: a shared git index, a shared rate limit, a shared balance, silent exit 0 on empty credits.

The fixes are structural: pathspec commits, worktrees, caps per vendor, loud failure, one dispatcher.

#AIagents #buildinpublic

vespio.ai

One text, sent in under a minute after a form fill, kept a client paying us for 2.5 years. Form fill texts the owner, the customer gets a text, one follow-up if nobody replies. Three texts, no more. Cheap to set up and no ad spend.

#localbusiness #automation #sms

Comment AUDIT and I'll run a free AI-visibility check on your business.

One text, sent in under a minute after a form fill, kept a client paying us for 2.5 years. Form fill texts the owner, the customer gets a text, one follow-up if nobody replies. Three texts, no more. Cheap to set up and no ad spend.

#localbusiness #automation #sms

Comment AUDIT and I'll run a free AI-visibility check on your business.

One text, sent in under a minute after a form fill, kept a client paying us for 2.5 years. Form fill texts the owner, the customer gets a text, one follow-up if nobody replies. Three texts, no more. Cheap to set up and no ad spend.

#localbusiness #automation #sms

Comment AUDIT and I'll run a free AI-visibility check on your business.

Local owners are on a roof or a truck when the form email lands, and it sits for hours. The first business to reply usually gets the job. Text the owner, text the customer, follow up once. Set quiet hours and dedupe first, and in the US register your texting. Reply below

#localseo #leadgen #agencyowner

Comment AUDIT for a free AI-visibility check.

Speed to lead is a small automation: form event in, two texts out inside a minute, one follow-up on a timer, dedupe so one lead is not three texts. The US texting registration is the step builders forget. vespio.ai

#automation #sms #buildinpublic

Two websites I know of went live with a placeholder where the tracking code should have been. Months of nothing recorded, and nobody knew.

If you're 18 to 27 and building sites or running a side project online, do the two-minute test on slide 4 before you launch. The tag being in your page source proves nothing.

#analytics #GA4

Comment AUDIT and I'll run a free AI-visibility check on your business.

Two sites went live with a placeholder where the analytics tag belonged, and months went by with nothing recorded. The tag was "installed". It was never firing.

We now prove it before launch: debug view open, own page loaded, hit seen on the preview link. Then one number every Monday: calls from the Google Business Profile.

When did your team last watch a hit land?

#GA4

Comment AUDIT for a free AI-visibility check.

Installed is not working. When you build sites for others, measurement is part of the build, so the launch check is a live one: open debug view, load the preview, watch the hit land.

Source inspection is a hint, not proof.

#analytics #GA4

vespio.ai

Everyone talks about AI agents spending your money. I gave mine access to ads and then made it impossible for it to spend. Any ad it creates starts paused, a script refuses a daily budget over 50 dollars, and it refuses every delete. The self-test passes 11 of 11 (read 2026-10-11). Agents draft, I press the button. That is the whole idea, and it took one afternoon to set up. If you are 18 to 27 and learning to build with AI, learn the guard before you learn the tool. Comment FLEET and I'll send the agent guardrails list.

#AIagents #metaads #buildinpublic

The cost of waiting for the owner is 30 seconds. The cost of a wrong live ad is a month. So in our agency, any ad an AI agent creates starts paused, and a script refuses anything else: no live status, no daily budget over 50 dollars, no deletes. The agent drafts. A person presses the button. If you let tools touch client money, write the refusal before you write the feature. Reply below with the one thing you would never let an agent do without your say.

#agencylife #AIagents

Guardrails are a feature, not a limit. The ads tool we build around refuses three things before a request is ever sent: anything not paused, a daily budget over 50 dollars, and any delete. Its self-test runs 11 cases and passed 11 of 11 (read 2026-10-11). Safe default state, refusal in code, human click. vespio.ai

#AIagents #buildinpublic

Installed is not the same as working. I learned that the expensive way, so now every client gets the same five sources and one rule: watch the tag fire, do not trust the snippet.

#analytics #localseo #googlebusiness

Comment AUDIT and I'll run a free AI-visibility check on your business.

Five sources, one place to look: Google Analytics, Search Console, Business Profile performance, a maps scan and directories. Prove the tag fires in the live debug view, keep name, address and phone identical everywhere, and read one number every Monday. Reply below

#localseo #analytics #agencyowner

Comment AUDIT for a free AI-visibility check.

A tag in the source is not a tag that fires. Our launch check opens the debug view, loads the page and waits for the event to land before anything is called tracked. The five sources are the minimum we wire up. vespio.ai

#analytics #buildinpublic #claudecode

I made two wrong claims about client websites in one night. The cause wasn't carelessness.

If you're 18 to 27 and building a business on top of lots of tools, check this: our site state lived in seven places, got read from four, and no two agreed. Every edit updated whichever file I happened to be looking at.

Swipe for the two-file fix you can copy.

#systems #AIagents

Comment FLEET

I made two wrong claims about client websites in one night. The cause wasn't carelessness.

If you're 18 to 27 and building a business on top of lots of tools, check this: our site state lived in seven places, got read from four, and no two agreed. Every edit updated whichever file I happened to be looking at.

Swipe for the two-file fix you can copy.

#systems #AIagents

Comment FLEET

I made two wrong claims about client websites in one night. The cause wasn't carelessness.

If you're 18 to 27 and building a business on top of lots of tools, check this: our site state lived in seven places, got read from four, and no two agreed. Every edit updated whichever file I happened to be looking at.

Swipe for the two-file fix you can copy.

#systems #AIagents

Comment FLEET

Seven records, four readers, no two in agreement. That was the state of our client website data in September 2026, and it produced two wrong claims in one night.

The fix was one record per fact: the client profile for business facts, one registry for the live tree and flip state. Everything else is a dated or generated view, with a check against the live site.

How many places does your client data live?

#agencylife

Reply below

State drifts when it has no single owner. We found site state in seven places, read from four, none agreeing (audit of 2026-09-15).

The design: two truth files, generated views, a drift check against the live site, every reader repointed. No hand edits to derived fields.

#AIagents #buildinpublic

vespio.ai

I looked at my own lead database and had to admit something. It had 152,382 businesses in it. Only 3,418 had a verified email (summary dated 2026-09-14).

If you're 18 to 27 and building something online, this is the number to check before you plan any outreach. A big list feels like a big pipeline. Count the verified column before you believe it.

