One crawler made most of our bot traffic
Counters here keep robots in a column of their own. They are recorded, but they do not move the number on the badge unless the setting is switched on. That separation makes it easy to ask a question that is otherwise awkward: who actually crawls a page small enough to need a visit counter?
Thirteen days of data across every counter on the service, read on 27 August 2026: 47,841 robot hits in total.
| robot | hits | counters | hits per counter |
| Meta AI | 36,670 | 151 | 243 |
| Googlebot | 3,272 | 148 | 22 |
| Baiduspider | 3,258 | 345 | 9 |
| Bytespider | 419 | 94 | 4 |
| ClaudeBot | 225 | 29 | 8 |
| GPTBot | 97 | 86 | 1 |
One crawler is 76.6 % of all robot traffic on this service. And it is not the one with the widest reach — Baiduspider touches more than twice as many counters. Meta AI goes deep on few.
What it looks like day by day
It did not arrive gradually. 148 hits on 14 August, 613 the next day, then 8,311 on the 16th, and a few thousand a day ever since, spread across fifty to eighty counters.
The peak on a single counter on a single day was 2,488 hits from this one crawler.
Why that number matters to a small site
Consider the receiving end. A page normally gets a handful of visits a day. One morning it shows two and a half thousand. If robots were counted in with everyone else, the reasonable conclusion would be that something had happened — a link somewhere, a mention, an audience. Nothing had. A crawler had walked that site 2,488 times.
That is the whole argument for keeping robots in a separate column rather than filtering them out or folding them in. Folded in, they produce false excitement. Filtered out entirely, the question "was that real?" has no answer at all. Kept apart, the question has an answer, and it takes one glance.
The part we cannot explain
Why this crawler is so concentrated is not visible from here. It could be following one platform's link structure, or one sitemap, or the counters could simply sit on the kind of pages it is currently interested in. We can see what it did, not what it wanted.
Also worth stating plainly: this is thirteen days, because the robots table only starts on 14 August. It is a young table and it does not know last month. Reading a share off it as though it described a year would be exactly the mistake this service keeps writing about.