Ruli In-house Legal AI entered the 2026 Legal Benchmarks cycle to have its work judged by practicing lawyers, and it came out as Industry Leader in three categories: Best Overall Legal AI, Best Data Extraction, and Best Product Experience. On product experience, Ruli scored a perfect 100.0 against a 72.2 benchmark average.

Our team has been waiting to share this one. That 100.0 belongs as much to the in-house lawyers who use Ruli every day as it does to us, because every redline they pushed back on, every playbook rule they refined, and every question that stumped an earlier version shaped the product Legal Benchmarks tested.

Lawyers wrote the test, and performance decided the result

Most claims about legal AI come from vendors, which is why an independent result carries real weight. Legal Benchmarks, co-founded by Anna Guo, is a lawyer-led group that pairs practitioners with researchers to test legal AI on real work. It awards its certifications on performance alone. More than 500 practicing legal professionals have designed, run, and reviewed its benchmarks to date, and its advisory network includes in-house legal leaders at Zoom, Morgan Stanley, Henkel, Dentsu, and Santander.

As a result, the tasks closely match the work an in-house team buys legal AI to do, so a win here reflects the contracts and questions that land on a legal team’s desk every week.

A perfect 100.0 for work that’s ready to use

Legal AI pilots usually stall for one reason: fixing the output takes longer than doing the work.

Legal Benchmarks scores product experience on ease of use and on reliably delivering legal-quality work that’s ready to use, and Ruli’s 27.8-point lead over the average points to output that needs less cleanup before it reaches the business.

Adoption decides the return on legal AI, and lawyers keep using the tool whose work they can send.

Best overall Legal AI, with an 8.3-point lead

A tool that handles one task well but stumbles on the next still leaves the team checking everything, so consistency is what lets lawyers trust Ruli on the next task.

Ruli scored 72.0 across the full benchmark against a 63.7 legal AI average, and Legal Benchmarks named it Best Overall Legal AI for results across every task.

The pass rate tells the more useful story. Ruli fully passed 31.1% of tasks against a 24.7% legal AI average, which means even the top-ranked tool clears fewer than a third of tasks outright.

Lawyer review remains an essential part of every legal AI workflow, and any vendor claiming otherwise is worth testing hard.

Best Data Extraction, with 41.7% of tasks fully correct

A simple question about a renewal date or a liability cap can mean digging through the original agreement in one folder, an amendment in another, and a side letter in someone’s inbox.

Ruli turns every contract and amendment into one source of truth. Any contract question gets a plain-English answer, and redlines start from what the company already signed.

On grading criteria, Ruli scored 63.1% against a 59.3% average, which is close. The real gap shows up in fully correct answers: Ruli got 41.7% of extraction tasks fully right, against a 31.3% average.

That’s the number to check when comparing legal AI vendors, because a partly right answer still sends a lawyer back to the source.

Your company’s history is the asset. The model is just the tool.

Extraction is where that difference shows up, because a system that reads every contract and amendment correctly is the one that can answer from what the company has actually signed.

What verified users report

Verified users rate Ruli 4.9 out of 5 on Ruli’s Legal Benchmarks profile, ahead of Harvey at 4.5 and Legora at 4.0 as of Q3 2026.

The reviews put the result in cost terms. A general counsel wrote, “Without Ruli, undoubtedly I would need to add another person to the legal team.” Another reviewer, a lawyer on a team of six to 20, credits Ruli with lowering outside legal spend and faster review on high-volume agreements like NDAs.

Reviewers also ask for faster results and clearer separation between sourced facts and inference, which is fair feedback and the kind that only reaches a vendor through an independent channel.

Thank you, and the next test

Thank you to Anna Guo and the Legal Benchmarks community for testing these tools the hard way, and to every in-house team that held Ruli to the standard they hold themselves to.

Scores only go so far, and a week of a team’s own inbound paper will show whether these results hold. We’d ask every legal team evaluating legal AI to run that test on any vendor it considers, including us. See how in-house teams use Ruli to keep more work in-house, from contract review to regulatory monitoring.