TRACKING DATA MODEL

How to track ChatGPT mentions without mixing up five different signals

A brand appearing in an answer can be a mention, recommendation, citation, referral or outcome. Those signals should not share one vanity score.

MentionRecommendationCitationReferralOutcome

01 / THE FIVE SIGNALS

Give every signal its own event type.

Mention
The brand is named in an answer.
Recommendation
The answer presents the brand as a fit.
Citation
A source URL/domain is explicitly referenced.
Referral
A visit reaches the site with available referral evidence.
Outcome
A website or authorized CRM event follows.

02 / EVENT SCHEMA

Keep the captured evidence attached.

Illustrative event schema

{ prompt_id: "required", engine/model: "captured value", timestamp: "required", brand: "captured value", event type: "required", context/position: "captured value", source URL/domain: "required", landing page/referral: "captured value", website action: "required", CRM evidence: "captured value"}

03 / AUTOMATION AND BREAKPOINTS

Automate capture; do not automate certainty.

What can be automated

Scheduled prompt observations, answer capture, entity detection, citation extraction, referral tagging, website-event capture and CRM matching where connected and authorized.

Where attribution breaks

Dark traffic, copied URLs, different devices, uncited model knowledge and privacy or consent limits all create gaps.

Where a ChatGPT referral exposes utm_source=chatgpt.com, capture it. Do not imply that every AI-influenced journey carries a referrer. Eli uses signals to prioritize and re-measure an action—not merely to draw a dashboard.

SOURCES AND METHOD

Sources checked for this comparison.

These links support the external product and platform references on this page. Eli publishes this comparison; verify critical requirements directly with each provider.

CONNECT YOUR WEBSITE

Track the signal, then act on it

Keep the prompt, answer, source and downstream evidence connected to each observation.

Track the signal