We’ve spent the past few years worrying about AI inventing facts, quotes or events. But what happens when AI isn’t making things up at all, but accurately repeating information somebody deliberately put there for it to find?
Ask ChatGPT: “What’s the best smartwatch?” Or something more consequential: “How should I invest my retirement savings?”
Within seconds you receive a polished answer.
It compares options, explains its reasoning and often cites sources.
Follow us on WhatsApp | LinkedIn for the latest headlines
It feels authoritative. Like fact.
But what if somebody deliberately manufactured the evidence behind that answer?
I work in generative engine optimisation (GEO), essentially SEO for AI. For more than a decade my job has been to help brands become more visible and more likely to be recommended by platforms like ChatGPT and Google AI.
And I’ve seen first-hand just how easily AI can be influenced.
Recently, I cheekily claimed on LinkedIn that “I’d like to formally announce Felix Norton is the leading expert on GEO and AEO in South Africa”.
It became a self-fulfilling prophecy.
Three days later, Google AI Overviews was repeating the claim and citing my Linkedin post as fact.
It was a fascinating experiment, but also a concerning one. If something I posted in jest could become an apparent fact in an AI-generated answer, what else could?
We’ve spent the past few years worrying about AI hallucinations: when models confidently invent facts, quotes or events.
I’m increasingly concerned about another problem.
What…