15% of Large Firm Lawyers ‘Now Dependent on AI’

“15% of Lawyers Are AI-Dependent” Is a Vendor-Commissioned Number — Read It That Way

A LexisNexis survey of more than 500 lawyers in the UK, reported by Artificial Lawyer on 2nd September 2026, found that 15% of lawyers at large firms now describe themselves as "dependent on AI" to do their job, against 13% of in-house lawyers. Separately, 34% of respondents said they use AI tools every day and another 32% use them multiple times a week — putting roughly two-thirds of the sample in frequent, non-experimental use. The top three reported use cases were drafting, legal research and document review.

How reliant on AI are large-firm lawyers today?

Frequent use is now mainstream: two-thirds of surveyed lawyers use AI daily or multiple times weekly, and 15% call themselves dependent on it.

That gap matters. "Uses AI often" and "dependent on AI" are different claims, and the survey collapses them into one headline. Frequency is observable; dependency is self-reported and undefined — nobody knows if a lawyer who says they're "dependent" means they'd miss a deadline without it, or just that the work would be slower without it.

Who defines "dependent," and should buyers trust it?

LexisNexis both commissioned the survey and sells the tools respondents are describing, so the metric was never independently audited or defined.

There's no methodology detail on what counted as "dependent" beyond a self-selected label, no breakdown of which tools respondents meant, and no error-rate data tied to that dependency. A vendor publishing a survey that shows its own market growing and locking in is not evidence of quality — it's evidence of a sales narrative. Buyers should ask their own AI vendor for a defined, measured usage metric, not a self-reported one.

What does the document-review finding mean for evidence workflows?

Document review sits in the top three reported AI use cases, meaning review tooling is now core infrastructure, not a pilot project, at large firms.

Artificial Lawyer also notes that most surveyed lawyers remain concerned about hallucination risk and want tools that show verified sources — the right instinct but the wrong stopping point. If review is becoming something firms can't do without, the question isn't whether the AI is used often, it's whether every output it produces during review can be traced back to a specific document and exhibit when a partner or opposing counsel asks "how do you know that." Frequency of use says nothing about that traceability.

What should a firm ask before this dependency deepens?

Ask the vendor to show, per document, a cited link from every AI answer back to the source exhibit — not just an accuracy percentage.

Also worth asking: where does the model run, does privileged material ever leave the firm's environment or train a shared model, and what happens on a low-confidence flag — does the tool fail open and guess, or fail closed and say so? None of that is answered by a usage survey, and it's the part that actually determines whether "dependent" is a safe word to use about a litigation record.

Frequently asked questions

Does the survey say AI dependency causes errors?

No. It reports concern about hallucination risk alongside high usage, but it does not measure error rates or link dependency to accuracy.

Is 15% dependency high compared to in-house teams?

It's higher than the 13% reported for in-house lawyers, but both figures come from the same self-reported survey question, not an independent audit.

Does frequent AI use mean firms already have verified sourcing in place?

Not necessarily. The survey shows lawyers want verified sources, which implies many tools already in daily use may not reliably provide them.

Source: Artificial Lawyer.