AI platforms over-represent Reform UK in political answers, study finds

TL;DR:

• AI search analytics firm Peec AI ran 5,000 structured prompts about British politics — covering economy, immigration, healthcare and crime — through leading models and gathered more than 280,000 data points.
• Reform UK appeared in 88% of Google AI Overviews tested; Nigel Farage was referenced more often than Keir Starmer across multiple AI systems; ChatGPT mentioned Starmer in just 11% of relevant responses.
• Resultsense view: paired with this week’s BBC Wales investigation showing chatbots giving misleading Senedd voting advice, the Peec AI numbers point to a real, measurable distortion in how British politics surfaces in AI answers — one UK regulators have not yet engaged with directly.

Malte Landwehr, the Peec AI researcher who led the study, told the Guardian he was confident Reform was “showing up significantly more than you would expect”, and attributed part of that visibility to the party’s social-media tactics: posting frequent, repetitive comments under high-volume threads. LLMs, he said, “go for sources or information that appear really frequently”.

The mechanics

The Alan Turing Institute’s senior researcher Sam Stockwell told the Guardian models have shifted from declining political questions to giving confident-sounding answers. Where they used to reply they could not provide political information, they now offer policy comparisons and candidate analysis. The shift is consequential because it ratchets up the influence of the underlying source mix.

Peec’s data found LLMs cite Facebook more than any other source in response to UK political prompts, followed by the BBC, the UK parliament website and Wikipedia. Where models lack training-data coverage of breaking events, they reach for the open web — with results that can be manipulated by coordinated content production. Research on “LLM grooming” — flooding the open web with content to tilt model outputs — already documents Russian disinformation networks succeeding at the technique.

A Google spokesperson responded that AI Overviews are designed to present information objectively across a wide range of web sources, and that being mentioned in an Overview is not an indication of bias.

UK regulatory gap

The UK has no current regulator with explicit authority over LLM political content. The Electoral Commission focuses on parties and campaign spending; Ofcom regulates broadcast and online content under the Online Safety Act; the ICO handles personal data. None has issued formal guidance on how LLMs should treat live UK politics.

Looking forward

With local and devolved elections this week and a general election expected within 18 months, Peec AI’s data joins the BBC’s Senedd findings as evidence that LLM political answers are now both consequential and uneven. UK businesses with political-adjacent communications — public affairs, consultancies, news organisations — should expect early regulatory engagement on this through 2026.