News — On 13 August 2026, Microsoft added an “AI Scrape-to-Referral” metric to Clarity's Bot Analytics report, comparing — operator by operator — how much AI crawlers extract from your pages against the number of visits those same operators send back (Microsoft Clarity, official blog, 13 August 2026).
Until yesterday you could see what AI took, or what it gave back. Never both on the same screen. Server logs and your CDN dashboard counted crawler hits. GA4 counted visits arriving from ChatGPT or Perplexity. Nobody lined the two columns up per operator. That is exactly what Clarity has now shipped.
What Microsoft added to Clarity
The example screenshot released with the announcement, picked up by Search Engine Land on 13 August, shows 41 visits attributed to one AI operator against a ratio of roughly 6,000 scrapes per visit. This is not a market average: it is a product screenshot. Its value is not the number itself, it is that any site can now compute its own.
In practice Microsoft shipped five blocks: a scrape-to-referral ratio card, an operator-level breakdown separating those who send traffic back from those who extract and return nothing, mapped-domain coverage safeguards to stop the calculation going wrong when CDN scope and Clarity scope do not line up, direct access to session recordings filtered by AI referral source, and the engagement metrics attached to them (scrolling, conversions, form fills).
It is the logical follow-up to the work started in spring, when Microsoft made the Citations dashboard that measures your citations inside AI answers generally available. Citations told you whether you were being talked about. The ratio tells you what that is worth.
Why this number was missing
The AI-crawler debate has run for eighteen months without an instrument. Everyone knew the volume had become enormous: Cloudflare documented the moment bots overtook humans on the web. Everyone could measure the return, once the plumbing was done: we walked through how to isolate ChatGPT, Claude and Gemini traffic in GA4. What was missing was the relationship between the two, operator by operator.
Without it, “should we block AI crawlers?” was settled on instinct. Publishers who chose to pull out of Common Crawl did so without knowing what traffic they were giving up. Sites that will tick Cloudflare's default AI-blocking box on 15 September will do the same, unless they look at their ratio first.
The other half of the story: Shopify's data
Two days earlier, on 11 August, Shopify published its second-quarter 2026 figures. AI-referred sessions across its merchants are up 197% year over year and the associated orders tripled, while organic search grew 12% on a much larger base. In spec-led categories, a shopper arriving from an AI converts at roughly twice the rate of one arriving from organic. Half of those AI sessions land straight on a product page.
What connects the two announcements: Shopify shows the return can be real and high-quality. Clarity shows it is very unevenly spread across operators. So an average ratio tells you nothing — only the per-operator breakdown lets you decide anything at all.
What this changes for a small business
First, measure before you block. Clarity installs with one tag and the Bot Analytics report needs no extra budget. Watch the operator-level breakdown for three or four weeks. You will then know which ones cost you bandwidth for nothing, and which ones send you buyers.
Second, judge on engagement, not volume. Forty visits that fill in a form beat four thousand that bounce. That is precisely why Microsoft wired session recordings into this report: the ratio asks the question, the session answers it. We saw the same gap in Adobe's data, which showed AI traffic converting markedly better than conventional traffic.
Third, structure what you expose. The most actionable detail in the Shopify report is this one: shoppers arriving via structured product data converted twice as well as those arriving via a scraped listing or a third-party feed. What you hand over cleanly comes back better than what gets taken from you sideways.
The limits of this reading
6,000-to-1 is not a benchmark. It is an interface screenshot Microsoft published to illustrate a product. Nobody knows which site it came from or over what period. Treating it as a reference point would be a misreading.
A scrape is not a citation. The ratio measures traffic return, not visibility. An operator can scrape you heavily, cite you in its answers, and send almost no clicks because the user got what they wanted without leaving the conversation. Blocking on that criterion alone would remove the citation without recovering the traffic.
Clarity only sees what is attached to it. Microsoft added mapped-domain coverage safeguards precisely because the scope observed on the CDN side and the scope observed on the Clarity side often diverge. A ratio computed over a partial scope is a wrong ratio.
Shopify's numbers describe e-commerce. These are merchants on a single platform, in categories where the product page does the work. A consultancy or a tradesperson has nothing to infer directly from a conversion rate on watches. Finally, this article does not cover the technical implementation of blocking or allowing crawlers: that is a separate discussion.
Frequently asked questions
What is Microsoft Clarity's AI Scrape-to-Referral ratio?
It is a metric added to Clarity's Bot Analytics report on 13 August 2026. It compares how much an AI operator scrapes your pages against how many visits that same operator sends back. A 6,000-to-1 ratio means 6,000 crawler hits for a single human visit attributed to that source.
Does a high ratio mean you should block that AI crawler?
No. The ratio measures traffic return, not visibility. An operator can scrape you heavily, cite you in its answers and still send almost no clicks: blocking it would remove the citation without recovering the traffic. The decision should rest on the value of the sessions you receive, not on scrape volume.
Is AI-referred traffic actually worth anything?
Shopify data published on 11 August 2026 shows AI-referred sessions up 197% year over year in Q2 2026, with conversion roughly twice the organic rate in spec-led categories. Volume stays small, per-session quality is high.
The Cicero take
This ratio is going to be misread, and fast. The number is spectacular, it lands a month before Cloudflare flips AI blocking on by default, and it hands anyone who wants it a ready-made argument: they take everything, they give nothing back, cut them off.
Except the ratio is not the decision metric. It is the question, not the answer. The answer sits two screens further on, in the session recordings: those forty visits — what do they actually do once they land? Microsoft put both in the same report, and that is the real news here. A crawler's cost is measured in bandwidth. Its value is measured in customers.
Sources
- → Microsoft Clarity (official blog): “Understand Which AI Crawlers Drive Real Visits with AI Scrape-to-Referral Insights”, announcing the five new Bot Analytics blocks, 13 August 2026.
- → Search Engine Land: Barry Schwartz's write-up of the announcement, including the example screenshot showing 41 visits and the roughly 6,000-to-1 ratio, 13 August 2026.
- → Shopify (Enterprise blog): “AI and organic search are doing different jobs”, Q2 2026 data on AI-referred sessions, orders and conversion rates, 11 August 2026.
Growth and SEO content strategist, I founded Cicéro to help businesses build lasting organic visibility — on Google and in AI-generated answers alike. Every piece of content we produce is designed to convert, not just to exist.
LinkedIn