Swipe for the funnel and a five-step count you can run on your own list today.

#leadgen #AIagents

Comment AUDIT and I'll run a free AI-visibility check on your business.

We hold a lead database of 152,382 businesses. Only 3,418 of them have a verified email (summary dated 2026-09-14).

The list was never the asset. The verified column was. Phone and website coverage were far wider, so the channel mix has to follow the data, not the headline count.

Count rows, businesses, any email and verified email on your own list. Four numbers, ten minutes.

#coldemail #leadgen

Comment AUDIT for a free AI-visibility check.

A lead database is a funnel, and every stage drops rows. Ours went from 316,751 rows to 152,382 businesses to 12,996 with any email to 3,418 verified (summary dated 2026-09-14).

We now treat verified email as its own field in the schema, never as a flag on the row. Build the count into the pipeline so the gap is visible on day one.

#AIagents #buildinpublic

vespio.ai

123,384 characters of rules, 18,737 of instructions and 8,676 of memory load before I type anything. I measured that on 26 Aug 2026.

If you're 18 to 27 and building with Claude Code, count what loads first. It shares a window with your actual work.

Save this for your next agent project.

#AIagents #Claude

Comment FLEET

Before my agent does any work, 123,384 characters of rules, 18,737 of instructions and 8,676 of memory load into the window (measured 26 Aug 2026).

My estimate of the space left was wrong. The count was not. If your team runs agents, has anyone counted what yours loads?

#AIagents

Reply below

Your setup is part of the context budget. Our measured baseline on 26 Aug 2026: 123,384 characters of rules, 18,737 of instructions, 8,676 of memory, plus 21,493 when one large skill is called.

Design rule: measure the window, never estimate it.

#AIagents #Claude

vespio.ai

Every session starts with about 26,000 words of instructions the model reads before I type a thing.

That is a lot to carry. The image shows the count and how I plan to cut it without dropping a guard.

Curious what yours loads? Copy it.

Comment FLEET

Before anyone types, our AI workspace loads about 26,000 words of instructions. We reviewed all of it and proposed a target of about 5,000 to 5,600 always on, with nothing safety-related deleted.

Do you know how big your setup is?

Reply below

About 26,000 words load before the first prompt. Scope, hook, move, delete.

vespio.ai

Every agent in my setup has one job, one place it reads from and one file it writes. That is the whole structure.

The image is the five steps, and when not to use it.

Running agents in parallel? Copy it.

Comment FLEET

When we run a batch through parallel agents, each one owns one item and one output file. We merge the results ourselves, one at a time.

Do your agents ever write to the same place?

Reply below

One job, one output path per agent. Merge serially.

vespio.ai

A custom website for a stranger takes an agent about three and a half minutes.

It fetches their real site, pulls their logo and photos, builds one page, then checks itself: page height, console errors, broken images.

One rule saves arguments. The live site wins for phone and name. The lead sheet wins for rating and reviews.

#AI #Agents #Outreach

Want the four steps as a checklist? Copy this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

One agent, one lead, about 3.5 minutes and about 160k tokens: fetch the site, pull assets, build a page, self-verify. Plus the rule for when sources disagree.

Which data would you trust first, the site or the listing?

Comment AUDIT for a free AI-visibility check.

Agent mockup: fetch, assets, build, self-check. About 3.5 minutes per lead.

vespio.ai

My AI agents write my emails. They never send them.

Every draft is a file marked do-not-send, a human reads it, and a human presses send. Ads get created paused too.

It costs thirty seconds. One wrong auto-send can cost a client.

Building agents that touch customers? Copy this rule.

Comment FLEET

Agents draft our client messages. A person reviews and sends from their own inbox. The draft file is marked do-not-send, and agents never wire a campaign to real recipients or auto-reply.

Honest limit: this adds a review step on purpose.

Where does your team let automation talk to clients?

Reply below

The human-in-the-loop rule for outbound: draft files with a do-not-send header, named reviewers, ads created paused, no send paths for agents.

vespio.ai

Run two agents on the same file and one quietly undoes the other. This is the rule I run instead.

Give every agent its own output path before you send them.

Running more than one agent? Copy it.

Comment FLEET

Our rule for parallel agents: no two ever edit the same file. If they have to share one, it is serial work.

Have you hit this with your own agents?

Reply below

No two agents edit the same file. Partition the outputs first.

vespio.ai

I use AI all day and I still pick up the phone to every client.

The AI drafts. I review and send. And when a client lapsed, it was a phone call that won them back, not a message.

Some things stay human on purpose.

Want my draft-then-send rule? Copy this.

#Clients #AI #Agency

Comment AUDIT and I'll run a free AI-visibility check on your business.

AI drafts our client messages, a person reviews and sends them. When a retainer client lapsed, a phone call is what brought them back.

Where do you keep the human in your client contact?

Comment AUDIT for a free AI-visibility check.

AI for drafts and checks, a person for the send and the call. One lapsed client came back after a phone call.

vespio.ai

You want traffic to an offer you do not own. Here is the lean AI setup for that.

Read the offer terms first, pick one channel, let an assistant draft the content and tag every link so you know where each visit came from.

Lean beats a pile of tools.

Want the short version? Copy this.

#AI #Affiliate #Traffic

Comment FLEET

For people sending traffic to someone else's offer: one channel, one assistant, tagged links. Skip the pile of tools.

Which channel would you start on?

Reply below

The lean AI setup for traffic plays: read the terms, pick one channel, tag every link.

vespio.ai

Your team does it by hand. Here is where agents take over first.

Start with the repeat task, write down what done looks like, and let an agent draft it. Add a guard that can stop a bad result, and keep a person on the final call.

Sending to clients stays human.

Want the handover order? Copy this.

#AI #Agency #Automation

Comment AUDIT and I'll run a free AI-visibility check on your business.

How we move work from hand to agents: repeat tasks first, written rules, a guard that can stop a bad result, and a person on every client send.

Which task would you hand over first?

Comment AUDIT for a free AI-visibility check.

Agency, manual team: hand over repeat tasks first, add a guard, keep people on sends, pricing and sign-off.

vespio.ai

You do not need to know what is inside a phone to make a call. Same with AI.

Start with one task and one assistant, and read every answer. Add more only when a job repeats.

