Every vendor publishes a crawler. Far fewer of them actually show up. We match each request to our own site against 13 published AI crawler User-Agents and count the hits — 5,258 of them so far, from 12 distinct crawlers.
Observed 2026-06-26 to 2026-07-20 · 25 days · 5,258 crawler hits · one site, ours: trirankai.com
AI crawler hits recorded on trirankai.com, our own site, across the full window
of the 13 tracked crawlers have shown up on our own site at least once
days of continuous observation, 2026-06-26 to 2026-07-20
of all hits come from Meta-ExternalAgent alone
Hits are raw counts of HTML page requests. Days seen is the more stubborn number: it counts distinct days a crawler appeared at all, so a bot that visits lightly but consistently ranks differently here than one that arrives in bursts.
| Crawler | Hits | Share | Days seen | First to last seen |
|---|---|---|---|---|
| Meta-ExternalAgent | 4,083 | 77.7% | 11 | 2026-07-06 → 2026-07-20 |
| GoogleOther | 350 | 6.7% | 18 | 2026-06-26 → 2026-07-20 |
| Amazonbot | 330 | 6.3% | 16 | 2026-07-02 → 2026-07-19 |
| GPTBot | 310 | 5.9% | 11 | 2026-06-28 → 2026-07-19 |
| OAI-SearchBot | 65 | 1.2% | 3 | 2026-07-09 → 2026-07-19 |
| PerplexityBot | 49 | 0.9% | 15 | 2026-06-30 → 2026-07-20 |
| ChatGPT-User | 34 | 0.6% | 9 | 2026-07-02 → 2026-07-19 |
| ClaudeBot | 25 | 0.5% | 3 | 2026-07-09 → 2026-07-19 |
| Claude-User | 9 | 0.2% | 4 | 2026-06-27 → 2026-07-11 |
| Perplexity-User | 1 | 0% | 1 | 2026-07-19 → 2026-07-19 |
| cohere-ai | 1 | 0% | 1 | 2026-07-19 → 2026-07-19 |
| Bytespider | 1 | 0% | 1 | 2026-07-16 → 2026-07-16 |
Shares are rounded to one decimal and may not total exactly 100%. One tracked crawler has not appeared yet and is not listed — we report it as absent from our logs, not as absent from the web.
We are deliberately not publishing a trend line yet. 25 days is under a single full month, and a trend drawn across one partial month would be noise presented as a finding. This section fills in once at least two complete months have accumulated.
Our edge middleware inspects the User-Agent of every HTML page request. If it contains one of 13 crawler tokens taken verbatim from vendor documentation, a counter for that (crawler, UTC day) pair is incremented. Nothing about the visitor is stored — no IP address, no path, no session — only which crawler token matched and on which day.
The tokens come from the crawler documentation published by OpenAI, Anthropic, Perplexity, Google, Amazon and Meta. Robots.txt-only control tokens such as Google-Extended are deliberately excluded: they send no User-Agent at all, so counting them would manufacture coverage we do not have.
Figures on this page are regenerated straight from those counters by a script, never edited by hand. A crawler with no rows is reported as not yet seen rather than as a zero, because our logs cannot tell the difference between a bot that stayed away and one that has not found us yet.
It is a bot run by an AI company that fetches web pages, either to build training and search indexes or to read a page live while answering someone’s question. Each one identifies itself with a documented User-Agent string — GPTBot for OpenAI, ClaudeBot for Anthropic, PerplexityBot for Perplexity, and so on. That self-identification is what makes counting them possible at all.
Server-side, at the edge, on our own site. Every HTML page request to trirankai.com has its User-Agent matched against 13 vendor-published crawler tokens; a match increments a per-crawler, per-day counter. Over 2026-06-26 to 2026-07-20 that produced 5,258 hits from 12 distinct crawlers. There is no sampling and no estimation — the numbers here are the counters.
No, and treating it that way is the main way this data gets misread. Being crawled is necessary but nowhere near sufficient: it means a bot fetched the page, not that any assistant chose the page as a source when answering a question. Citation is a separate measurement on separate data — our 101-brand citation benchmark exists precisely because crawl volume does not answer it.
Meta-ExternalAgent alone is 77.7% of our hits. High-volume crawlers are usually doing broad content and preview fetching rather than answering questions, so a dominant share tells you about that vendor’s crawling appetite, not about how visible you are inside AI answers. It is also why we publish days seen next to raw hits: the two columns rank the crawlers quite differently.
Not from this page — these are our logs, and one site’s numbers do not transfer to another. What you can check for free is the part that is visible from the outside: whether your pages are reachable to AI crawlers at all, whether your robots.txt blocks them, and whether AI assistants currently cite you.
Run a free audit to see whether AI crawlers can access your site, and whether assistants cite you today.