Jo Dunning, Charity and Housing Account Director at Nexer Digital writes
People seeking help are increasingly turning to artificial intelligence before contacting a charity or advice service. For charitable organisations whose work depends on providing trusted and accurate information, this raises the question of how they ensure people find the right guidance when AI may be providing the first answer.
Research from Which? found that 51% of UK adults use AI search tools to look for products, services and advice online. Among those using AI search, 55% use it to seek knowledge or advice.
Ofcom has also highlighted the changing nature of online search. Its 2025 Online Nation report found that around 30% of keyword searches provide AI-supported overviews, while 53% of UK adults say they often see AI-generated summaries. ChatGPT recorded 1.8 billion UK visits during the first eight months of 2025, up from 368 million during the same period in 2024.
For charities, this matters because the problems people bring to them rarely have straightforward answers. Questions about debt, benefits, housing, immigration or health can all depend on details that the person seeking help may not realise are relevant.
Someone facing a housing problem, for example, may ask an AI tool whether their landlord can evict them or what rights they have as a tenant. The answer could depend on their type of tenancy, the notice they have received, where they live, the reason for the proposed eviction and what has already happened.
An adviser would typically establish those details before providing guidance. An AI system can respond only to the information included in the prompt. If an important detail is missing, its answer may still appear clear and authoritative even though it does not reflect the person’s circumstances.
At Camp Digital 2026, Shelter content designers Andy Parnell and Kate Belgrave discussed a year-long investigation into how AI is disrupting advice journeys. Their session considered how people use AI-generated housing advice and what this changing behaviour could mean for organisations providing trusted information and support.
The issue is therefore not simply whether an AI-generated answer is factually correct. It is whether it is appropriate to the individual situation.
There is another consideration, too. AI can sometimes reinforce the assumptions contained within a user’s question rather than challenge them.
Research from the UK AI Security Institute found that AI models became more likely to give user-affirming responses as the level of certainty expressed by the user increased. In an advice setting, this matters because an adviser would normally test the assumptions behind a question, whereas an AI system may reinforce them.
The advice journey is changing
Traditionally, someone with a problem might search online, find a trusted charity or advice organisation, read its guidance and then contact an adviser if they needed further help. AI can remove several of those steps.
A person can now describe their situation conversationally and receive an immediate response without necessarily visiting the organisation that produced the original guidance.
There are clear benefits to this. Conversational interfaces can make information easier to access, particularly for people who find complex websites or formal language difficult to navigate. Being able to explain a problem in everyday language and ask follow-up questions can feel considerably simpler than working through a large website. But there is a risk that the organisation providing the trusted information disappears from the journey.
That matters because charity and advice websites are rarely just repositories of answers. They can help someone identify related problems, understand their wider rights, find additional support or recognise that their situation requires professional advice.
A person may believe they need help claiming a particular benefit, for example, when their circumstances also involve debt, caring responsibilities, ill health or difficulty paying for essentials. Similar connections exist across many areas of charitable advice. A trained adviser can identify them by asking questions and considering the person’s circumstances as a whole.
What should charities do?
The answer is not to tell people to stop using AI. Nor is it realistic to assume that users will always follow the digital journeys organisations have designed for them. Instead, charities need to understand how those journeys are changing.
That means testing the kinds of questions people are asking AI about their areas of work and checking whether the answers reflect the organisation’s published guidance. It also means identifying where AI-generated information may be incomplete, misleading or lacking important context.
The role of the charity’s website may need to change, too. If fewer people arrive through traditional search, it needs to provide value beyond simply answering an initial question. It may need to help users verify information, understand their circumstances, discover related support and recognise when they need to speak to a person. This also makes accessibility an important consideration too.
For some people, AI may lower barriers to finding information. For others, a confident but incomplete response could create a new barrier by presenting information as authoritative without making clear what is missing.
Charities should consider testing common questions in widely used AI tools, particularly in areas where incomplete advice could cause harm. They should check whether essential qualifications, eligibility criteria and geographical differences are retained. Their websites should also distinguish clearly between general information and advice about an individual’s circumstances, while providing simple routes to human support.
The priority should therefore be ensuring that people can access information in a form they understand, can assess whether it applies to their circumstances and can easily move from information to human support when required.
AI is changing the front door to online advice. The challenge for charities is not to prevent people from walking through it, but to understand where it leads.
Wherever someone begins their journey, they must still be able to establish whether the information they have received applies to them—and find the right information and human support when it does not.
