ATTRIBUTION EPISTEMOLOGY
How to attribute ChatGPT traffic without pretending you can see everything
Use the strongest available evidence, label uncertainty and leave dark-funnel influence distinct from confirmed attribution.
- PossibleChatGPT discovery
AI answer evidence
- UnknownReturn via Google
Dark-funnel influence
- ObservedDemo
Website action
- ConfirmedCRM opportunity
Joined record
01 / FIVE EVIDENCE TIERS
Confidence should travel with every outcome.
- DirectionalAI answer evidence
A captured answer or prompt observation.
- DirectReferral evidence
Referrer or utm_source=chatgpt.com where available.
- ObservedWebsite evidence
Landing URL, session/event IDs and timestamps.
- CorroboratingBuyer-reported evidence
An optional “How did you hear about us?” response.
- ConfirmedCRM-confirmed evidence
Lead/contact, opportunity and closed-revenue records with defined windows and consent.
02 / PRACTICAL TABLE
Never collapse three states into one source field.
03 / DARK FUNNEL
Unknown does not mean unimportant.
A buyer can discover a brand in AI and later return through branded search, direct traffic, Slack or another device. Capture explicit referrers and utm_source=chatgpt.com where they exist; preserve landing URL, event IDs and timestamps.
Self-reported attribution is useful corroboration, not perfect truth. Automation can act on patterns, but Eli should keep the evidence state visible on every outcome.
NEXT DECISION
Use attribution to improve—not overstate—the chain.
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.
- PrimaryPublishers and developers FAQOpenAI · Checked September 2026 · Supports: publisher access and referral context
CONNECT YOUR WEBSITE
Measure AI influence with evidence, not guesses
Keep confirmed referrals, supported influence and unknown journeys visibly separate.
Measure AI influence