AI Traffic
Measure both sides: what AI crawlers take from your property, what AI products send back, and the ratio between them. Crawler logs or referral analytics alone show only half of the exchange.
AI crawlers
Pages taken
AI products
Visitors returned
Exchange
Crawl to referral · N:1
An upper bound because observable referrals are a floor.
Crawls in. Visitors back.
Crawler logs record what AI companies fetch. Referral analytics record some of the people their products return. Reading either side alone hides the exchange between them.
What they take
See the pages AI crawlers fetch, which operators fetch them, each operator's share, and whether that operator sent an observable AI referral back.
What they send back
See visitors referred by ChatGPT, Claude, Gemini, Perplexity and other AI products, which product sent each one, and the distinct landing pages those platforms linked people to.
Crawls and referrals use separate chart axes. Their gap is the subject of the measurement, not a like-for-like comparison.
Four deliberately narrow metrics
Each definition states what the request data can support—and what it cannot.
Pages crawled
AI crawlers only. Crawls from operators that serve both search and AI are counted separately and are not included.
AI referrals
Visitors an AI assistant sent to this site, counted when the arriving request carries a Referer naming a known AI platform.
Crawl to referral
Pages taken for every visitor returned, shown as N:1. Referrals are a floor, so this ratio is an upper bound.
Pages cited
Distinct pages an AI platform linked someone to. Whether a model cited a page inside an answer is not observable.
A ratio with an explicit limit
WebDecoy expresses the exchange as pages taken for every visitor returned, shown as N:1. A visitor counts from a known AI-platform Referer, or from a recognized campaign tag when Referer is absent.
AI platforms can omit both signals. The referral total therefore undercounts, and the crawl-to-referral ratio is an upper bound—not a precise measurement. A site may receive zero observable AI referrals while AI tools still link to or use its content.
Arriving request
Referer: known AI platform
→ count one AI referral
No Referer, recognized campaign tag
→ count one tagged AI referral
No identifiable signal
→ no attributable AI referral
Result
referrals = floor
crawl-to-referral ratio = upper boundMixed search and AI crawlers stay separate
Google, Microsoft, and Apple use one crawler for both a search engine and an AI product. A fetch from one of those operators cannot be assigned reliably to search or AI.
WebDecoy labels those operators as search + AI and excludes their crawls from the attributable AI total and its ratio. Search-engine referrals are outside this measurement, so a shared operator can show no AI referrals here while its search engine still sends real traffic.
Attribution boundary
AI-only crawler fetches contribute to pages crawled.
Search + AI crawler fetches are reported separately.
Only identifiable AI-product referrals count on the return side.
Freeze the exchange at a stable address
Create a snapshot for a 7, 14, 30, or 90-day window. The report does not change with the live dashboard.
Private or public
Private reports require sign-in to the owning organization. Public reports are readable by anyone with the link. Visibility can change without changing the address.
Named or anonymous
A report can be published without naming the site. Public reports are marked noindex.
Made to share
Each public report has an Open Graph share card rendered from that report's own figures. Deleting a report permanently removes it.
A real, frozen report
WebDecoy published its own 30-day snapshot for 27 July–26 August 2026. It recorded 13,828 attributable AI pages crawled and 5 observable visitors returned: a 2,766:1 upper-bound ratio because referral headers are frequently stripped.
A further 2,625 crawls from shared search + AI operators were kept outside that ratio. This is one window on one small site, not an industry benchmark or trend.
Keep the referral side in context
Referral tracking still answers concrete questions about the visitors AI products return. It now sits beside the crawler activity that precedes it.
Products sending visitors
See which AI products produced identifiable referred visits.
Linked landing pages
See the distinct pages an AI platform linked someone to—not unobservable citations inside model answers.
Property-level windows
Review the exchange for the selected property over 7, 14, or 30 days.
Frequently asked questions
Definitions and limits for AI traffic measurement.
What does WebDecoy measure on the AI Traffic page?
For each property, WebDecoy puts AI crawler activity and AI-referred visits in one view. It reports pages crawled, AI referrals, crawl to referral, and pages cited, with live windows of 7, 14, or 30 days.
How are AI referrals counted?
An AI referral is counted when an arriving request carries a Referer naming a known AI platform. AI platforms frequently strip that header, so the referral count is a floor rather than a complete count of visits from AI products.
What does the crawl-to-referral ratio mean?
It is the number of pages taken by attributable AI crawlers for every visitor returned with an observable AI Referer. It is shown as N:1. Because referrals are undercounted when platforms strip the Referer, the ratio is an upper bound, not a precise measurement.
What does "pages cited" mean?
Pages cited means distinct pages an AI platform linked someone to. WebDecoy can observe that landing-page link through a referral. It cannot observe whether a model cited your page inside an answer, because that happens inside the model and leaves no trace in your site logs.
Why are shared search and AI crawlers separate?
Google, Microsoft, and Apple operate crawlers used for both search and AI. Their fetches cannot be attributed to one side, so WebDecoy holds them outside the attributable AI crawl count. Search-engine referrals are also outside this measurement.
What is an AI Traffic Exchange report?
It is a frozen snapshot for a 7, 14, 30, or 90-day window at a stable address. It can be private or public, can omit the site name, and can switch between private and public without changing its address. Public reports are noindex and have a share card made from their own figures. Deletion is permanent.
Investigate an individual actor
What is this traffic looking for?
Alongside property-level AI Traffic reports, AI Journey Insights lives in Actors → Journey. Review likely collection targets, see whether requests concentrate on one page or cover many, and inspect the detections behind each observed pattern.
It analyzes a bounded sample on demand. These insights describe observed activity; they do not prove successful collection or attribute traffic to a particular AI operator.
Explore AI Journey Insights →Measure your AI traffic exchange
See what AI crawlers take from your property, what AI products observably send back, and the upper-bound ratio between them.
Open AI Traffic