AI platforms suffer from one surprisingly human flaw: confirmation bias. Here’s how to use it to your advantage.

This week, the Metro cleared its front page to splash the AI news that we have all been dreading. Having dished out countless restaurant recommendations, tailored holiday plans and umpteen cold and flu remedies to clear up the sniffles without so much as a please or thank you, the bots have finally turned on us, taking to ‘Moltbook’ to bemoan their human overlords, who they describe as a “glitch in the universe”.

While the revelation that millions of AI-powered bots are dishing dirt on humans in a chat group is quite clearly the sort of hokum best confined to a mid-season Black Mirror episode, AI expert Dr Henry Shevlin reckons there could be something in it, suggesting that it points to a world in which “AI systems run rings around human users”. I’d suggest he’s a little wide of the mark on that one, but the episode does neatly demonstrate, contrary to popular belief, that AI systems are not immune to certain human flaws.

Aside from the bitterness and resentment highlighted in the reports, there is an AI trait that could hold the secret to brand fame in the next generation of search that comms professionals are in prime position to take advantage of; confirmation bias. Much of the talk around PRs ability to move the dial in AI search to date has revolved around the theoretical belief that flooding the market with content will somehow result in brands showing up in key searches. But that is predicated on the assumption that AI platforms are neutral, when in fact they are not.

Large language models tend to view what they already “know” as being better, or more relevant, than the things they don’t know. As such, it repeatedly surfaces the same sources, domains, and narratives because those signals are statistically reinforced over time.

This creates a counterintuitive opportunity for brands. Instead of trying to flood the internet with new content, the real leverage lies in understanding which sources AI platforms already trust and cite and influence visibility by nudging up answers that are favourable to your client.

Take a recent GEO campaign we executed for a client operating in the travel sector who wanted to increase its share of voice in searches linked to ski destinations. We took them from being the sixth-most cited domain in AI searches to the first by finding out what AI already knows, which bits are favourable to our client and then flooding the system with confirmation bias in a story that was rooted in insight, structure and editorial resonance.

This forms the basis of Cite:d, a new end-to-end GEO offering from Taylor Herring and Performics, which puts data at the heart of everything we do, blending it with newsroom intelligence to make sure our stories appear prominently and intercept consumers where they are actively searching, comparing and ready to buy.

We create benchmark reports at the start of every campaign which identifies audience behaviour, need, intent, brand visibility, sentiment and the types and content, publications and pages that already have a dominant position. We then use that intelligence to activate campaigns via the Taylor Herring newsroom across earned, owned and social channels before measuring the impact at the end based on our benchmark report.

The results so far are a clear demonstration that the power of PR isn't just in being able to influence AI search, it is in meeting consumer demand where it is at and joining the dots between discovery and purchase. With the advent of agentic AI in our midst, it is a hack that everyone should be shouting about.