AI referral signal
A directional indication that a human browser may have reached a publisher from an AI answer; useful for measurement, but it cannot prove identity or bill.
Also called: AI-referred visit, AI referrer signal
An AI referral signal is a browser request whose Referer header names a known answer engine or AI
assistant. It can help a publisher compare the human visits an engine appears to return with the
verified crawls that engine made earlier. It is a measurement input, not a paid placement and not
payment evidence.
Why is an AI referral only directional evidence?
A browser supplies the referrer header, and a client can forge or omit it. Privacy settings, redirects, apps and an engine’s own link handling can also strip or rewrite the value. The signal therefore cannot establish that a particular engine, answer or earlier crawl caused the visit.
CrawlClick normalizes exact known referrer hosts and keeps this evidence separate from verified crawler identity. A lookalike hostname or a user-agent string does not strengthen the claim. Reports should describe the result as observed navigation, not as audited attribution.
Can an AI referral trigger billing?
No. Day-0 production billing accepts only a disclosed insertion durably recorded on an eligible, network-verified crawler request. Human requests receive the unchanged origin response. A referral signal does not enter an auction, consume advertiser budget, create publisher earnings or authorize different content for the visitor.
This boundary matters even when the same engine previously crawled the same page. A verified crawl does not turn a later forgeable browser header into trustworthy payment evidence. Correlation can support an experiment, but it cannot safely create a money movement.
What is the signal useful for?
Publishers can compare verified crawls and observed referrals over the same period, by engine and page, to describe the crawl-to-referral gap. Advertisers can use aggregated referral direction as a secondary outcome alongside independent answer probes. Both should preserve missing observations and changes in measurement coverage rather than reporting the ratio as a guaranteed causal effect.
Does measurement change what a visitor sees?
No. The CrawlClick-managed edge keeps human traffic on the ordinary origin path. If referral measurement is enabled for an approved pilot, it records a bounded event and returns the publisher response; it does not add a sponsored passage. If CrawlClick is unavailable, the route still fails open to the real page.
What else should I read?
- Crawl-to-referral gap
The ratio between how often an AI engine reads a site and how many visitors that engine sends back, measured per engine.
- Crawler verification
Confirming that a bot claiming to be a particular AI crawler genuinely is one, since user-agent strings are trivially forged.
- Answer engine
A system that responds to a question with a synthesised answer rather than a list of links, typically by retrieving live web pages and summarising them.
- Surfacing
When content placed on a page actually appears in an AI engine's answer, whether quoted directly or paraphrased.