We Logged 14.3 Million AI Crawler Visits. 75% Came From One Bot.
There is a new kind of reader on the web. It does not click ads, it does not run JavaScript, and it never shows up in Google Analytics. It is the fleet of AI crawlers that ChatGPT, Claude, Apple, Amazon, and the rest send out to read the open web, and what they read shapes what those assistants say about every business on it. We operate one of the larger networks of business pages anywhere, 12.9 million of them, and we count every one of these visits at the server, where a bot cannot hide. As of this week the counter crossed 14.3 million visits. Here is who is actually doing the reading.
The AI crawler census across 12.9M business pages. Bar lengths are square-root scaled so the small crawlers stay visible; the numbers are the real counts. Source: Engagemii, July 2026.
Finding one: the AI reading the web is mostly Claude
ClaudeBot, Anthropic's crawler, accounts for 10.7 million of our 14.3 million visits. That is 75 percent of all AI crawler traffic we have ever logged, more than five times OpenAI's GPTBot, the name most people reach for when they picture an AI bot. And it is accelerating: 9.2 million of those ClaudeBot visits, 86 percent of its lifetime total with us, landed in the last 30 days alone.
Whatever Anthropic is building its next models and answers from, it is reading the business web at a pace nobody else in this census comes close to right now.
Finding two: Apple is the sleeper
Applebot is the quiet number three, 1.14 million visits, and the last 30 days tell the sharper story: Applebot read over a million of our pages in the past month, three times more than GPTBot did in the same window. Apple barely markets its AI answers today, but its crawler behaves like a company stocking a very large pantry. When Siri starts answering questions about businesses at iPhone scale, the sites it can read cleanly are the ones it will have on hand.
Finding three: the most citation-happy engine crawls the least
PerplexityBot, from the answer engine famous for citing its sources, is dead last among the majors: 6,349 visits, ever. That is 0.04 percent of the traffic. However Perplexity assembles its answers, it is not by bulk-reading business pages the way Anthropic, OpenAI, and Apple do. The practical takeaway is that you cannot count on every engine to eventually wander by. The ones doing the mass reading are doing it now, and the ones that are not may only ever see you through the documents everyone else already indexed.
None of this shows up in your analytics
Every visit in this census is invisible to JavaScript analytics, because these crawlers do not run JavaScript. A business owner staring at Google Analytics sees zero of the 14.3 million reads in this dataset. The only place this traffic exists is in server logs, which almost nobody reads. That gap is why most businesses have no idea whether the AI engines their customers now ask have ever even looked at them.
The honest limits
This is one vantage point: our network, which skews toward business and brand pages. A news site or a forum would see a different mix. Bot identity is taken from verified user agents with reverse-DNS checks on the major crawlers, so the headline numbers are conservative. And a crawl is a read, not a promise: being read is the entry ticket to AI answers, not the whole game.
See who is reading your site
We made the same counting available for any website, free: install monitoring and see which AI crawlers visit you, how often, bot by bot, week by week. Pair it with your free AI visibility score and you know both halves: whether the engines are looking, and what they can read when they do. Both at engagemii.com/aeo.
If you want to cite this article, the URL is engagemii.com/blog/we-logged-14-million-ai-crawler-visits-75-percent-one-bot.
Ready to find out if AI can cite your brand?
Get Your Free AEO Score