LexisNexis: Why lawyers still don’t trust AI accuracy

Lawyers are using AI more than ever, and for good reason.

Our latest research found that 34% of legal professionals use AI every day and another 32% use it several times a week. Legal research is now the most common use case, used by 69% of respondents, followed by document summarisation at 62%. More than half are already using AI for drafting and document review.

That level of adoption shows how quickly AI has become useful in everyday legal work.

At the same time, lawyers remain cautious about accuracy. 83% are concerned about relying on inaccurate or fabricated information.

 

It remains the biggest AI concern in our research. Confidential-data leakage comes next at 53%, followed by becoming too reliant on AI at 52%.

Rather than suggesting lawyers are reluctant to use AI, the findings point to something more interesting: they are becoming more experienced users. They recognise the value of AI, but they also have a clearer sense of where human judgement still needs to come in.

Useful answers still need checking

Anyone who has spent time with generative AI will recognise the appeal. It can produce a useful first answer quickly, help lawyers get through large volumes of material and make everyday tasks more efficient.

In legal work, though, the quality of the answer still depends on context. A research response may need another authority. A summary may need to be checked against the source material. A draft may need to be adjusted for the facts or the client.

That changes the role of the lawyer rather than removing it. AI can accelerate the first stages of the work, while the lawyer remains responsible for testing, refining and applying the result.

One in-house legal leader described the trade-off in our survey:

“Quicker to draft emails, summarise documents, obtain an initial steer. Basically it enables you to work quicker but sometimes reviewing AI output can take as long as if you reviewed it yourself but that human review cannot be lost.”

That is a useful way to think about AI productivity. The benefit is not simply producing an answer faster. It is helping lawyers get to a useful starting point more quickly, while keeping professional judgement in the loop.

Lawyers want to see what sits behind the answer

The survey gives a fairly clear indication of what increases confidence.

81% of respondents say they feel more comfortable using AI when it is grounded in legal sources. Among the 541 people who answered the question, 35% strongly agreed and 46% agreed.

That preference fits naturally with the way lawyers work. Legal professionals are used to checking authority, understanding context and tracing conclusions back to reliable sources.

A Senior Associate put it this way:

“Yes, the general tools are available to the clients at the moment – but the tools with the right harness are the ones that they can rely on and don’t yet have access to.”

As general-purpose AI becomes easier for everyone to access, that ability to connect AI with trusted legal information could become increasingly important.

Producing an answer is becoming easier. Giving lawyers confidence in how that answer was reached is where much of the value sits.

Trust is becoming part of the buying decision

The same pattern appears when lawyers choose technology providers.

59% select strong data privacy and security standards as one of their most important considerations. Transparency about how a provider uses AI follows at 40%, with human oversight at 39%. By comparison, 28% select cost and commercial value.

This suggests buyers are looking beyond raw capability.

They want to understand the source material behind the system, how information is handled, how easy outputs are to verify and where human review fits into the workflow.

As the in-house legal leader said elsewhere in the survey:

“Yes, it will be an advantage as the trust is already there but of course trust can be easily lost hence governance and a human in the loop is essential for use of AI.”

Trust, in that sense, is not a sign of resistance to AI. It is part of what allows lawyers to use it more confidently.

Clients will have AI answers too

AI is also changing the lawyer-client relationship.

Clients increasingly have access to the same general-purpose tools and may arrive with more information, more questions and sometimes their own interpretation of the law.

Our research found that 77% of legal professionals believe AI is causing clients or the general public to misinterpret the law.

A barrister in our survey offered an interesting prediction:

“Increasingly I suspect that the only reason people will come to lawyers is that they need someone they trust to verify the AI advice they have already sourced.”

That could also create a new opportunity for lawyers.

Clients may arrive better informed than before, but still need help understanding whether the information they have found is accurate, relevant and applicable to their circumstances.

AI is becoming very good at helping people find and generate answers.

The role of the lawyer is increasingly to add the context, judgement and verification that turns those answers into reliable legal advice.