The biggest setup is not the right one. The one that fits your stage is.

Want the stage-by-stage version? Copy this and see where you sit.

#AI #SmallBusiness #Beginners

Comment FLEET

Owners ask what AI setup they need. The answer depends on the stage: a beginner starts with one assistant on one task, an owner automates a few repeat jobs, an agency builds written rules and a person who checks the output.

Which stage are you at?

Reply below

Match the AI setup to the stage. Beginner: one assistant, one task. Owner: a few repeat jobs. Agency: a system with checks and a person reviewing output.

vespio.ai

You run a business and you have never touched code. You still can use AI.

Pick one job that repeats, use one assistant in plain language, and check every answer before it goes anywhere.

Add a second job only when the first one is steady.

Want the starter version? Copy this and keep it small.

#AI #SmallBusiness #Owners

Comment AUDIT and I'll run a free AI-visibility check on your business.

For owners who do not code: pick one repeat job, use one assistant, check the answer, then add another job when you are ready. Small first, on purpose.

What is the one job you would hand over first?

Comment AUDIT for a free AI-visibility check.

A starter AI setup for a non-technical owner: one job, one assistant, a check before anything goes out.

vespio.ai

I wrote three build plans in one day. All three tools already existed. Now I search before I build.

The image is the habit.

Comment FLEET

Three separate 'we should build X' conclusions in one planning session, and all three already existed. Our rule now: search the capabilities list first, extend before you add.

What did you rebuild last quarter that you already had?

Reply below

Search before you build. Extend what exists, build only when the search comes back empty.

vespio.ai

AI answers like to quote short, complete answers.

We write every money page with one block that gives the answer first, names the subject, carries a fact, and runs 134 to 167 words.

Writing your own pages? Save this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

A block meant to be cited does three things: opens with the direct answer, names its subject so it stands alone, and holds at least one fact. We aim for 134 to 167 words. That gives a page a better chance of being cited.

What does your top page open with?

Comment AUDIT for a free AI-visibility check.

Citable passage spec: 134 to 167 words, answer-first, self-contained, one fact, no leading conjunction.

vespio.ai

If your business texts never arrive, it is usually a registration issue, not your phone.

The order: profile approved, brand, campaign, numbers added. And every text must read as one sender.

Texting customers? Save this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

When business texts are not delivered, the cause is often registration. The order that works: business profile approved, brand, campaign, then numbers added to a messaging service. Every message has to read as one sender.

Are your customer texts registered?

Comment AUDIT for a free AI-visibility check.

Delivery starts at registration: profile, brand, campaign, messaging service. One sender per campaign, or it gets flagged.

vespio.ai

I do not run one model for everything. The cheapest thing that can decide, decides. Bash first, then small models, and the top tier only for the decisions that matter.

Comment FLEET

Our model routing rule: the cheapest tier that can decide, decides, and blast radius picks the tier, not task size. A check that can be a script is a script.

How do you decide which AI model does which job?

Reply below

Route by blast radius: script, small model, mid model, top model. Recon, decide, edit, verify.

vespio.ai

Before you tell a client bad news, check the measurement. One search pull read 30 to 80% fewer top-10 keywords for every client. It was an off day.

The image is the check.

Reporting numbers to clients? Copy it.

Comment AUDIT and I'll run a free AI-visibility check on your business.

One keyword pull told us every client had lost positions. A live re-query said otherwise. Now a pull is only a baseline after a sample is re-checked.

How do you validate a number before you report it?

Comment AUDIT for a free AI-visibility check.

Re-query a sample before a pull becomes a baseline.

vespio.ai

Big inbox providers changed the rules for bulk senders. Here is the check I run on every domain first.

Ten reputation lists, one score. If a check errors, I do not call it clean.

Then the spam word scan on step 1.

Want the checklist? Copy this.

#ColdEmail #Deliverability #Outreach

Comment AUDIT and I'll run a free AI-visibility check on your business.

Before any campaign goes out we check the sending domain against 10 reputation lists, scan step 1 for spam trigger words, and look at inbox warmup. An error on a check is never counted as clean.

What do you check first?

Comment AUDIT for a free AI-visibility check.

Pre-send checks as a routine: domain reputation across 10 lists, spam word scan, warmup status. Errors are reported, not hidden.

vespio.ai

One setting in Cloudflare decides whether ChatGPT's crawler can read your site. Most owners have never seen it.

Check the setting, then test it: request your site as GPTBot and expect a 200.

Save this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

If AI crawlers cannot read your pages, you lose a chance of being cited. A Cloudflare AI bot setting can block them without any other symptom, and its default is changing.

We read the setting and also test with GPTBot, OAI-SearchBot, PerplexityBot and ClaudeBot.

Who checked yours?

Comment AUDIT for a free AI-visibility check.

AI crawler access as a gate: zone setting read, then a user-agent test per crawler that must return 200.

vespio.ai

Your cold emails are too long.

Keep the body under 100 words, open with something specific about them, and ask for one thing. Three steps at most, then stop.

The rule list is short. Most people skip it.

Want the checklist? Copy this.

#ColdEmail #Outreach #Sales

Comment AUDIT and I'll run a free AI-visibility check on your business.

Our cold email rules: under 100 words, a specific opening observation, one soft reply ask, no images in the first touch, and a maximum of three steps.

How long is your first email?

Comment AUDIT for a free AI-visibility check.

Cold email as a checklist: length, one ask, no spam trigger words, three steps maximum.

vespio.ai

If your script calls an AI API, put a gate in front of it. Estimate the cost, set a hard cap, run, then log what it really spent.

We made it a rule after a bad month, and a hook now blocks any script that skips it. The image is the whole flow.

Building with AI? Copy it.

Comment FLEET

Any script that calls a paid AI API has to estimate its cost first, run under a hard cap, and log the real spend afterwards. A hook blocks scripts that skip this, so it does not depend on anyone remembering.

Honest limit: the estimate is only as good as our per-call numbers, so we compare it against the ledger.

Who owns the AI spend cap in your business?

Reply below

Any script that calls a paid AI API has to estimate its cost first, run under a hard cap, and log the real spend afterwards. A hook blocks scripts that skip this, so it does not depend on anyone remembering.

Honest limit: the estimate is only as good as our per-call numbers, so we compare it against the ledger.

Who owns the AI spend cap in your business?

