How to Track Citations from ChatGPT to Your Website

Why ChatGPT citations are hard to track

ChatGPT links open in a new tab with limited referrer information, and many sessions strip the referrer entirely. The result is that traffic which actually originated from a ChatGPT citation often lands in analytics as direct or social traffic, distorting the apparent contribution of AI search.

The problem is compounded by the fact that the highest-value impressions never produce a click at all. A user reading a synthesised answer that mentions your brand has been influenced even if they never visit your site. Click-based attribution alone will systematically undercount AEO impact.

Signals you can actually capture

• Referrer header — `chat.openai.com` or `chatgpt.com` when not stripped; treat as a strong but partial signal.
• User-agent — ChatGPT's browsing tool identifies itself; log requests at the edge and attribute them as AI-driven.
• URL parameters — when you control the citation surface (e.g. earned media), add a campaign parameter so downstream clicks are unambiguously tagged.
• Prompt-level monitoring — run your target prompts on a schedule and record every citation, regardless of whether a click follows.
• Brand search lift — track branded query volume in classic SEO tools as a lagging indicator of AEO impression growth.

Combining the signals

No single signal is sufficient. The strongest measurement stack layers prompt-level citation data as the leading indicator, referrer and user-agent logs as the on-site confirmation, and branded search lift as the trailing macro signal. Together they give a defensible picture of AEO contribution.

Where attribution remains ambiguous, lean on incrementality testing. Pause a content cluster for a month, watch what happens to citations and branded search, then restart. The delta is your most credible measure of the cluster's true contribution.

Reporting it to the business

Report AEO impact in three layers. Citations and share of voice answer 'are we in the answer?'. Referral and user-agent traffic answer 'is the answer producing clicks?'. Branded search lift and pipeline data answer 'is any of this turning into revenue?'.

Stakeholders care most about the third layer. The first two are the leading indicators that let you act before the third layer moves. A good AEO dashboard makes all three visible side by side so causation can be reasoned about, not guessed at.

Frequently asked questions