Build Your Lead-Gen AI Analyst in ChatGPTGA4, Google Ads, Meta and your lead listGA4, Google Ads, Meta and your lead list

Real Leads, Not Reported Leads In-depth guide to building your own AI analyst in ChatGPT that combines GA4, Google Ads, Meta and your lead list - and where it breaks.

All three platforms count spam

What you get (+1 bonus)

Three platforms, one ChatGPT project.
Meta's connector added by its URL, a third-party connector for GA4 and Google Ads, or plain CSV exports, with the developer-only steps boxed so you can hand them over.
A lead list with no names.
Mark each lead real, junk or duplicate in your own tool, strip the list to 11 columns with no names or contact details, and check ChatGPT's training setting before you upload.
Rules ChatGPT reads first.
Instructions that tell ChatGPT to count only the leads you marked real, plus definitions and settings files. All of them, with the lead list columns, also come as a separate download.
15 prompts, labelled by trust.
From real leads by week to Meta instant-form leads against what Meta reports: four SAFE lookups, eight VERIFY prompts that say what to check, and three that won't work without joined data.
A 20-minute trust test.
Five questions on one closed week, checked against each platform and your own list, then scored. Re-run it whenever you change ChatGPT's model or thinking level.
Where it breaks in ChatGPT.
Spam all three platforms count, UTC exports that put leads made near midnight on the wrong day, app responses that can be cut short, and Python that counts a double-firing form event twice, precisely.
+1 bonus
An AI skill for Claude, ChatGPT, Gemini or Copilot
that helps you set up the project and strip your lead list, then runs the trust test and the monthly check on your own numbers.

Free · about 15 seconds

Who it's for, and who it's not for

The owner of a small service business at the front desk after closing, with a phone, an appointment book, enquiry notes and a laptop open to a chat window.

Who it's for

Businesses that collect leads through website forms, calls, bookings or Meta and Google lead forms, such as clinics, trades, agencies and B2B services, and run Google and Meta ads.
Whoever runs their marketing, in-house, at an agency or freelance, and has to say which of the lead numbers is real.
Anyone who wants ChatGPT to compare what Google Ads, Meta and GA4 claim with the enquiries that arrived, without uploading a single name or email.

Who it's not for

Businesses that already join ad, analytics and CRM data in a data warehouse such as BigQuery - you'll know most of what's in here.
Anyone who can't export the lead list or get read access to the ad accounts - the comparison this guide builds needs both.

What's inside

PDF · About 40 min read

Honest note

Passing the trust test proves ChatGPT reads your sources accurately. It can't prove your tracking is right, and Python can't fix the inputs: unmarked spam counts as a real lead, and a form event firing twice counts twice. The permissions Meta lists for its connector don't include lead retrieval, so you download instant-form leads by hand. A stripped list is still personal data under GDPR; counts-only uploads avoid that. All hours are our estimates: three to five a month, mostly the lead list.

Free · about 15 seconds