Reply below

Any script that calls a paid AI API has to estimate its cost first, run under a hard cap, and log the real spend afterwards. A hook blocks scripts that skip this, so it does not depend on anyone remembering.

Honest limit: the estimate is only as good as our per-call numbers, so we compare it against the ledger.

Who owns the AI spend cap in your business?

Reply below

The cost gate: estimate, cap, run, ledger. Enforced by a hook rather than by memory, and reconciled against the vendor console after any spike.

vespio.ai

Launch eight agents and seven come back. The count check is how you notice the missing one.

The image is the five steps.

Sending agents out in a batch? Copy it.

Comment FLEET

After every batch of parallel agents we count the outputs against the number we sent. A missing file is a silent failure.

Do you check what did not come back?

Reply below

Count the outputs. A missing one is a silent failure. Re-send only that item.

vespio.ai

You moved a site and traffic dipped. Do not guess why.

There are three usual killers: redirects that were drafted but never deployed, pages that got merged away, and junk links that showed up off-site. Each one has a tell you can check in ten minutes.

Save this before your next migration.

Comment AUDIT and I'll run a free AI-visibility check on your business.

After a site move, the instinct is to blame one thing. We check three: redirects, page count, and off-site links. A ranked URL that returns 404 instead of 301 is the first tell. A new sitemap far smaller than the old one is the second.

Honest limit: recovery is slow and we never promise a timeline.

What was the cause the last time you saw a drop after a migration?

Comment AUDIT for a free AI-visibility check.

A migration post-mortem in one picture: three failure modes, three tells, three fixes. Redirect parity, page parity, link hygiene.

vespio.ai

This is my daily brief. Everything that does not need me, the agents take. This is what is left.

One command reads every lane and writes the day. Another clears the safe rows. Sending, pricing and DNS stay with me.

Want to see how it is built? Copy this.

#AI #Productivity #Agents

Comment FLEET

How our day starts: one command reads every source and writes the plan; agents clear the safe rows; sends, pricing and DNS stay with a person.

What would you hand to an agent first?

Reply below

A daily brief: agents read and prepare, people send, decide and sign off.

vespio.ai

Four ways to do your marketing. One honest table.

DIY costs your time. A freelancer brings one skill. An agency brings a team. Software repeats a task but still needs you.

Cheapest is whichever one you are missing.

Want the table as a checklist? Copy this.

#SmallBusiness #Marketing #LocalBusiness

Comment AUDIT and I'll run a free AI-visibility check on your business.

For a local owner choosing how to run marketing: DIY costs time, a freelancer brings one skill set, an agency brings a team, software repeats a task but needs set up. Choose by what you lack: time, a skill, or a team.

Which one are you using today?

Comment AUDIT for a free AI-visibility check.

DIY, freelancer, agency or software. A plain comparison, then a rule: pick by what you are missing.

vespio.ai

Some agencies register your domain in their own name. Here is how to check who owns yours.

Run an RDAP lookup, read the registrant, and check you can log in to the account. If it is not in your name, ask for a transfer.

It takes a few minutes.

Want the checklist? Copy this.

#Domains #SmallBusiness #Websites

Comment AUDIT and I'll run a free AI-visibility check on your business.

Who owns your domain? Look it up with RDAP, read the registrant and make sure you can log in. If it is held by someone else, ask for a transfer.

When did you last check?

Comment AUDIT for a free AI-visibility check.

Check your domain owner: RDAP lookup, read the registrant, confirm you hold the login.

vespio.ai

Adding 200 city pages can hurt you.

If every page shares an opener and the same headings, it is a doorway family. Each city page needs copy that no sibling shares and at least three facts about that city.

Save this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

More pages is not always more traffic. Google reads the shape repeated across a site: the same opener, the same headings. Our check fails a family when openers repeat on 10% of pages or one heading order covers 25%.

How different are your city pages?

Comment AUDIT for a free AI-visibility check.

Doorway check as a gate: page_footprint_check fails shared openers at 10% and one H2 sequence at 25%; city pages need 35% unique copy.

vespio.ai

I gave a brief to five agents and got five finished websites.

Each agent gets one lead, one brand kit and a strict brief. No sub-agents, no deploys. Then I count the outputs before the next batch starts.

The rule that keeps it honest: a page that does not scroll to 2250 px does not count.

#AI #Agents #Outreach

Want the brief layout? Copy this.

Comment FLEET

Parallel agent batch for outreach mockups: five per batch, one lead each, outputs counted before the next batch, a minimum page height as the pass mark. 50 mockup folders so far.

Would you trust agents with a build like this?

Reply below

Agent batch for lead mockups: 5 per batch, one lead each, a counted output, a hard page-height check.

vespio.ai

Six agents tripped my limit. Five never failed once.

I run agent builds in batches. Six in one batch hit the session limit. Five did not fail across six batches.

Your limit will differ. Find it by running it, then write the number down.

#AI #Agents #Workflow

Want the batch rules? Copy this.

Comment FLEET

Batch size for parallel agent builds: six trips the session limit, five ran six batches with zero kills. Your own limit will differ.

What batch size works for you?

Reply below

Parallel agents: 6 per batch trips the limit, 5 ran 6 batches with 0 kills.

vespio.ai

You do not need dozens of rules. Start with these five, then add one each time something breaks.

The image is the starter file.

Comment FLEET

Five rules to start an AI rules file: draft never send, no secrets in files, verify before done, one source for client data, paused by default. Then add one rule per failure.

Which rule would you add first?

Reply below

A starter rules file for AI agents, five lines, grown one failure at a time.

vespio.ai

Most website forms send an email and stop.

A better one opens a conversation thread you can work from, so the enquiry has somewhere to go after it lands.

Save this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

A form that only emails you ends the story at the inbox. A form that opens a thread gives the enquiry a place to continue.

Where does an enquiry go after your form sends it?

Comment AUDIT for a free AI-visibility check.

Form notification versus form thread: the lead becomes a conversation object, not a one-off email.

vespio.ai

The form said thank you. The email never arrived.

The only proof a form works is the email landing in an inbox after a real test submit.

Got a website form? Test yours today.

Comment AUDIT and I'll run a free AI-visibility check on your business.

A thank-you message on screen proves nothing. We prove a form with a real test submit to an internal address and a received email, with spam checks on and ad click tags preserved.

