How a Chatbot became Britain’s new favourite comparison site

A few years ago, someone wanting a better savings account began the classic ritual. They opened a comparison site, sorted by headline rate, squinted at the small print, and clicked through to whichever provider topped the table.

The journey typically began in a search box and ended on a bank’s own website. As a meerkat would say -‘simples’. But that ritual is quietly dissolving.

Increasingly, the journey begins and ends in a conversation with a machine, and the bank never sees the customer until the account is already open.

More than half of UK adults, 54%, now use AI tools such as ChatGPT, Copilot or Gemini, rising to 79% of those aged 16 to 24.

Ofcom, which published those figures in April, found that AI-generated answers are increasingly replacing the older pattern of searching, clicking and browsing, with 42% of people who see AI search summaries now reading them often or always.

What began as a tool for drafting emails and planning kitchens has become, for a large slice of the country, the first place they go to ask a question. And some of those questions are about money.

A recent EY survey of consumers across 23 countries captured the rapid pace of change. Roughly half of respondents had used AI in the previous six months to help with savings and investment decisions, a figure that climbed to 68% among Gen Z and 65% among millennials.

In the UK sample, around 35% had turned to AI for savings and investment, with ChatGPT the most popular tool.

A separate Sky News investigation put the headline number higher still. It found that 40% of Britons now use AI chatbots for financial advice, a trend driven partly by a widening gap in affordable human help.

Why banking is unusually exposed

Not every purchase is up for grabs in this way. You do not deliberate over a chocolate bar. But financial products are researched, not grabbed off a shelf, and a recent report from Boston Consulting Group and FT Partners argues that this makes banking especially vulnerable to the change.

Accounts, cards, loans and investment tools are chosen through comparison, context and trust, the report notes, and the decisive question is shifting. It is no longer only whether a product ranks highly in a list, but whether a model will retrieve it, trust it and recommend it. In the report’s words, discoverability still matters, but recommendation-worthiness matters more.

That has turned a decade of marketing orthodoxy on its head. The discipline once known as search engine optimisation, the craft of ranking on Google, is giving way to what the industry now calls generative engine optimisation: making a product legible and trustworthy to the AI systems that increasingly sit between a customer and a decision.

The same forces are reshaping discovery far beyond finance, as HotMinute reported when Google overhauled its search box to lead with AI answers rather than blue links, and when YouTube and TikTok overtook Google as the default discovery channel for UK Gen Z shoppers. Banking is simply the latest, and highest-stakes, version of the same story.

The strategic worry for providers is what the BCG report calls the “zero-click” journey. A customer can form a firm preference, and rule out three rivals, without ever landing on a single bank’s website.

The recommendation comes to them, pre-digested, through an intermediary they did not choose and cannot audit. Tellingly, the report says conversational advertising is already emerging as a distinct budget line, separate from search and programmatic spend, with more than half of organisations setting money aside to appear inside these answers.

In other words, the race to influence the machine’s recommendation has already begun, and consumers cannot see it happening.

What happens when it goes wrong?

Here is the part the chatbot rarely volunteers: when it recommends a bank, nobody is on the hook if it is wrong. The Financial Conduct Authority has been blunt on this point.

General-purpose tools such as ChatGPT are not regulated, and any advice they give is not covered by the Financial Ombudsman Service or the Financial Services Compensation Scheme; the regulator advises consumers to check other trusted sources before acting.

The safety net that customers instinctively assume sits behind financial guidance simply is not there when the guidance comes from a chatbot.

Nor is the advice reliably good. When Sky News tested ChatGPT, Copilot and Gemini with a hypothetical £16,000 in savings, a Hargreaves Lansdown strategist found all three produced recommendations that were US-biased and incomplete, with no recourse if a customer acted on them.

The pull towards these tools is partly a symptom of a shrinking alternative. The share of human financial advisers willing to take on clients with less than £50,000 to invest has roughly halved over six years, from 52% to 25%, pushing younger savers online by default.

Regulators are scrambling to map the territory rather than fence it off. In January the FCA launched the Mills Review, a long-term study of how AI will reshape retail finance, due to report this summer.

Its framing is instructive: it splits the future into assistive AI, the explain-and-compare tools available today; advisory AI, systems that actively nudge people to switch supplier or refinance; and autonomous AI, agents that would move money and even switch current accounts within boundaries the customer has set.

The FCA has also been told to publish practical guidance by the end of 2026 on how existing consumer-protection rules apply to AI. Until it does, Britain is choosing where to keep its money with the help of a tool the rules have not yet caught up with.

What it means for the rest of us

The banks themselves are not standing still. As HotMinute noted earlier this year, one in five queries to NatWest’s chatbot Cora are now handled by AI, a sign that the technology is being built into the relationship at both ends, the choosing and the using.

The likely winners, the BCG report suggests, will be the firms that treat this as a product and distribution problem, not merely a marketing one, structuring their information so a model can trust it and reducing their dependence on discovery channels they no longer control.

For everyone else, the practical takeaway is smaller and older than it sounds.

The comparison table you used to scroll has not vanished; it has been replaced by a single, confident answer with the workings hidden. That answer can be a genuinely useful starting point. It should not be the finish line.

The rate is still worth checking against a regulated comparison site, the protection still worth confirming on the provider’s own pages, and the recommendation still worth treating as a lead to verify rather than a verdict to act on.

The machine is good at sounding certain, but it is not the one whose savings are on the line.

Now read: The UK now levies 90 taxes – the most since 1834