A growing wave of AI-powered ‘bill-cutting’ assistants that promise to find households the cheapest energy deals may be quietly steering users towards suppliers that pay referral fees, according to a new investigation by Citizens Advice. The charity claims that several of the most widely used consumer tools failed to surface genuinely cheaper tariffs, leaving some households as much as £180 a year worse off than if they had switched to the market’s lowest available rate.
The findings, published this week, land at a sensitive moment for both consumers and regulators. With energy prices still well above pre-2022 levels and AI chatbots increasingly marketed as effortless money-savers, Citizens Advice warns that opaque commercial incentives could undermine trust in a technology many households are only beginning to rely on. The charity is now urging Ofgem and the Competition and Markets Authority (CMA) to consider mandatory disclosure rules for AI-driven recommendation engines.
What the investigation found
Citizens Advice researchers tested a sample of consumer-facing AI switching assistants over a six-week period, feeding each tool identical household profiles ranging from single-occupancy flats to larger family homes on standard variable tariffs. In a significant proportion of cases, the charity says, the ‘recommended’ tariff was not the cheapest on the market — and in several instances, the supplier highlighted by the assistant had a commercial arrangement with the tool’s operator.
The gap between the AI’s top suggestion and the genuinely cheapest deal ranged from a few pounds to £180 a year, depending on the household’s usage profile. The disparity was most pronounced, researchers noted, for higher-consumption homes where small per-unit differences compound quickly across a year.
“What concerns us most is not that these tools earn money — that’s a legitimate business model — but that the financial relationship is almost entirely invisible to the user,” said Priya Mehta, a senior policy researcher at Citizens Advice. “People are being told they’re getting the best deal when, in reality, they’re getting the best deal for the platform.”
The charity stressed that not every tool behaved this way, and that some clearly labelled sponsored results. But it argued that the inconsistency itself is the problem: consumers have no reliable way of knowing whether an AI recommendation reflects the market or a commercial deal.
How the AI tools work — and where the incentives lie
Most consumer switching assistants combine a conversational interface with a back-end comparison engine. Users describe their household and usage, and the tool returns a tailored recommendation in plain language. The convenience is precisely the appeal — but it also conceals the underlying logic.
Unlike traditional price-comparison websites, which typically display a ranked list users can scroll through, AI assistants often present a single confident answer. That design choice, analysts say, magnifies the impact of any embedded bias.
“A ranked table invites scepticism — you can see what’s below the top result,” said Dr Callum Reeves, an algorithmic-transparency researcher at the fictional Centre for Digital Markets. “A chatbot that says ‘I’d recommend Supplier X’ carries an authority that’s very hard for an ordinary consumer to interrogate. The interface is doing persuasive work, whether or not that’s intended.”
Referral fees are common across the comparison sector and are not illegal. The issue, regulators have previously signalled, is whether consumers are given enough information to understand that a recommendation may be commercially influenced.
Pressure builds on Ofgem and the CMA
The Citizens Advice report explicitly calls for clearer rules. Among its recommendations:
- Mandatory, prominent disclosure when an AI recommendation is linked to a referral fee or commercial partnership
- A requirement to show users the genuinely cheapest available tariff, even where it is not the ‘recommended’ option
- Independent auditing of recommendation algorithms used in regulated markets such as energy
Ofgem has previously taken action against misleading switching practices, and the CMA has been increasingly vocal about the risks of AI in consumer markets. A spokesperson for the CMA said the regulator was “aware of the concerns raised” and that AI-driven recommendation systems “fall squarely within our interest in fair and transparent digital markets.”
Industry voices were more cautious. A representative for one switching platform, speaking on background, argued that referral models keep tools free to use and that blanket rules could push providers towards subscription fees. But consumer advocates counter that transparency and affordability are not mutually exclusive.
The trust problem for AI consumer tools
Beyond energy, the report raises a broader question about AI assistants positioned as impartial advisers. As similar tools expand into broadband, insurance and banking, the same tension between helpfulness and hidden incentives is likely to recur.
“This is a test case for AI in everyday life,” Mehta said. “If the first generation of mass-market AI advisers turns out to be quietly monetised, the reputational damage could be lasting.”
What this means
For households, the immediate takeaway is to treat AI switching recommendations as a starting point rather than a verdict — cross-checking against Ofgem-accredited comparison sites before committing. For the industry, the report signals that the era of unscrutinised AI ‘advice’ may be ending. If Ofgem and the CMA act on Citizens Advice’s recommendations, disclosure rules for AI recommendation engines could become a regulatory norm, reshaping how these tools are built, marketed and trusted across the UK consumer economy.
Photo by Helena Lopes on Pexels
