A BBC investigation shows how a single, well-crafted web post can skew answers in leading AI products, including Google’s AI Overviews, OpenAI’s ChatGPT and Anthropic’s Claude—steering users toward biased or false claims on health, finance and more. Google calls its recent spam-policy edits a clarification, but search experts say the company is quietly tightening defenses, from downranking suspected self-promotion to adding caveats and confidence labels. The stakes are high: more than a billion people use chatbots, and Google says 2.5 billion users see AI Overviews monthly. SEO consultants warn that manipulators will keep evolving tactics—shifting from blogspam to influencer content on platforms like YouTube, which AI systems now cite. For consumers and businesses, it’s an arms race that could reshape traffic, trust and liability online—and, for now, the safest stance is skepticism toward any single “AI answer.”
Related articles:
Google’s March 2024 core update and new spam policies
Indirect Prompt Injection Attacks on Large Language Models
