Banking customer trust 2026: why AI personalisation stalls

Customers say they want banks that know them. However, they are far less sure they want an AI acting on their behalf. That gap sits at the centre of banking customer trust 2026. It matters now because banks, fintechs and insurers are rolling out AI assistants quickly. They are testing how customers feel about them far more slowly.

Public surveys from technology vendors, consultancies and Gallup point in the same direction. For example, trust in the bank as an institution is holding up. However, trust in AI in banking lags well behind it. This is most true where the AI makes decisions or takes actions for the customer. Our reading is simple. AI personalisation in banking will be won on control and transparency, not on the sophistication of the technology.

This insight sets out what the published evidence suggests and where the gap comes from. It also covers what retail banking, fintech and insurance insight teams should measure before scaling AI-led personalisation. As a result, we have kept to what the public sources support. We have also flagged where you need your own data.

Banking customer trust 2026: high in general, uneven by product and generation

The headline picture is encouraging. Recent coverage in the ABA Banking Journal describes a banking trust survey. In it, banks regain status as one of the most trusted institutions. In addition, trackers such as the Fairer Finance Trust in Banking Index show how that trust moves over time.

However, a general figure hides a lot. Customer trust in financial services is not one thing. For example, a customer may trust a bank to keep deposits safe. Yet they may doubt it on lending decisions, advice, fees or how it uses their data. Insurance adds another layer, because trust there often depends on claims experience rather than day-to-day service.

Generation matters too. The MX research on the next generation of financial trust is a useful starting point. It helps frame how younger customers think about trust differently. Younger customers tend to be more comfortable with digital channels and app-based providers. However, that does not mean they extend the same comfort to automated decisions about their money.

For insight leaders, the lesson is to stop reporting a single trust score. Instead, break it down by product, by channel and by the type of action being taken. The picture then becomes much more useful.

Financial strain and thin margins: the context behind customer behaviour

Trust does not exist in a vacuum. Many households are managing tight budgets, and that shapes how they judge their bank. When money is short, a fee, a declined payment or a poorly timed offer feels personal. As a result, customers become more alert to anything that looks like the bank benefiting at their expense.

Banks face their own pressure. The Integris banking trust and technology outlook describes rising technology spending alongside fragile trust. The Accenture Banking Top Trends FY26 report sets out the shifts banks are planning around. Put those together and you see the tension. Institutions are investing in AI to protect margins and serve customers more efficiently. Meanwhile, customers are watching closely for signs that efficiency means less care.

This is the same behavioural shift we have seen in consumer goods. There, financial anxiety changes how people weigh value and risk. Our piece on the global rise of the value seeker covers that pattern. Similarly, in banking, the equivalent is a customer who scrutinises every automated nudge.

The practical point: customers will read personalisation that feels like selling as pressure, however accurate the targeting.

The personalisation paradox: wanting to be known, not managed by AI

Customers are not asking banks to know less about them. They want relevant service, fewer irrelevant messages, and a bank that remembers their situation. However, the paradox appears in how the bank uses that knowledge.

Being known feels good when it saves effort and the customer can see why. However, being managed feels uncomfortable when the bank makes decisions out of sight. For example, a reminder that a bill is due is welcome. An unexplained change to a credit limit, or a product suggestion that seems to know too much, is not.

This is why trust in the institution and trust in AI diverge. For example, a customer may trust their bank enough to hold their salary there. They may still not trust an AI agent to move money, switch products or negotiate on their behalf. Banks build institutional trust over years, and regulation, branches and a name back it. An AI assistant, however, has to earn its own trust, and the bank's reputation only partly transfers.

It helps to separate three levels of AI involvement, because customers respond to them differently:

• AI that informs: summaries, spending insights and alerts the customer reads and acts on.
• AI that recommends: suggested products or next steps, with the customer deciding.
• AI that acts: moving money, switching providers or making changes on the customer's behalf.

Therefore, we expect acceptance to fall as you move down that list, and the drop is where personalisation programmes stall. That is a hypothesis to test in your own market, not a finding to assume.

What is driving the gap: fraud fears, data breaches and AI errors

Three concerns come up repeatedly in public commentary on AI in financial services, and each one is rational.

Fraud fears

Customers know that scams are increasingly sophisticated, and many have either experienced fraud or know someone who has. As a result, an AI that can act on an account may look like another route in for a criminal. That is how it looks to a worried customer. Anything that blurs who is really contacting them, human or machine, bank or impostor, raises suspicion.

Data breaches

Personalisation depends on data, and customers understand that more data held means more data exposed. Every breach in the news, in any sector, reminds them of the trade-off. Therefore, the more a bank says it knows, the more a customer may ask how well it protects that knowledge.

