Google has strengthened its AI search policies following a public test by BBC journalist Thomas Germain. In February, Germain showed that generative models could be manipulated within 20 minutes: Gemini and ChatGPT produced misleading outputs, including a joke labeling Germain as a “hot-dog eating champion.”

The company now classifies such manipulations under its anti-spam policy, extending traditional SEO enforcement to AI search. Measures include content filtering, model behavior adjustments, and ongoing testing against “prompt injections” — hidden instructions embedded in webpages, documents, or emails that can distort AI responses.

Google emphasizes that AI use is allowed as long as the content serves user interests rather than manipulative purposes. In May, the company launched Gemini 3.5 Flash, its most powerful AI model for programming and agent creation.