Across 20 sites, AI crawlers logged 114,644 visits in 40 days, more than 2,800 a day. The finding that surprised us most: the majority of that traffic has nothing to do with citing you. About 57 percent came from AI training crawlers harvesting content, and only about a third came from the AI search crawlers that can actually place you inside an answer. Crawl volume and getting cited are not the same thing.
AI crawlers visited 20 sites 114,644 times in 40 days. Fifty-seven percent were training crawlers, thirty-six percent were AI search crawlers, and the rest were scrapers and assistants. Twenty-nine different AI crawlers were active, and the average page was crawled almost six times.
The sample nobody else has
Most writing about AI crawlers is guesswork. People read one server log, or they repeat a number a vendor put in a press release, and they call it a trend. Aggregate sources like Cloudflare Radar confirm crawler traffic is climbing across the web, but a macro trend cannot tell you which bots reached your pages.
We had a different starting point. We record AI crawler visits at the server level across 20 live sites, every crawler, every page, every hit. In the first 40 days of tracking, June 5 to July 15, 2026, those 20 sites logged 114,644 AI crawler visits.
That is the number in the title rounded down. The real figure passed 100,000 well before the 40 days were up, and it is still climbing. Here is what it taught us.
Finding 1: The volume is far past what most owners assume
114,644 visits in 40 days is more than 2,800 AI crawler hits every day across the network.
These are not people. This is software reading your pages on repeat, at a pace no human audience could match. If you are still thinking about your site only in terms of human visitors, you are watching the smaller half of your traffic.
Finding 2: Most AI crawling is not about citing you at all
This is the one that changes how you should think about the whole game.
Of the 114,644 visits, the split by crawler type was:
- AI Training: 65,969 visits (57.5%) - crawlers reading your content to train models.
- AI Search: 41,029 visits (35.8%) - crawlers feeding live answer engines.
- Data Scraper: 3,965 visits (3.5%)
- AI Assistant: 3,681 visits (3.2%)
Read that top line again. The majority of AI crawler traffic on these sites was training harvest, content being taken to train a model, not to answer a question a person just asked.
Only the AI Search share, about a third, is the crawling that can put you inside an AI answer. So when someone says "the AI crawled my site," the honest response is: which kind? Because two out of every three of those visits were never going to cite anyone. Crawl volume feels like a win. Only a slice of it is the win that matters.
Finding 3: Your pages get read again and again
The 20 sites had 19,443 pages crawled to produce those 114,644 visits. That is almost six crawls per page in 40 days.
AI crawlers do not read a page once and move on. They come back. A page you published in June was revisited repeatedly through July. Whatever an AI model believes about your content, it is refreshing that belief on a short cycle, not caching it once and forgetting you.
Finding 4: It is a swarm, not a bot
29 distinct AI crawlers were active across the network in 40 days.
The conversation online tends to fixate on one or two famous names. The data says otherwise. Nearly thirty separate crawlers touched these sites, each with its own behavior, its own appetite, its own reason for being there. Optimizing for a single bot is planning for one guest at a party of 29.
Finding 5: The line is not flattening
Plot the 40 days out and the trend climbs, with repeated spikes pushing daily visits well above the average. There is no sign of the crawlers losing interest. If anything, the network drew more AI attention at the end of the window than at the start.
What we measured, and what we kept
For the record: these numbers are AI crawler visits recorded at the server level across 20 live sites between June 5 and July 15, 2026, sorted by crawler category.
What is not in this article is how those 20 sites are built to earn that attention in the first place. That part is the point, and that part stays ours. The data is the proof. The method is the moat.
What this means for you
If you cannot see your own crawler data, you are making decisions about AI visibility blind. You do not know which of the 29 crawlers are on your pages, whether the ones hitting you are training bots or search bots, or whether the pages you care about are being read at all.
That gap is the whole reason this data is rare. Almost nobody is watching. We are, across 20 sites, and it is already rewriting what we thought we knew about how AI reads the web.
Want to act on this? Compare the best AI visibility tools, or start reading your own crawler data.