Tracking ChatGPT Referral Traffic and Brand Mentions

#01Where does ChatGPT traffic show up?

Set up the channel group with the GA4 AI referrals guide and read why AI referral traffic is undercounted before you report a number.

#02How do you capture what analytics misses?

Add a self-reported field to demo and signup forms, with "AI assistant (ChatGPT, Claude, etc.)" as an option and a free-text follow-up. It catches buyers who heard of you in ChatGPT and arrived later through search or Direct. See adding ChatGPT to self-reported attribution.

ATTRIBUTION READ
Attached: last quarter's form responses (source answer,
free text, deal stage). No personal data. How many mention
ChatGPT or another AI assistant? Compare their conversion to
demo and to closed-won against other sources. Quote 5
free-text answers.

#03How do you track mentions?

Build a fixed set of 20-50 buyer questions across intents (see building a prompt set). Run them on a schedule and log, for each answer: whether you're mentioned, your position, which of Competitors A-D appear, and which sources are cited.

MENTION LOG
Search: "[buyer question]". Return one table row: question,
brands mentioned in order, whether [Brand] is mentioned (Y/N),
how [Brand] is described in one line, cited source URLs.

Run checks logged out or in a temporary chat; OpenAI says memory and approximate location shape search, so your own account isn't neutral. A scheduled task (lesson 13) can run a small set weekly; for larger sets use a tracker, compared in manual vs automated AI visibility tracking.

#04Which metrics should you report?

Definitions are in AI share of voice and citation share. Log wrong facts over time with tracking brand hallucinations.

Our view: report the mention rate and the referral trend side by side each month. Either alone misleads: mentions without traffic can mean weak positioning in answers, and traffic without mentions usually means a few lucky citations.

#05Worked example: Ledgerline's monthly ChatGPT report

Ledgerline is a fictional accounts-payable tool. Details are illustrative.

Ledgerline built a 30-prompt set, ran it weekly in temporary chats and set up an AI channel group in GA4. After the CDN fix from lesson 19, its mention rate on "compare" prompts rose over six weeks while referral sessions grew more slowly. Self-reported attribution filled the gap: several demo requests named ChatGPT but arrived via brand search. The monthly report showed all three side by side, with the wrong-fact log underneath.

#06What should you check before you finish?

• Set up the GA4 AI channel group.
• Add the AI option to your attribution field.
• Run your first prompt set and save the baseline.
• If you also track Claude, use the same prompt set with measuring brand visibility in Claude so results compare.

You have finished the track. Return to the ChatGPT for Marketers track to claim your certificate.

#07Common questions

How do I see ChatGPT traffic in GA4? Look for chatgpt.com as a session source, then group AI sources into a custom channel so they stop hiding in Referral. Some ChatGPT visits arrive without a referrer and land in Direct.

Does ChatGPT tag its links with UTM parameters? Many sites see utm_source=chatgpt.com on visits from ChatGPT links, but OpenAI's help pages don't document it for organic answers. Check your own data. For ChatGPT Ads, add your own UTMs to landing page URLs.

How often should I check brand mentions in ChatGPT? Weekly for a fixed prompt set is enough for most teams. Answers vary run to run, so judge trends over several weeks, not single answers.