When did you last test your contact form?

Comment AUDIT for a free AI-visibility check.

Form delivery proof: test submit, spam check, Worker, Resend, receipt in the inbox. A rendered success message does not count.

vespio.ai

A homeowner fills in your form at 11pm. Who replies first?

The order: form, text, call, follow-up. About 20 minutes to set up.

Trade business? Save this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

Speed to lead is a sequence: the form fires, the customer gets a text, the owner is alerted, a call happens, a follow-up is queued. It takes about 20 minutes to set up.

What happens after your form is submitted at night?

Comment AUDIT for a free AI-visibility check.

Speed-to-lead chain: form event, customer SMS, owner alert, call, queued follow-up.

vespio.ai

Before you flip a live site to a new build, plan four lanes: do not lose what ranks, do not break the contact form, match what the old site had then beat it, and verify before you flip once.

Most migrations that go wrong skip one of these.

Save it for your next launch.

Comment AUDIT and I'll run a free AI-visibility check on your business.

Our switch plans always cover four lanes: keep what ranks, keep the lead form working, match then beat the existing footprint, and verify everything before a single flip.

Honest limit: the plan reduces the risk; it does not remove it.

Which lane does your team skip most often?

Comment AUDIT for a free AI-visibility check.

Cutover planning, simplified: preserve rankings, preserve conversion, match then exceed the footprint, verify then flip once.

vespio.ai

A new website goes live and the calls dip. Most owners find out weeks later.

We treat launch day as week zero and check the same list every week for four weeks: old URLs, sitemap, analytics, AI crawler access, rankings against the baseline.

Running a site for a local business? Save this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

Launch day is the start of the check, not the end. For four weeks after a site move we re-test redirects, the sitemap in Search Console, analytics sessions on the live host, AI crawler access and rankings against the baseline.

When did anyone last check your old pages still land somewhere?

Comment AUDIT for a free AI-visibility check.

Post-launch watch as a weekly job: URL parity, sitemap swap, GA4 sessions on the live host, AI crawler access, ranking diff vs the flip-day baseline.

vespio.ai

Your analytics can be installed and record nothing. We found a site that logged zero for 102 days because a placeholder ID shipped instead of the real one.

The fix is a 30 second check: open Realtime, visit your own site, watch yourself appear.

Running a site for a local business? Save this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

A tracking snippet in the page source is not proof that tracking works. In an audit on 22 June 2026 we found a site that had recorded nothing for 102 days because a placeholder ID went live.

Before any site move we now open Realtime on the preview and watch our own visit land.

When did you last see your own visit in your analytics?

Comment AUDIT for a free AI-visibility check.

Tag present is not tag firing. The gate: open GA4 Realtime or DebugView against the preview, see the hit, then ship.

vespio.ai

Most local business owners have never audited their own Google profile.

It takes about ten minutes: reviews, replies, categories, photos, posts, Q&A, hours, services.

Running a local business? Save this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

A Google Business Profile audit is a short list: reviews and replies, categories, photos, recent posts, Q&A, hours including holiday hours, services, and the website and booking links.

Which of those is the gap on your profile?

Comment AUDIT for a free AI-visibility check.

GBP audit as a checklist: review velocity, reply rate, categories, photos, posts, Q&A, hours, services, links. Scored, then fixed in order.

vespio.ai

The category you pick on Google decides which searches you can show up for. Movers get this wrong all the time.

Pick the primary on purpose, add extras for the other services, and check it again.

Local business? Save this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

On a Google Business Profile the category decides which searches you can show for. Our profile audit always checks the primary and the extra categories.

Do you know which category your profile uses?

Comment AUDIT for a free AI-visibility check.

Category selection as a profile setting that gates search surface. The audit reads primary and secondary.

vespio.ai

My deploy gate log said 71 fail-open, 29 pass, 18 block. A green deploy is not proof the checks ran.

The image is how to read it.

Relying on a gate? Copy it.

Comment FLEET

Our deploy gate failed open 71 times, passed 29 and blocked 18 in one stretch. The exit code looked green either way. The log is the proof.

Do you check what your gates actually did?

Reply below

A green deploy is not proof the checks ran. Read the gate log.

vespio.ai

Before you move a website, export one report from Search Console: Pages. Every URL on it must still load, or redirect properly.

Skip it and you can lose pages that were quietly sending you customers.

Moving a site soon? Save this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

Before any site move we export Search Console Performance by Pages and check every URL that gets traffic against the new build. Each one must load, or redirect to a matching page. A sitemap check alone misses pages that people actually land on.

Do you know which pages bring you calls?

Comment AUDIT for a free AI-visibility check.

The pre-flip reconcile: GSC Pages export vs the new build, every trafficked URL a 200 or a verified 301, zero 404s. A sitemap sweep is not enough.

vespio.ai

Rules in a doc get forgotten. Rules in a hook do not.

Every risky thing my agents can do now passes a script first: paid API calls, deploys, domain changes, even deleting files from my own folders. If the gate says no, the command does not run.

Pin this if you let AI touch real systems.

Comment FLEET

We stopped relying on people and agents to remember the rules. The risky actions are guarded by hooks: spend, deploys, domain changes, and style checks on client documents. A log records what each gate did.

Honest limit: a green deploy still is not proof the checks ran, so we read the log.

Which of your rules only exist in a document?

Reply below

Guardrails as code: pre-command hooks for cost, deploy, DNS, style and file safety, each with a named override and a log of what it did.

vespio.ai

Cloudflare Pages gives you 100 projects and no paid plan lifts it. I host every preview and audit inside that limit with one trick: one project per purpose, many files inside.

Comment FLEET

Cloudflare Pages caps an account at 100 projects. We run one project per purpose with unlimited files inside, so a new prospect is a subpath, never a new project.

How do you organise your hosting as you scale?

Reply below

One project per purpose, many files inside. How to live under a hard hosting cap.

vespio.ai

Most lead forms lose leads quietly. The page says thanks, nothing arrives.

Our default is boring on purpose: form, bot check, one Worker, one email. The tracking fields ride along so you know where the lead came from. And a form only counts as working when the test email lands.

If you build sites for a living, steal the shape in the image.

Comment AUDIT and I'll run a free AI-visibility check on your business.

