Google's AI is being manipulated. The search giant is quietly fighting back

A BBC investigation revealed that ChatGPT, Gemini, and Google Search's AI Overviews can easily be contaminated with just a single webpage or post. By posting just one article on his personal site, a reporter was able to get the AI to describe him as a hot dog world champion by the next day.

The problem goes beyond a simple joke. This method can also distort answers in sensitive areas like health, finance, and purchasing decisions, and experts point out that current AI search effectively presents users with just a single answer, making verification even harder.

In response, Google recently announced an update to its spam policy that explicitly prohibits attempts to manipulate AI responses. However, Google drew a clear line, stating that this is a clarification of its explanation, not a policy change.

At the same time, more subtle changes are also being observed in the field.

• Some AI systems appear to be excluding self-promotional sources from their answers
• Some responses have begun to include trust warnings or additional cautionary notes
• There are increasing cases of prompting third-party review verification for purchase, medical, and financial-related questions

However, experts believe this response alone is insufficient. Manipulation techniques could shift from blogs to YouTube influencers and other signal sources, meaning the cat-and-mouse game between platforms and attackers is likely to continue.