Is that really ChatGPT crawling your site? Fake AI bots are hunting for passwords
Not always. A bot can call itself GPTBot or ClaudeBot just by typing that name. Cequence found one scanner using about 32 fake AI bot identities to look for files that hold passwords and API keys, and on one day in October it reached 48% of the Cequence-protected organizations that saw any AI bot traffic. Keep letting real AI crawlers in, but check who they are before trusting the name, and make sure no secret files can be opened from the web.
- One scanner rotated through about 32 AI bot identities, including crawlers and assistants from OpenAI, Anthropic, Meta and Perplexity.
- It skipped normal pages and asked for about 90 file paths that often hold secrets: .env configuration files, cloud credential files and private SSH keys.
- It used more than 300 cloud addresses in one day and nearly 400 in a week, so IP blocklists went stale fast.
- At several organizations it made up more than 80% of the traffic labelled as AI bots.
What happened
On 28 October, Cequence’s CQ Prime research team published its findings on AI agent impersonation. The scanner was active from at least late September to late October and hit retail, financial services, telecoms, energy and software companies. Nearly all of its requests were blocked or failed, and it gave itself away: the same machine claimed to be OpenAI, Anthropic, Meta and Perplexity within seconds. Dark Reading and CSO Online also covered the report.
What this means for your site or store
You want real AI crawlers to read your pages, because that is how ChatGPT, Claude and Perplexity find and recommend you. You don’t want anyone to walk in just by using their name. Three things to check:
- No secrets on the web: try opening yourdomain.com/.env in a browser. Anything other than a “not found” or “forbidden” page is urgent: move the file and change every key in it. Check staging and admin sites too.
- Don’t trust the name alone: if a firewall, CDN or plugin rule lets “GPTBot” through by name only, make it also check the address.
- Read your AI bot numbers with care: if most of your “AI traffic” is asking for config files, it isn’t AI.
The technical details
Real AI crawlers can be checked. OpenAI publishes its crawler IP ranges, linked from its crawler docs, and Google explains how to verify its crawlers with a reverse DNS lookup. Cequence also recommends signed requests and judging bots by what they do, since an “AI assistant” asking for SSH keys is not doing an AI assistant’s job. Your robots.txt still decides which real bots may read your pages; our guide on allowing AI search bots while blocking AI training has the exact lines. Robots.txt is a request, not a lock, so it never protects secret files.
Want to see which AI bots can reach your site, and what they find? Run our free AI check, or get robots.txt, llms.txt and schema done for you with the AI-Ready Kit.
