Fraud Blocker lets advertisers choose which AI screens clicks

The click-fraud tool now checks ad traffic with OpenAI or Anthropic models of the customer’s choosing and explains every visit it blocks.

By Alexandre S. 28 Oct 2026, 04:40 UTC

Fraud Blocker, a Los Angeles company that blocks fake clicks and fake form leads on ads, launched an AI detection engine on 26 October 2026. Its customers now decide which AI model judges their traffic: one from OpenAI, one from Anthropic, or the one Fraud Blocker recommends, its announcement says. Each blocked visit gets a written explanation; the release’s example is a single IP address that made 31 clicks from 38 devices on 15 sites in under 15 minutes.

The company says the engine catches about 60 per cent more sophisticated invalid traffic, meaning bots built to act like shoppers, than tools that score fixed signals. The only basis given is “initial customer results”; The Agile Brand Guide points out that the release does not say how many customers, over what dates or against which tools. Fraud Blocker says it draws each year on more than 500 million ad clicks and more than 1 billion IP addresses from over 25,000 customer websites, and that traffic sent to the models is held under zero-retention controls and never used to train them.

Choosing the model also means choosing whose data rules apply. For its own Team plans, Anthropic does not train on your content by default, our Claude entry records (checked 27 October 2026); that is a seat plan, and the release does not say which terms cover the traffic Fraud Blocker sends to the models.

The 60 per cent gain rests on unpublished customer results, so the one part an advertiser can check today is the written reason on each blocked click, such as the IP with 31 clicks from 38 devices; the zero-retention promise is Fraud Blocker’s word in a press release.

Fraud Blocker lets advertisers choose which AI screens clicks

Sources