AI errors

Customers have seen chatbots give wrong or confusing answers. In banking, however, the cost of an error is higher than in retail. A wrong payment, a misread instruction or a bad recommendation affects real money. Customers also want to know who is accountable when the AI gets it wrong. They also want to know whether they can reach a person quickly.

A better model alone does not resolve any of these concerns. Instead, design choices resolve them. They cover what the AI may do, what the customer can see, and how easily a mistake can be reversed.

What it means for banks, fintechs and challengers: control, consent and community

If trust is the constraint, the response is to give customers more say, not to hide the technology. Three themes stand out.

Control

Let customers decide how much AI does for them. For example, offer clear settings that separate information, recommendation and action. Make it easy to switch features off, to approve actions before they happen, and to undo them. As a result, control turns AI from something done to the customer into something the customer uses.

Consent and transparency

Explain in plain language what data the AI uses and why. Show the reason behind a recommendation at the moment it appears. Consent should be specific and revisitable, not a single tick box at onboarding. In addition, this fits the privacy rules that research and banking both operate under. Our guide to privacy-first research methodologies covers the same principles from the data collection side.

Community

Trust travels through people. Customers listen to friends, family and peers. They also respond to evidence that others like them have used a feature safely. Challengers and fintechs often have an advantage here because of strong user communities. However, established banks can build the equivalent through advisers, branch staff and visible human support alongside the AI.

Each type of provider starts from a different place. For example, incumbents have institutional trust but may be slower to show transparency. Fintechs and challengers, in contrast, are often more open about technology but have less history to lean on. Insurers face the added test of trust at the moment of a claim. In every case, the winner is the provider that makes the customer feel in charge.

Differences by market and generation worth testing

Trust and AI comfort do not travel evenly across borders. Many factors vary by market. These include regulation, the history of banking failures, the strength of local fintech, fraud experience and attitudes to data. As a result, a message that reassures customers in one country may feel intrusive in another.

We would test the following before assuming any pattern holds in your markets:

• Generation: whether younger customers accept AI that recommends but resist AI that acts, and whether older customers prefer human contact for the same tasks.
• Provider type: whether trust in AI differs between a high street bank, a digital-only challenger, a fintech app and an insurer.
• Product: whether comfort changes between everyday current accounts, savings, credit, investments and insurance claims.
• Market: whether the same AI feature earns different levels of trust under different regulatory and cultural conditions.
• Experience of fraud: whether customers who have been victims or near-victims react differently to automated actions.

Therefore, averages across markets and age groups will mislead you. The segments that matter are the ones where the gap between institutional trust and AI trust is widest. That is where personalisation is most likely to stall.

What to find out about your own customers and which research approach answers it

Public surveys tell you the direction. However, they cannot tell you how your customers feel about your assistant, in your app, handling your products. To understand banking customer trust 2026 in your own market, answer these questions. Each one has a method that suits it.

1. Where does trust in the bank end and trust in AI begin? A quantitative banking trust survey with a representative sample can measure trust by product, task and level of AI involvement. Using online data collection, it can also compare segments.
2. Why do customers hesitate? Depth interviews and focus groups uncover the reasoning behind fraud fears, data worries and error concerns. Good qualitative recruitment matters here. You need the right mix of ages, provider types and fraud experiences in the room.
3. What controls and explanations reassure them? For example, concept testing of consent screens, settings and explanations shows which design choices lift acceptance. It does so before you build them at scale.
4. Does it hold across markets? Multi-country work needs consistent questionnaires, translation and quality control, so that differences reflect customers, not method.
5. Is trust moving over time? Finally, a tracker lets you see whether transparency changes shift behaviour after launch.

Sample design deserves particular care. Hard-to-reach groups often hold the views that matter most. Examples are affluent consumers, small business owners and customers who have experienced fraud. In addition, rigorous screening and fraud controls keep poor-quality respondents from distorting a topic as sensitive as money.

If you are planning this kind of retail banking customer insight work, our financial services market research team can help. The team can help you design the sample, run the fieldwork and process the data.

Frequently asked questions

Do customers trust banks in 2026?

In general, yes. Recent survey coverage of banking customer trust 2026 suggests banks have regained standing as one of the more trusted institutions. However, trust varies by product, generation and market, so a single score can mislead.

Why does AI personalisation stall at trust?

Customers want to be known, but they trust AI acting for them less than they trust their bank. Fraud fears, data breaches and AI errors all feed that gap, and a better algorithm does not remove them.

How can banks build trust in AI assistants?

Give customers control over what the AI does. Explain what data it uses and why, and keep human support easy to reach. Then test these choices with your own customers before scaling.