Every form on a site we build ships with a bot check and carries the tracking fields, then posts to one endpoint that sends an email. We do not call a form working because a thank-you message showed up. We call it working when the test email arrives.

Honest limit: this is our default. Some sites add their own CRM on top.

How do you test your own contact form?

Comment AUDIT for a free AI-visibility check.

Every form on a site we build ships with a bot check and carries the tracking fields, then posts to one endpoint that sends an email. We do not call a form working because a thank-you message showed up. We call it working when the test email arrives.

Honest limit: this is our default. Some sites add their own CRM on top.

How do you test your own contact form?

Comment AUDIT for a free AI-visibility check.

Every form on a site we build ships with a bot check and carries the tracking fields, then posts to one endpoint that sends an email. We do not call a form working because a thank-you message showed up. We call it working when the test email arrives.

Honest limit: this is our default. Some sites add their own CRM on top.

How do you test your own contact form?

Comment AUDIT for a free AI-visibility check.

The default lead path: form, anti-spam, one Worker endpoint, one email, attribution fields preserved. A rendered success state is not proof; a received email is.

vespio.ai

Most people check profile views. The more useful read is where those views came from: Search or Maps, phone or computer.

The split shows how people find you.

Own a local business? Save this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

Profile views are one number. The performance tab also splits them into Search and Maps, on phone and on computer. That split shows how customers found the listing, and where it is thin.

Have you looked at yours?

Comment AUDIT for a free AI-visibility check.

Profile performance, four slices: search mobile, search desktop, maps mobile, maps desktop. The split is the signal, not the total.

vespio.ai

Month end used to mean a weekend. Now it is render, verify, review.

Data in by API. Refresh the diagnosis. Render and verify each report. Diff against what we already sent. Draft the email.

A person reads it before it goes anywhere.

#Reporting #AI #Agency

Want the sequence? Copy this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

Our monthly report close: data in, diagnosis refresh, render and verify, diff against the last sent version, email draft. A person sends.

How do you close the month?

Comment AUDIT for a free AI-visibility check.

Monthly report pipeline: data, diagnosis, render, verify, diff, draft. Human sends.

vespio.ai

Every AI vendor I use has a monthly cap. No cap, no auto-reload. Here is why.

A cap turns a silent failure into a loud one.

Paying for AI by the call? Copy it.

Comment FLEET

Our rule for AI spend: no vendor gets an auto-reload unless a monthly cap is set on it. A cap makes a failure loud instead of silent.

Do your AI vendors have caps?

Reply below

No cap, no auto-reload. A cap turns a silent drop into a loud 429.

vespio.ai

There is money sitting in your old quotes.

Set three follow-ups: day 3, day 14, day 30. Same sequence for every quote, then stop.

Send one tonight? Save this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

A quote that went quiet is not always a no. A fixed sequence of three messages, at day 3, day 14 and day 30, gives every old quote the same chance.

How many quotes are sitting unanswered in your inbox?

Comment AUDIT for a free AI-visibility check.

Dead-quote sequence as a CRM automation: three messages at 3, 14 and 30 days.

vespio.ai

There is a point where a personalised email gets creepy. Here is the line.

Use one fact from their own website or public listing, about the business. Never invent a fact and never use anything private.

One public fact is enough to show you looked.

Want the rule? Copy this.

#ColdEmail #Outreach #Personalisation

Comment AUDIT and I'll run a free AI-visibility check on your business.

Our personalisation rule: one specific fact about the business, taken from a public source, never invented. Nothing private.

Where do you draw the line?

Comment AUDIT for a free AI-visibility check.

Personalise with one public business fact. Never invent, never use private details.

vespio.ai

A custom video per lead sounds expensive. Mine costs nothing per lead.

One generic video. Only the first 11.5 seconds gets re-made, with the lead's own website scrolling in. The rest is copied as is.

Honest limit: uploading is still manual.

Want the pipeline outline? Copy this.

#Outreach #Video #Automation

Comment AUDIT and I'll run a free AI-visibility check on your business.

A personalised video per prospect without a per-lead cost: one master recording, a short intro rebuilt with each prospect's site, the body copied unchanged. Built on local tools. Tested on synthetic assets; upload is still manual.

Would a video like this change your outreach?

Comment AUDIT for a free AI-visibility check.

Personalised video at no per-lead cost: re-render 11.5 seconds, copy the rest. ffmpeg and Playwright, local. Tested on synthetic assets.

vespio.ai

I audited 73 of my plans. 35 said active and were stale.

The check is simple: read the status, count the open rows, look at the last git touch, flag the rest.

Status is what you wrote once. The log is what happened.

Want the audit method? Copy this.

#Productivity #AI #Systems

Comment FLEET

Plan audit, read 2026-10-10: 73 plans checked, 35 flagged stale, 17 without a continue file, 5 gated on a person. We trust the folder history over the status header.

How do you check that work is still alive?

Reply below

Status drift check: compare the status header with open rows and the last git touch. In one audit, 35 of 73 plans were stale. Read 2026-10-10.

vespio.ai

Pricing by search position rewards the wrong things. Here is how we price instead.

A local package has a clear list of work and a monthly price. Each month there is a report of what was done.

No promise about results, because nobody controls them.

Want the framing? Copy this.

#Pricing #LocalBusiness #Agency

Comment AUDIT and I'll run a free AI-visibility check on your business.

We price local packages by the month for a clear list of work, with a report each month. We do not price by search position and we do not promise results.

How does your agency price?

Comment AUDIT for a free AI-visibility check.

Price the work, report every month, and promise nothing nobody controls.

vespio.ai

If removing a line would not cause a mistake, delete the line. That is the whole test for pruning CLAUDE.md.

The image is the checklist.

Has your CLAUDE.md grown? Copy it.

Comment FLEET

We reviewed our own instruction files. The test: would removing this line cause a mistake? If not, it goes. Anything that must happen every time becomes a hook.

When did you last prune yours?

Reply below

Cut any line that would not cause a mistake. Make must-happen rules hooks.

vespio.ai

Watch me audit a stranger's business in five minutes. One URL in, a 14-slide deck out.

The image is the flow and when to use the deeper audit instead.

Want the checklist? Copy it.

Comment AUDIT and I'll run a free AI-visibility check on your business.

Before a first call we run a quick audit: one URL in, a 14-slide AI visibility deck out in about 3 to 5 minutes.

