26% of UK adults trust ChatGPT for financial advice. The FCA counted.

Two regulators, two surveys, one uncomfortable pair of numbers. The first is what your clients are doing. The second is what your firm is doing. Neither is what the compliance manual assumes.

What did the FCA actually find?

The review commissioned a nationally representative survey of 5,026 UK adults in April 2026. Its executive summary reports that “around 26% trust general-purpose tools such as ChatGPT, Claude or Gemini for financial advice, despite limited awareness that formal routes to recourse will not apply”. It also reports that one in five UK adults is already open to AI making decisions for them, “with demand strongest where choices feel complex or high-stakes, particularly debt advice, pensions and investments”. The press release put the fifth at eleven million people.

Read that second sentence again. The appetite for handing a decision to a machine is strongest in exactly the areas — pensions, investments, debt — where regulated advice exists because the decisions are hard to undo.

What is the recourse gap?

Regulated advice from an authorised firm comes with the Financial Ombudsman Service and the Financial Services Compensation Scheme behind it. A general-purpose chatbot is not an authorised firm and does not give regulated advice; the review’s phrase is that formal routes to recourse “will not apply”. A client who acted on a chatbot’s suggestion about a pension transfer has nobody to complain to and nobody to compensate them. The review’s first recommendation, of seven, is to secure and adapt the regulatory perimeter — which is the regulator saying out loud that the line between a tool and an adviser is no longer where the rulebook drew it.

And what are firms doing?

The Bank of England and the FCA surveyed 118 firms in 2024 and published the results that November. 75% were already using AI, with another 10% planning to within three years. A third of all use cases were third-party implementations, and the top three providers accounted for 73% of cloud, 44% of models and 33% of data — the concentration that the review and the Treasury Committee both worry about. 55% of use cases involved some automated decision-making, though only 2% were fully autonomous. And 46% of firms reported only a partial understanding of the AI technologies they use, against 34% who said their understanding was complete.

The Treasury Committee, reporting on 20 January 2026, cited the same 75% and asked the FCA to publish comprehensive guidance on applying its existing rules to AI, including accountability under the Senior Managers and Certification Regime, by the end of 2026. It also asked the Treasury to designate the major AI and cloud providers as critical third parties by the same date; as of October 2025 no firm had yet been brought into that regime.

What this means for an advice firm

Three things follow, and none of them is “buy a chatbot”.

1. Ask what the client has already been told. A quarter of adults trust a general-purpose tool; the fact-find should treat “I asked ChatGPT” the way it treated “I read it on a forum” — as a belief to be surfaced, dated and addressed in the suitability report, not ignored. The tool’s answer has no recourse; yours does, which is the whole of your value in that conversation.
2. Know which of your own systems decides anything. Half of firm use cases involve some automated decision-making, and nearly half of firms only partly understand their AI. If a system in your firm scores, ranks, drafts or chases, somebody should be able to say what it did, from what inputs, and who confirmed it. In a regulated firm the audit trail is the product, and the SM&CR question the Committee wants answered is “who was accountable for that decision?”
3. Expect the rules to be written for you if you do not write them first. The Committee has asked for guidance by the end of 2026; the review has asked for the perimeter to move. A firm that can already show what its AI does, where a human sat, and what it refused to do will find that guidance describes what it already does. That record is a thing you can build now.

The client-side number will not go down. The firm-side number will not go down either. What changes is whether, when a client asks why they should pay for advice a chatbot gave them for free, the answer is a sentence about recourse and a file that proves it.