Some agents are customers. Most are not.
A person asking ChatGPT to find a gift is a shopper. A cloud browser on a home address that reads every price is not. Reputable tells them apart, and shows you how it knows.
3×
more AI-agent sessions than a year ago
58%
of agent and bot traffic was agents you want
42%
never named itself, or lied about who it was
Sample store. Illustrative figures, not customer data.
Agents you want.
They name themselves, and we check the claim against the operator's published network.
AgentHow we identify itShare of agent traffic
Googlebot, Bingbot
Search crawlers
Verified against the operator's published network
31%
ChatGPT agent
OpenAI
Names itself in a signed-request header
9%
ChatGPT-User, Claude-User, Perplexity-User
Fetches a person asked for
Verified network, or named in the user agent
14%
Google-Agent
Google
Named in the user agent, verified network
4%
Agents you don't.
They never say who they are. Several rent home internet connections and solve CAPTCHAs, so edge bot scores pass them. We flag them from connection evidence and behaviour, on the visit and across visits, with a likelihood rather than a name.
AgentHow we identify itShare of agent traffic
Cloud browsers on residential proxies
Real Chrome, a new home address per visit, solves the CAPTCHA
21%
Undeclared agentic browsers
Drive the person's own browser or a stock Chromium, never name themselves
12%
AI training scrapers
Ignore robots.txt, rotate networks, read every product page
7%
Crawler impostors
Claim to be Googlebot or GPTBot from networks those operators never use
2%
You set the policy.
- Let it in
- Verified crawlers and signed agents. They bring customers and keep you in AI answers.
- Count it apart
- Agents shopping for a person. Keep them out of your conversion rate, and keep their orders.
- Slow it down
- Training scrapers that ignore robots.txt. Rate limit them, or send them a smaller page.
- Stop it
- Cloud browsers on residential proxies, and impostors. Challenge them, or block them at your edge.