
OAI-AdsBot is a real OpenAI bot: it checks the pages advertisers submit as adverts on ChatGPT, and only those. This site has never advertised on ChatGPT, and this OAI-AdsBot wasn’t OpenAI’s: where its network was recorded, it came from Google’s cloud, not from OpenAI’s published addresses, and it only looked for weak spots in website software.
Worth knowing: OpenAI began testing adverts in ChatGPT on 9 February 2026, and by April its list of bots included OAI-AdsBot (Search Engine Journal). In August 2026 the security company GreyNoise reported scanners using forged AI crawler names, six of them from the same 824 addresses, asking for files of passwords and keys and never once for robots.txt (GreyNoise).
How it makes money: Not adverts but keys: scanning like this pays off when a site has left its secrets out. In 2024 one gang copied exposed .env files from at least 110,000 domains and used the cloud keys in them to empty storage and leave ransom notes. (unit42.paloaltonetworks.com)
On this site: Not a page, not an advert: only addresses that would give away a server’s insides, /dashboard, /proc, /fs and a hole in a developer’s tool (CVE-2025-68155). OpenAI’s name was a disguise.
A few are security companies mapping what is exposed on the internet and selling that picture. Most are automatic attempts to find a site with a door left open: a leaked password file, a WordPress login, an admin page. A site that is broken into can be resold, used to send spam, or added to a network of hijacked machines. New sites are found within seconds, because every new site’s security certificate is written into a public log (SANS). What they sell →