How to track ChatGPT traffic and connect it to form leads

Google introduced the AI Assistant channel on 13 May 2026. Its release notes describe automatic classification when the referrer matches a recognised AI assistant: the channel is AI Assistant, the medium is ai-assistant, and the campaign is (ai-assistant).

Start with the data your property has recorded:

1. Open Reports > Acquisition > Traffic acquisition.
2. Select a date range and inspect Session default channel group for AI Assistant.
3. Switch the table dimension to Session source / medium and search for chatgpt to inspect the matching source rows.
4. Review the source spelling and medium shown in those rows before using them in a report.

Google's Traffic acquisition guide documents the report and its search. If the report is missing from your navigation, an Editor can add it back.

The AI Assistant total includes multiple assistants. Inspecting source rows lets you separate the ChatGPT entries from other platforms. Historical data and different incoming signals may also produce different source/medium combinations, so a filter limited to chatgpt.com / referral can miss relevant rows.

These steps describe the current documented GA4 method. The figures in your property depend on the data it receives; a recognisable source isn't present on every visit.

Keep Google's AI search features separate too. Its channel definitions place Google AI Overviews and AI Mode within Organic Search, outside the AI Assistant category. An AI Assistant total therefore doesn't represent all AI-influenced discovery.

A ChatGPT answer can mention your business without sending someone to your website. A person who does click can browse without contacting you. Keep those stages separate when reporting results.

A session can include more than one submission, and a visitor can return in another session. Comparing session totals with unique CRM enquiries won't produce a one-to-one match.

Use a verified successful-enquiry event if you want to examine form activity in GA4. Check that it represents the result you intend to count and compare it with saved submissions. Opening a form or clicking Submit can happen without a valid enquiry reaching your team.

For session traffic, use session-scoped dimensions. First user source answers how the user was first acquired, which can differ from the source of a later visit. Google's scope guide explains that distinction.

Capture the source context when the visitor arrives, then retain it until the enquiry is submitted. Otherwise, someone can arrive through ChatGPT, browse several pages and submit from a contact page that no longer has the original information.

The useful inputs include an available referrer and any campaign tags in the arriving URL. For example, if a URL contains utm_source=chatgpt.com, preserve that supplied value. It doesn't supply a medium or campaign that is absent from the link. Google's traffic-source processing guide describes how campaign parameters and referrer information contribute to analytics attribution.

A tagged link can also be copied elsewhere. Treat its label as recorded evidence about the link, rather than proof of the exact answer someone read.

TrackLeadSource collects available source, medium and campaign information, along with the recorded source journey and pages visited. Its workflow passes that context through hidden form fields. Your form connection maps the submitted values into your CRM or spreadsheet, where sales adds qualification and customer outcomes.

Imagine an operations manager researching warehouse layout advice. They arrive through ChatGPT, return later through a newsletter, read a service page and case study, then request a consultation. An illustrative record could contain:

This is a fictional example, not a customer result or a live integration test. The labels describe the information to keep, rather than required TrackLeadSource field names.

Keeping the journey lets your team see the earlier ChatGPT arrival as well as the newsletter return. For reporting, choose a consistent rule such as first recorded source or latest recorded source. Those views can group the same enquiry differently. A page visit alone doesn't prove buying intent or establish which interaction caused the sale.

Our website form tracking guide explains the general capture and transfer process. Use your form's supported method and inspect the saved destination record before relying on it.

Check the saved enquiry, not only the source shown on the page. A value visible in the browser can still be missing from the form submission or its CRM mapping.

For a controlled check, use a form you own and record the input you supplied. Then inspect the saved result:

• Open a tagged landing page, browse to an untagged form page and submit a test enquiry. Check whether the intended source context survived.
• Repeat through a returning visit to check which source values change and which earlier arrivals remain in the journey.
• Compare the saved form entry with the corresponding CRM or spreadsheet record, using an enquiry ID to match them.
• Test an untagged visit too, so a fixed test value doesn't label unrelated enquiries as ChatGPT.

Manually adding a ChatGPT source tag tests your form's handling of that input. A real click from ChatGPT is a separate check of what that particular link and browser pass. Record those tests separately; one cannot establish the result of the other.

When neither a useful referrer nor a source tag arrives, leave the origin unknown rather than assigning it to ChatGPT. Google's Direct traffic explanation lists several reasons for unclear origins, including missing campaign information, redirects and blockers. An increase in Direct traffic alone doesn't tell you which of those happened.

A person may also read a recommendation, then search for your company name or type the address. That later arrival gives you limited evidence about the earlier recommendation. An optional “How did you hear about us?” question can add context. Keep that answer separately from the recorded source, as explained in our survey options guide.

Once the source context reaches the destination, keep it linked to the sales process. Record qualification, associate the relevant deal and update its outcome. The CRM lead source guide covers consistent fields and retaining that relationship when a lead becomes a deal.

For a first comparison, select a defined group of enquiries and state your source rule. You might review enquiries received in July whose recorded journey includes ChatGPT, with sales outcomes updated through October. Keep open deals visible and count each associated deal once.

A journey-includes-ChatGPT report can overlap with reports for other sources. The example above would also appear in a newsletter journey report, so adding the two reports together would double-count that enquiry. Use mutually exclusive source groups when you need a total that adds up.

Look at suitable projects and customers alongside the enquiry count. A source bringing enquiries for a service you don't offer calls for a different investigation from one bringing suitable projects that remain open. Our lead-quality guide explains fair comparisons and allowing time for leads to buy.

You can use an ordinary CRM report or an export. If an approved AI tool helps you explore that export, a starting request is:

Find July enquiries whose recorded source journey includes ChatGPT. Show unique enquiries, qualification status, associated won deals and open deals. Count each deal once. List missing source or outcome data, and explain how you matched enquiries to deals. Keep self-reported ChatGPT answers in a separate group.

Check the returned totals against the saved records before deciding where to spend time or budget. To track lead source through to customers, your source context and sales outcomes need to remain connected to the same enquiries.