An analyst found 81.8% of traffic claiming to be from AI assistants was fake, with the Googlebot share worse. Bots self-report through user-agent strings that cost nothing to fake.
Claiming to be Googlebot costs nothing and proves nothing. The address is the only thing that does not lie.
Everyone is watching their AI assistant traffic right now, trying to work out how much ChatGPT, Claude and the rest are sending. An analyst decided to check whether that traffic was even real. It mostly was not. In his analysis, 81.8% of the traffic claiming to come from AI assistants was fake, and the Googlebot number was worse.
The reason is simple. Bots announce themselves through user-agent strings, names like ChatGPT-User, Claude-User and Googlebot. Claiming to be Googlebot costs nothing and proves nothing. The uniform is easy to put on. Anyone can send a request that says it is a friendly AI crawler when it is something else entirely.
The fix is to verify the address, not the name. Bot operators publish the actual IP ranges their crawlers use as plain files. A request is only legitimate if the name matches and the IP sits inside the published list. The IP is the proof. The name is just a claim.
Why it matters
A lot of businesses are about to make decisions off AI referral numbers, and a lot of those numbers are noise. If you are reporting how much traffic AI assistants send and you have not verified it, you are likely reporting fiction. Decisions made on fake data are worse than decisions made on no data, because they come with false confidence.
This is the same discipline that has always separated real measurement from grading your own homework. The number that matters is the one you can verify, not the one that happens to flatter the story you want to tell.
Of one analyst's "AI assistant" traffic turned out to be fake. The Googlebot share was worse (via SEJ)
What to do about it
Verify bot traffic by IP, not by name. Match the user-agent against the operator's published IP ranges before you trust it.
Do not report unverified AI referral numbers. If you cannot confirm the source, label it as unconfirmed rather than passing it off as fact.
Filter your analytics. Fake crawler traffic inflates your numbers and distorts every metric built on sessions.
Be sceptical of the AI traffic gold rush. Before you build a strategy on AI referrals, confirm the referrals are real.
Treat clean data as the job. Measurement is only worth something if the inputs are true.
The businesses that stay clear-eyed here will be the ones who check the address before they believe the name, while everyone else celebrates traffic that was never there.