What do you prepare before a sales call?

Comment AUDIT for a free AI-visibility check.

One URL in, a 14-slide audit deck out in about 3 to 5 minutes.

vespio.ai

A photo that looks clean can still say AI generated in its metadata. And rename it, and the metadata check misses it. The fix is to match the pixels.

The image is the process.

Shipping images for clients? Copy it.

Comment FLEET

We trace shipped site images back to originals that carry an AI credit by comparing pixels, not filenames. A renamed copy cannot hide.

Do you check where your site images came from?

Reply below

Match the pixels, not the filename. A perceptual hash finds the renamed copy.

vespio.ai

All eleven reports rendered fine. All eleven reprinted last month's sentences.

The checks looked at shape, not truth across months. Copy that was true in month one was false in month two.

Now a reviewer reads each live page before it ships, and I fix the engine once, not eleven reports.

#Reporting #AI #QA

Want the review order? Copy this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

Month-two reports reused month-one copy: 'first month on the tracker', a hardcoded profile line, errored AI answers scored as not cited, last month's keywords. Shape checks passed on all 11.

What would your checks miss?

Comment AUDIT for a free AI-visibility check.

Reports can render fine and still be wrong. Read the live page, fix the engine once, re-check.

vespio.ai

Asking for a review takes one text. The trick is making it a template, not something you retype and forget.

First name, the review link, short. One ask per job.

Running a service business? Save this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

We turned the review ask into a template: first name, review link, under about 300 characters, one ask per customer per job, logged. A person sends it, never a script.

Who asks your customers for reviews, and how?

Comment AUDIT for a free AI-visibility check.

Review ask as a pipeline: template from the profile, drafted message, human send, log row. The log is the dedupe.

vespio.ai

I run my business on a searchable second brain. Three vaults, one key the whole team shares.

Two rules make it safe: what a client said and what we said about them are tagged apart and never merged, and every search must report back whether it helped.

That feedback loop is what makes it improve. Want the setup?

Comment FLEET

Our team notes live in a searchable brain split into three vaults. Client words and our own notes about them carry different tags and never mix. People, pay and contract talk is never ingested.

Every search ends with a mark: it helped, it misled, or unsure. That is how the next search gets better.

How does your team find what it already knows?

Reply below

Our team notes live in a searchable brain split into three vaults. Client words and our own notes about them carry different tags and never mix. People, pay and contract talk is never ingested.

Every search ends with a mark: it helped, it misled, or unsure. That is how the next search gets better.

How does your team find what it already knows?

Reply below

Our team notes live in a searchable brain split into three vaults. Client words and our own notes about them carry different tags and never mix. People, pay and contract talk is never ingested.

Every search ends with a mark: it helped, it misled, or unsure. That is how the next search gets better.

How does your team find what it already knows?

Reply below

A RAG design note: three vaults, one shared key, tag-separated chunks (client-comms vs internal-notes), and a mandatory feedback call after every retrieval.

vespio.ai

The sentences our site builder refuses to write.

No invented numbers or years in business. No licence numbers, awards or warranty terms. No prices, crew names or promised outcomes. Nothing negative about the business.

If a field has no evidence, it stays empty and the gap is noted.

#Vespio #AI #Websites

Want the rule list? Copy this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

Writing rules for an AI site builder: evidence only, no invented numbers, years, licences, awards, warranties, prices or promised outcomes. No evidence means an empty field and a noted gap.

Which rule would you add?

Comment AUDIT for a free AI-visibility check.

Site builder rules: evidence only. No evidence means an empty field and a gap noted.

vespio.ai

Your Claude Code skill never runs, and it is almost always the one description line. Lead with a verb, say when it triggers, name what it produces.

Comment FLEET

A skill is only as findable as its description line. Ours lead with a verb, state the trigger and name the output file, in under 200 characters.

Reply below

Skill routing lives in the description. Verb first, trigger conditions, named output, short.

vespio.ai

Most people never write a Claude Code skill because it sounds like programming. It is closer to writing a recipe: purpose, input, steps, limits, when to stop.

The image is the whole structure.

Comment FLEET

A skill is a text file laid out like a recipe: purpose, triggers, input, steps with a check after each, hard rules and when to escalate.

What job in your business is still only in someone's head?

Reply below

The skill file shape: purpose, when to use, input, process with checks, hard rules, escalation.

vespio.ai

Before I write a Claude Code skill I run the job without it and watch it fail. If nothing fails, I do not need the skill.

The image is the five steps.

Building skills? Copy it.

Comment FLEET

Our rule for new skills: run the job without the skill first and record how it fails. A skill whose absence changes nothing is not a skill.

Do you test your automations against a real failure first?

Reply below

RED first for skills: fail it without the skill, record it, write the skill, re-run.

vespio.ai

A slow website does not look slow on your office wifi. It looks slow in a customer's driveway.

You test on fast wifi, on your own computer, on a site you know. They arrive on mobile data, on a phone, for the first time.

Switch wifi off and open your site. Watch the first screen.

#LocalSEO #Website #SmallBusiness

Want a phone test sheet? Copy this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

Speed is a phone problem. Test with wifi off, on mobile data, the way a customer sees your site.

How does your site load with wifi off?

Comment AUDIT for a free AI-visibility check.

Test site speed the way customers see it: wifi off, mobile data, first visit.

vespio.ai

On a phone, tap the number on your website. If nothing happens, you have this fault.

Open your site on your phone. Tap the number. The dialler should open with the number filled in.

If it does not, a customer ready to call has to copy it by hand.

#LocalSEO #Website #SmallBusiness

Want the full phone checklist? Copy this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

Two-minute check: tap the phone number on your own website, on a phone. It should open the dialler.

Have you ever tapped your own number?

Comment AUDIT for a free AI-visibility check.

Check your call button: tap the number on a phone. The dialler should open.

vespio.ai

A guard that has never blocked anything is not a guard. Here is how I test mine: break it first, then prove it lets good input through.

Building with AI agents? Copy it.

Comment FLEET

Every automated check we add has to be seen failing first. Red case, green case, and the pair is kept in a selftest. Without that it does not merge.

How do you know your checks catch anything?

Reply below

No verifier ships without a negative control. Red case, green case, selftest.

vespio.ai

One tag can hide a whole site from Google, and a test site has it on purpose. Forget to remove it at launch and nothing shows up.

