"ChatGPT says I'm right": how AI is changing the clients walking into law firms

For years, lawyers have dealt with clients who arrive at their first appointment having already searched their problem online.

Artificial intelligence is taking that behaviour considerably further.

Instead of reading a few websites before speaking to a solicitor, clients can now have a lengthy conversation with an AI chatbot about their dispute, ask follow-up questions, upload documents, request an assessment of their position and ask the technology to draft correspondence or suggest what they should do next.

By the time they approach a law firm, they may already have developed a detailed view of their legal position.

Sometimes that information may be useful. Sometimes it may be wrong. The bigger problem for lawyers is that it can be presented with enough confidence and detail to make the client believe there is very little left to debate.

The result is a new challenge for legal practice: lawyers are increasingly being asked not only to advise clients, but to review, verify and sometimes undo the legal advice their clients have already obtained from AI.

Consumers are already using AI for legal problems

This isn't a future problem.

Research published by the Legal Services Board in June 2026 found that consumers are already using AI-powered tools for legal issues including employment rights, housing disputes, debt and family matters.

The potential benefits are significant.

The Legal Services Board's research, based on a representative survey of 1,000 adults in England and Wales, found that 70% expected AI to make legal services easier to access, 66% expected improved accessibility and 64% believed it could make legal services more affordable.

For someone who cannot afford an initial consultation with a solicitor, the attraction is obvious.

They can ask a chatbot whether their employer can dismiss them, whether they have grounds to challenge a Will, what happens if their landlord refuses to undertake repairs or what their options are following a relationship breakdown.

It is available immediately and, in many cases, at little or no direct cost.

AI could therefore help people recognise that they have a legal problem and encourage them to seek professional advice when they might otherwise have done nothing.

But there is a significant difference between using AI to understand that a legal issue exists and relying upon it to determine exactly what the law says and what someone should do about it.

Lawyers are already dealing with the consequences

There is growing evidence that AI-generated client material is creating additional work for lawyers.

The Financial Times has reported that law firms are receiving increasing volumes of AI-generated emails, questions and legal documents from clients, with lawyers having to spend additional time reading, checking and responding to the material.

The report included experiences from lawyers at firms including Addleshaw Goddard and Mishcon de Reya and described clients using AI to suggest legal strategies and produce correspondence. One litigation partner reported receiving such a volume of AI-generated emails from a client that the firm decided it would respond only at appropriate intervals to points it considered material.

That creates an interesting contradiction.

AI is generally promoted as a technology capable of reducing the cost of legal services. In some circumstances, a client's use of it could actually increase their legal bill.

If a solicitor receives a concise email containing three questions, responding may be relatively straightforward. If those three questions have been transformed by AI into several pages of legal argument, statutory references and suggested authorities, the solicitor may have to work through the entire document to establish what is relevant and whether any of it is correct.

Where lawyers charge by the hour, that time has a cost.

For firms undertaking fixed-fee work, it creates a different problem. Lawyers may find themselves absorbing the additional time required to review large quantities of AI-generated material that was never anticipated when the fee was agreed.

The problem isn't just whether AI gets the law wrong

An AI response does not have to be completely wrong to cause difficulties.

Legal advice depends heavily on context.

A general explanation of employment law might be accurate while being completely unsuitable for a particular employee's circumstances. A chatbot might correctly identify a legal principle but overlook a limitation period, evidential weakness, procedural requirement or commercially sensible alternative.

There is also the question of jurisdiction.

A convincing answer based partly on US law or another jurisdiction is of little assistance to someone involved in proceedings in England and Wales.

The Law Society's guidance on generative AI recognises both the opportunities presented by the technology and the associated technology and data risks.

For lawyers, the difficulty is that clients may not necessarily recognise those limitations.

A professionally written response containing legal terminology can look authoritative.

Unlike a traditional internet search, where somebody is presented with several different sources and has to decide which to trust, an AI chatbot can provide one clear answer directly to the user.

That can turn tentative research into apparent certainty.

"ChatGPT says my claim is worth £50,000"

This could also have a significant impact on client expectations.

A solicitor may increasingly encounter clients who arrive believing AI has already established that they have a strong case, that a particular argument should be pursued or that their claim is worth a certain amount.

The lawyer then has two jobs.

They must establish the client's actual legal position and explain why the answer they previously received may not apply.

That can be a more difficult conversation than advising someone who arrives without a predetermined view.

A client who has spent weeks discussing their case with an AI tool may have repeatedly refined their argument and received responses that appear to support it. Being told by a solicitor that the position is considerably less straightforward may feel, to the client, as though the lawyer is being unnecessarily cautious.

