Microsoft Clarity shipped an AI Scrape-to-Referral Ratio card inside its Bot Analytics dashboard on August 13, according to a Clarity Staff post on the Clarity blog. The card divides crawler-request counts by human referral sessions to produce one number. It also ranks individual AI operators against that number and opens a filtered view of session recordings for whichever operator a site owner selects. The result is a calculation SEO teams used to reconstruct by hand, now built into a report. It lands inside a measurement category this publication has watched take shape from two other angles this month.
The ratio itself is arithmetic, not a new data source. Bot Analytics already logged crawler requests through a connected content delivery network. Clarity’s behavioral analytics already logged human sessions through the platform’s JavaScript tag. The new card just divides one figure by the other. It requires a supported CDN, Fastly, Amazon CloudFront or Cloudflare, or the current Clarity WordPress plugin, before it can populate at all.
Microsoft’s own illustration of the card shows a ratio of 6,000 to 1 against 41 total referrals. That number is a screenshot from Microsoft’s own product, not a published cross-site benchmark. The company has never called it a finding about any named AI operator. Microsoft frames it only as an indication of scale, and says a reading in the low thousands is typical rather than exceptional for this style of measurement.
The operator-level breakdown carries more weight than any single blended average. A ranked list separates the crawler that accounts for most of the requests from the crawler that accounts for most of the referrals, a distinction one site-wide ratio erases entirely. From there, a click opens the Recordings tab pre-filtered to that operator. A site owner can then watch whether visitors from a given AI source scroll, convert or leave immediately, rather than trusting the number on its own.
Clarity is not introducing this measurement category. A widely circulated version of the same calculation came from Cloudflare first, in an August 2025 publication. Cloudflare turned it into a paid Attribution Business Insights product starting July 1, 2026, with ratios by operator running from 118 up to nearly 50,000 crawls for every referral, tracked through UTM parameters. Cloudflare’s tool sits in the request path and sees the crawl directly. Clarity’s tool sits on the page and sees the session directly. Neither vendor’s architecture sees both halves natively, which is why Microsoft built mapped-domain safeguards, stopping the ratio from running across domains where its two collection layers do not both have coverage.
This publication has been following the same underlying math from two other angles recently. Google confirmed on August 4 that every one of its crawlers pulls from a single shared crawl budget, not separate allotments. On August 10, a site owner blocked Amazon’s crawler after logging roughly 117,000 requests a day against it. That is the same scrape-to-referral tradeoff, worked out from raw server logs before any vendor packaged it as a dashboard. Clarity’s card is that same arithmetic, shipped as a product feature instead of a spreadsheet exercise.
Microsoft is not a neutral referee in this argument. Clarity is a Microsoft product. Microsoft operates its own web crawler and its own AI answer surface, the same category of operator the card exists to rank. A tool that scores whether an AI platform returns value doubles as a competitive argument for the platform selling the tool, a tension the announcement does not raise.
The referral side of that number still runs through channels that undercounted AI traffic for most of the past year. Google confirmed in May 2025 that AI Mode’s link code carried a noreferrer element that stripped outbound referrer values, misclassifying those clicks as direct traffic. Google Analytics did not add a native channel for AI assistant referrals until May 13, 2026. A ratio built on a denominator that went uncounted for months makes the underlying gap look larger than it is, a distortion no dashboard corrects after the fact.
For any team deciding whether to block or allow a specific crawler, the operator-level ranking replaces one averaged headline with a per-source decision, provided a supported CDN is already wired in. Teams without Fastly, CloudFront or Cloudflare should put that infrastructure gap, not the ratio, at the top of their next ninety-day list.
Per a PPC Land report by Luis Rijo, published August 13, 2026.