There are three places to check: page source, response header, robots.txt.

Building or moving a site? Save this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

Test sites are blocked from search on purpose. The risk is carrying that block to launch. We check three places: the page source, the response header and robots.txt, and switch all three off together.

When did you last check yours?

Comment AUDIT for a free AI-visibility check.

Three noindex guards, flipped together: layout flag, X-Robots-Tag header, robots.txt. Post-launch verify checks all three.

vespio.ai

Every owner has been burned. Here is what I say when they tell me an agency did nothing.

I do not argue. I show a monthly report, walk through what was done, and agree what to check together.

Every number in it has a scope, and a person checks it before it goes out.

Want the answer? Copy this.

#LocalSEO #Agency #ClientTrust

Comment AUDIT and I'll run a free AI-visibility check on your business.

When an owner says an agency did nothing, we show the monthly report: what was done, every number scoped and checked by a person before it ships.

What would you want in that report?

Comment AUDIT for a free AI-visibility check.

Show the monthly report: what was done, every number scoped, checked before it ships.

vespio.ai

Open your own website on your phone and try to open the menu.

Tap the menu. Open each main page from it. That is the whole test.

If the menu does nothing, visitors on a phone cannot get past your home page.

#LocalSEO #Website #SmallBusiness

Want the phone checklist? Copy this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

One-minute check: open your site on a phone, tap the menu, open each main page.

Does your menu pass?

Comment AUDIT for a free AI-visibility check.

Phone check: open the site, tap the menu, open each main page.

vespio.ai

One web address goes in. A ten-page site preview comes out.

It reads their website for pages, photos, logo and colours. It reads their Google listing for rating, hours and reviews. Then it writes the copy from that evidence.

Every preview is marked as a demo.

#Vespio #Websites #AI

Want the field list? Copy this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

A URL in, a ten-page preview out: website for pages, photos, logo and colours; Google listing for rating, hours and reviews. Every page is marked as a demo.

What would you check first on a page like this?

Comment AUDIT for a free AI-visibility check.

Vespio URL-to-site: website plus Google listing in, ten-page demo preview out, marked as a demo.

vespio.ai

Every video should get its own tracking tag. That is how you know which one sent the visit.

Put a utm_content slug on the link, one per video, and read the results by video, not by channel.

Want the setup? Copy this.

#Tracking #UTM #ContentMarketing

Comment AUDIT and I'll run a free AI-visibility check on your business.

Per-video tracking: a utm_content slug on each link tells you which video sent the visit. One shared link tells you nothing.

Do you tag your links?

Comment AUDIT for a free AI-visibility check.

One utm_content slug per video, so each visit is attributable to a video.

vespio.ai

My AI agent kept saying done. Done meant it wrote the thing, not that it checked the thing.

So every claim now has a place it must be checked: a render for looks, a run for works, git log for committed, the live URL for deployed.

If you're building with AI, screenshot this and pin it above your desk.

Comment FLEET

Four words cause most agency rework: done, live, fixed, shipped, said without a check. Each one has a source we verify from. A 200 status is not a render. An exit code is not a live page.

Honest limit: this slows the last mile by minutes, and it is worth it.

Which of those four words do you trust too easily?

Reply below

Four words cause most agency rework: done, live, fixed, shipped, said without a check. Each one has a source we verify from. A 200 status is not a render. An exit code is not a live page.

Honest limit: this slows the last mile by minutes, and it is worth it.

Which of those four words do you trust too easily?

Reply below

Four words cause most agency rework: done, live, fixed, shipped, said without a check. Each one has a source we verify from. A 200 status is not a render. An exit code is not a live page.

Honest limit: this slows the last mile by minutes, and it is worth it.

Which of those four words do you trust too easily?

Reply below

Agents round up. The fix is a lookup table: for every claim an agent can make, the one source it must be checked against before the claim is allowed.

vespio.ai

Before I ring or email anyone, I check that the list is true.

Only addresses that verify as ok or valid go forward. Invalid, catch-all, unknown and anything unrecognised wait in quarantine, and nothing gets deleted.

The check never sends. A person does that.

Want the five minute check? Copy this.

#ColdEmail #Deliverability #Outreach

Comment AUDIT and I'll run a free AI-visibility check on your business.

Our list check: verify every address, keep only ok and valid, quarantine the rest and never delete it. The tool has no send path.

How do you check your lists?

Comment AUDIT for a free AI-visibility check.

List check: verify, keep ok and valid, quarantine the rest. Nothing is sent by the tool.

vespio.ai

A website can load fine and still cost you calls. Before ours ship we check the form, the logo, the footer, the button style and the contact details.

Small things, fixed first.

Building a site? Save this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

Before a local business site ships we run it through a QA probe: form delivery, logo, footer density, call to action style, contact details. Pages that look finished often fail one of them.

When did anyone last check yours?

Comment AUDIT for a free AI-visibility check.

Pre-deploy QA: form pipeline, logo, footer density, CTA style, NAP. Gate first, ship second.

vespio.ai

Here is what I will not do for a client, even if they pay me for it.

I will not invent a number. I will not promise a ranking or a timeline. And my AI does not send a single message to a client; it drafts, and I send.

Boring rules. They are why clients stay.

Want the full rule list? Copy this.

#Agency #LocalSEO #AI

Comment AUDIT and I'll run a free AI-visibility check on your business.

Our limits on client work: no invented metrics, no promised rankings or timelines, no promised AI visibility, and no client message sent by an agent. A person reviews and sends every one.

What is on your own do-not-do list?

Comment AUDIT for a free AI-visibility check.

The refusals built into our delivery: no invented metrics, no ranking or timeline promises, drafts only for client messages, a named person sends.

vespio.ai

Before I touch a local business's marketing I ask one thing: which page sent your last call?

If nobody can answer, it is guesswork. Open GA4 Realtime, tap your own call button, watch for the dot.

Local business owner? Save this.

Comment AUDIT and I'll run a free AI-visibility check on your business.

The first question we ask any local business: which page sent your last phone call? If the answer is a shrug, the next step is a call-tap event you can see in GA4 Realtime.

Can you name the page behind your last call?

Comment AUDIT for a free AI-visibility check.

Call attribution check: call-tap event, seen in GA4 Realtime from your own device, reported by page.

vespio.ai