AI bots read our website 22 times as often as people did
On October 19 we switched on AI crawler counting for every Buildsy site, including our own. I expected a trickle. Over the next 23 days, AI crawlers fetched pages on buildsy.io 1,522 times. In the same period, people viewed 69 pages across 44 visits. That is 22 bot fetches for every page a person read.
The numbers
From October 19 to November 10, 2026, on buildsy.io:
- 1,522 page fetches by AI crawlers
- 69 page views by people, across 44 visits
- 0 visits referred by an AI assistant
- 0 leads from an AI assistant
Who the bots were
We watch for 17 crawler names used by OpenAI, Anthropic, Perplexity, Meta, Amazon, ByteDance and Common Crawl. Nine of them showed up.
- Meta-ExternalAgent (Meta): 635
- Bytespider (ByteDance): 599
- Amazonbot (Amazon): 129
- ClaudeBot (Anthropic): 79
- OAI-SearchBot (OpenAI): 33
- ChatGPT-User (OpenAI): 22
- CCBot (Common Crawl): 14
- Claude-User (Anthropic): 8
- Meta-WebIndexer (Meta): 3
Two companies, Meta and ByteDance, made 81% of all the fetches. OpenAI's training crawler, GPTBot, never came at all, and neither did Perplexity's.
What they wanted
Not all AI crawling is the same. We sort each crawler into one of three jobs.
Training: 1,456 fetches (96%). Crawlers that gather text for building AI models.
Search index: 36 fetches (2%). Crawlers that build the index an assistant searches when someone asks it a question.
Live fetch: 30 fetches (2%). A person asked ChatGPT or Claude something, and the assistant opened one of our pages to answer.
The live fetches are the ones I care about. Each one happened because a real person asked a question. There were 30 in 23 days: 22 through ChatGPT and 8 through Claude.
What the bots read
The home page was fetched 236 times and the pricing page 126 times. After that came our About page (82) and the help guides we write for clients (77 for the guide index alone). The bots were far more interested in pricing than people were: 126 fetches against 9 human views.
What this means if you run a small business
AI assistants are reading your site, but they are not sending you customers yet. In 23 days, not one visit to buildsy.io came from an AI assistant. Search engines sent 11 of our 44 visits. The rest came directly, from social media or from links we shared.
What the bots carry away is your facts. When someone asks an assistant for a roofer nearby who does metal roofs, the assistant can only use what it has read. So put the basics on the page as plain text: your services, your service area, your hours, and your prices if you publish them. A fact locked inside an image or a PDF is much harder for a bot to pick up.
You can say no. A robots.txt file lets you ask AI crawlers to stay away, and the well-behaved ones listen. The trade-off is real: block them, and an assistant cannot quote your site when someone asks about you. We leave ours open.
The fine print
This is one small website over 23 days, not a study. A crawler is identified by the name it gives, and a name can be faked. We count page requests only, so fetches of files such as sitemaps are not included. And buildsy.io is a new site with little traffic, so the ratio will look different on a busy one.
See it on your own site
Every Buildsy site gets this same count in its admin, next to its visits, calls and leads, with one switch to allow or refuse AI crawlers. If you want to know what the bots are doing on your site, get in touch.