Half the traffic on the web is no longer human
And AI training is now 52% of all crawler requests, up from 22% a year ago.
Cloudflare, July 2026 · June 26, 2026 · 3 min read
Cloudflare sits in front of a large fraction of the web and publishes what it sees. In July 2026 it published three numbers that, taken together, describe a different internet from the one most businesses think they are on.
More than 50% of traffic on the internet is now non-human
52% of crawler requests are for AI training — up from 22% in spring 2025
Over 36% of crawler activity is mixed-use, and even Cloudflare cannot tell
what the request is for
The first number gets the headlines. The third is the one that matters.
Mixed-use crawlers are the unsolved problem. Googlebot is the clearest example: the same crawler that indexes your site for ordinary search also feeds AI Overviews. There is no way to serve one and not the other. Over a third of all crawler activity now falls into this category — requests whose purpose cannot be determined from the request itself. Cloudflare names this as the central open problem in bot management, and it is why the simple advice “just block the AI crawlers” does not survive contact with reality.
The traffic imbalance is the second story. Cloudflare publishes crawl-to-refer ratios — how many pages a platform’s crawler takes for every visitor it sends back. For the first week of August 2025:
Anthropic 50,000 : 1
OpenAI 887 : 1
Perplexity 118 : 1
Two honest notes on those figures, because they get quoted carelessly. Cloudflare states in the same post that traffic referred by Claude’s native app carries no referer header, so “these calculations may overstate the respective ratios, but it is unclear by how much” — the Anthropic number is the least trustworthy of the three. And Cloudflare has not published refreshed per-platform ratios since August 2025. If you see a 2026 figure — the “4,580 to 1” number circulating widely is the common one — it does not trace to any Cloudflare publication.
What does this mean for a business with a normal website?
Less than you might fear, and something different from what you might expect.
Your site is almost certainly being crawled by these systems already. Cloudflare’s 2025 year in review put Googlebot at 4.5% of all HTML request traffic and every other AI bot combined at 4.2%. That is real load but it is not the thing to worry about.
The thing to worry about is the asymmetry of the exchange, and what people are doing about it. Over fifty publisher–AI licensing agreements have been signed since 2023. Cloudflare has deployed a Content Signals policy to 3.8 million domains — though note that number is mostly Cloudflare setting a default, not publishers choosing one — and launched a monetisation gateway that lets a site charge per crawl. No adoption figures have been published for that, by anyone, and anyone quoting one is inventing it.
For a small business the calculus is simply different from a publisher’s. A newspaper loses when a model answers instead of sending the click. A plumber does not have an advertising business to protect; the model naming them is the outcome. The publisher’s problem and the small business’s problem look similar and are opposite. Most of the commentary you will read about AI crawlers is written from the publisher’s side, and applying it to a service business is how people end up blocking the crawler that would have recommended them.
And a related trap. Cloudflare’s August 2026 launch of AI-visibility measurement came with a line worth sitting with: “fewer than half of all HTML page requests now come from a human.” If your analytics show flat traffic while your logs show growth, those are both true and they are measuring different populations.
