Track Google, ChatGPT and Ads From Website Visit to Qualified Enquiry
To learn whether Google, ChatGPT or ads generate useful enquiries, connect three records: the identifiable visit source, the accepted lead and the verified sales outcome. Page views alone cannot establish that chain. In particular, a Google-branded device is not evidence of Google Search acquisition.
Use the attribution test matrix. This guide defines measurement and verification; it does not claim that your current form-to-Sheets path or production analytics has passed every test.
Read the correct analytics scope
GA4's Traffic acquisition report describes session acquisition. Compare session source/medium with session outcomes, and label first-user acquisition separately if you use it. Mixing these scopes can create apparent disagreements that are actually different questions.
Check hostname and landing page so marketing visits and application usage are not mistaken for the same audience. A /login spike may come from returning customers, repeated views or other activity. Country and device breakdowns do not establish search intent, bot activity or acquisition source by themselves.
Identify AI referrals without guessing
OpenAI documents referral tagging with utm_source=chatgpt.com. Preserve identifiable tags and available referrers, then review how they appear in your analytics configuration. Referral visibility can be incomplete, so report identifiable ChatGPT traffic rather than all traffic supposedly influenced by AI.
| Evidence | Reportable interpretation |
|---|---|
| Google organic session source/medium | Identifiable organic search acquisition |
| Campaign tag or valid ad click attribution | Identifiable paid campaign under the declared attribution scope |
| ChatGPT referral or documented source tag | Identifiable ChatGPT referral |
| Direct, absent referrer or unknown tag | Uncertain source; do not assign to ChatGPT |
Do not manufacture source tags for visitors who did not arrive through those channels. Keep raw captured fields and your classification version so later changes are explainable.
Join the form to the actual outcome
Generate a unique enquiry identity when the backend accepts the form. Preserve permitted source, medium, campaign, landing URL and receipt time with that record. Avoid placing personal information in analytics URLs or event parameters. Respect the site's consent policy and minimize collected fields.
Verify the accepted backend record, Sheet/CRM delivery and qualification separately. A browser success message is not proof of a saved lead, and a saved lead is not proof of qualification. Deduplicate by enquiry identity and record staff-reviewed outcome definitions.
Compare complete periods and source cohorts
Use equal, complete weekly or 28-day periods with the same timezone. Report sessions, accepted unique enquiries, qualified enquiries, completed demos and cost where available. Keep brand/non-brand Google queries separate in Search Console, and acknowledge that query-level and lead-level data are not a perfect individual join.
Test organic, tagged campaign, identifiable AI referral, direct, rejected form, duplicate submit and marketing-to-app navigation. Compare analytics with backend counts and disclose blockers. This establishes an evidence chain for decisions without treating missing attribution as a reason to invent a source.
Frequently asked questions
Can Search Console tell me how many ChatGPT leads I received?
No. Use identifiable referrals or source tags in website analytics and join accepted enquiries to business outcomes. Search Console measures Google Search performance.
Does direct traffic mean ChatGPT?
No. Direct or missing attribution has many causes. Retain it as uncertain unless you have supporting evidence.
Does a Google device mean Google Search traffic?
No. Device information describes the device; source and medium describe acquisition. A Google device cannot establish a search referral.
Continue planning your workflow
See the parent guide, calling costs, or request a scoped demonstration. Examples and test resources are educational; verified provider configuration and measured customer results must be evaluated separately.
Put these ideas into production
Deploy AI voice agents in minutes and build outbound, inbound, and follow-up workflows on one platform.
Related Articles
What AI Voice Agent Metrics Should You Track After Launch?
Learn the AI voice agent metrics that matter after launch: task completion, containment, latency, transfer rate, tool-call accuracy, sentiment, cost per resolved call, and failure reasons.
AI Calling Cost per Qualified Lead: A Calculator With Clear Assumptions
Calculate supplier, acquisition, review and sales follow-up costs per verified qualified lead, including sensitivity analysis and zero-outcome handling.
AI Lead Qualification Scorecard: Budget, Need, Timing and Human Review
Build a reviewable AI lead qualification scorecard using stated need, budget, timing and next steps, with explicit unknowns and a downloadable evidence matrix.