The expertise of the solicitor can consequently be placed in competition with an answer produced instantaneously by technology.

The phenomenon has been compared with the effect experienced by doctors when patients arrive having researched and effectively diagnosed themselves online. Reporting in 2026 suggests lawyers are now encountering a similar problem, with clients arriving with AI-generated legal research, strategies and expectations.

What if the client has already acted on the advice?

Correcting a misunderstanding during an initial consultation is one thing.

The situation becomes considerably more serious when the client has already relied on AI-generated advice.

Someone representing themselves could potentially use AI to draft correspondence, respond to the opposing party, prepare a witness statement, formulate arguments or create documents for court proceedings.

By the time a solicitor becomes involved, decisions may already have been made.

Deadlines could have passed. Admissions may have been made. Poor arguments may have been advanced. Relevant evidence might not have been preserved. Documents could already have been submitted.

The lawyer is then no longer simply advising on the original legal problem. They are also dealing with what has happened since.

The risks of relying uncritically on AI-generated legal research have already reached the English courts.

In Ayinde v London Borough of Haringey and Al-Haroun v Qatar National Bank, the Divisional Court considered cases involving false or suspected AI-generated legal authorities. The judgment discussed the problem of AI hallucinations, where systems generate false or misleading material that can appear genuine.

The case primarily concerned the responsibilities of legal professionals, but the underlying problem applies equally to consumers. AI-generated legal information needs to be verified before somebody relies upon it.

Should lawyers ask clients whether they have used AI?

This raises an increasingly practical question for law firms.

Should AI use become part of client onboarding?

Depending on the nature of the matter, firms may eventually want to know whether a client has used an AI system to research their case, draft documents or prepare material that has already been sent to another party or filed with a court.

That doesn't need to involve interrogating every new client about whether they have ever used ChatGPT.

But where a client provides extensive pre-prepared legal arguments or correspondence, asking how those documents were produced could become important.

It may also be useful to establish whether the client has uploaded documents relating to the matter to an external AI service.

What has the client told the chatbot?

Confidentiality creates another problem.

A client trying to obtain the most accurate response from an AI system may provide significant amounts of information.

They might upload an employment contract, correspondence with an opponent, medical records, witness statements, commercially sensitive documents or information relating to a family dispute.

That could happen before they ever speak to a solicitor.

The Law Society warns lawyers to consider the data risks associated with generative AI, including how information entered into AI systems is processed and handled.

There is also a wider distinction consumers may not appreciate. Speaking to an AI chatbot is not the same thing as speaking confidentially to their lawyer.

That issue has already attracted significant attention in the US. In April 2026, Reuters reported on a federal ruling concerning AI conversations sought by prosecutors, prompting lawyers to warn clients against treating chatbots as though they were confidential legal advisers.

The precise privilege rules differ between jurisdictions, but the broader lesson is relevant to UK consumers too. People should not automatically assume that information shared with a consumer AI service attracts the protections associated with communications with their solicitor.

AI could still improve access to justice

None of this means lawyers should discourage consumers from using AI altogether.

That would overlook one of its potentially most important benefits.

There remains a substantial group of people who need legal help but struggle to afford traditional legal services. The Legal Services Board's research reflects the potential consumers see in AI making legal help more accessible and affordable.

Used appropriately, AI could help somebody understand terminology, organise information, identify questions to ask a solicitor or recognise when professional advice is necessary.

It could even make the first meeting with a lawyer more productive.

A client who understands the basic terminology surrounding their problem and has organised a chronology of events may be easier to advise than someone beginning with no understanding at all.

The distinction is between using AI as a tool for legal information and treating it as a substitute for individual legal advice.

For law firms, acknowledging that distinction may be more productive than simply telling clients not to use it.

Law firms may need to adapt to the AI-informed client

There is a temptation to view clients using AI as an irritation that will eventually disappear as people learn about the technology's limitations.

That seems unlikely.

AI systems are becoming easier to access and increasingly embedded in the technology consumers already use.

Lawyers should therefore probably expect more clients to arrive having already consulted one.

The challenge for law firms is deciding how to manage that behaviour.

Client care information could explain that AI-generated legal information should not be relied upon without verification. Lawyers could ask about AI-generated documents where relevant. Fixed-fee terms may need to address unusually extensive client-generated material. Fee earners may also need training in explaining the limitations of AI without making clients feel foolish for having used it.

Most importantly, firms may need to accept that the starting point of the solicitor-client relationship is changing.

Clients once walked into law firms primarily looking for answers.

Increasingly, some will walk in believing they already have them.

The lawyer's role may therefore involve something new: not simply explaining what the law says, but establishing whether the client's AI adviser got there first and whether it got it right.