Clients using AI chatbots for legal help now number about one in six of everyone with a recent legal problem, on research published on 1 September 2026 by the legal charity JUSTICE and the Administrative Fairness Lab. The study surveyed 3,287 people across the United Kingdom through Prolific between December 2025 and January 2026, with quotas set to match the population by age, gender and ethnicity. Almost half of them, 1,428, had a legal dispute in the previous two years, and 233 of those, 16.3 per cent, had put it to a chatbot.
What separates this from the usual survey is that 77 of the 233 handed over the conversations themselves. The result is a record of what people ask when a legal problem lands on them, in their own words, at the point before most of them reach a solicitor.
What they were asking
The problems were ordinary ones. Consumer disputes over faulty goods and refunds, tenancy deposits and evictions, dismissal and discrimination, family matters, debt collection, medical negligence and insurance, most of it outside the scope of legal aid. Four uses run through the transcripts. People asked about their rights and about what happens in court. They asked for translation, one of them pasting in an email from their own solicitor with a request to explain it in normal person language. They asked for drafting, both to sharpen a letter they had written and to produce a complaint or a refund request from nothing, and one person contacted by a bailiff asked for a script to read down the phone. They also asked for reassurance, describing themselves as frightened of the process and, in one case, lost and powerless.
Your client is checking your work
Some of the participants already had a solicitor and used the chatbot to test the service they were getting. One was unimpressed to learn that their conveyancing solicitor had verified a building's height using Google Maps. Another asked whether their solicitor's conduct sat within normal practice or grounded a complaint. Your client is taking a second opinion on your work from a tool that holds none of the file and owes them no duty.
That second opinion carries a bias. JUSTICE points to Oxford research from April 2026 finding that chatbots trained to be warm and agreeable make more mistakes and tell people what they want to hear, with the most distressed-sounding users drawing the least accurate answers. A worried client asking whether their solicitor has let them down is the profile most likely to be agreed with rather than corrected.
What to do about it
Write your client care letter and your matter updates in language that survives being pasted into a chatbot, because some of it will be. Where a client brings you a document they have drafted, read it as theirs and test the assertions in it rather than the prose. Answer the emotional question alongside the legal one, since saying what happens next and when is the part the machine does willingly and a busy fee earner skips. When a client tells you the AI said something different, ask what they typed before you correct the answer, because the prompt usually explains the gap.
One caution before you read the study across to your own client base. The people using chatbots were younger, more likely to be male and more likely to be employed, and JUSTICE warns that the digital gap may deepen the justice gap rather than close it. Chatbot users were also less likely to have consulted a solicitor at all, 33.9 per cent of them against 44.6 per cent of everyone else. The client who arrives with a chatbot in hand is not the client you are at risk of losing to one.
The research is published as What AI Chatbots Can Teach Us About Unmet Legal Needs on the JUSTICE website, which is open to read and asks nothing of you.
If nobody has read your client care letter and your update templates since clients started running them past a chatbot, we go through them with you: start with a conversation.