Westlaw AI vs Lexis+ AI vs Aewita: the research comparison.

Westlaw AI vs Lexis+ AI is the question every attorney evaluating research tools has asked in the last two years. The honest answer has always been that the two giants are more alike than either would admit. Both sit on top of the best case-law databases in the country. Both have layered AI features onto research platforms your firm has probably licensed since before the iPhone. Both depend, under the hood, on frontier models built and run by third-party AI providers.

Aewita is a newer entrant with a different architecture. We built the model. We host the model. The retrieval, the inference, and the citation verifier all run inside our boundary. This piece is a fair, sourced comparison of all three — written by the CEO of one of them, with the ground rules spelled out: we do not invent competitor internals, we do not reveal our own, and we cite every specific claim we can.

What Westlaw AI, Lexis+ AI, and Aewita have in common

Let us start with where the three are genuinely comparable, because it is more than the marketing would suggest.

Comprehensive U.S. case-law access. All three cover the full U.S. federal and state case-law corpus, plus federal statutes and statutes from all 50 states and D.C. If you are running a standard research query — a controlling case in a circuit, a statutory interpretation question, a current appellate opinion — all three will get you to primary source.

AI-powered search and drafting. All three have moved past keyword-and-boolean into natural-language queries, summarization, and assisted drafting. The UI metaphors differ; the category of capability does not.

Citation support and Shepardizing-style verification. All three help an attorney confirm that a cited case is still good law, and all three produce output that is intended to be verifiable against primary source. The quality of that verification is where we will spend most of this piece.

Coverage parity is the starting point, not the differentiator. The differentiator is what happens between the question and the answer.

Where they differ: the subprocessor question

Here is the architectural fork. When an attorney types a prompt, three things happen: the system retrieves relevant sources, an AI model generates an answer grounded in those sources, and a verification step checks the output. The question is who runs each step, and specifically who runs the inference.

Lexis+ AI (now marketed as Lexis+ with Protégé). LexisNexis’s own product page describes the offering as including “access to large language models from OpenAI, Google, and Anthropic,” positioning external frontier models as part of the user’s choice set. That is a publicly stated reliance on third-party AI providers in the inference path, alongside LexisNexis’s own grounding layer.

Westlaw AI (CoCounsel / AI-Assisted Research). Thomson Reuters has publicly described its generative AI capabilities as built in partnership with external frontier-model providers, with retrieval and grounding layers added on top. We will not invent a specific vendor relationship that has not been disclosed — but the public posture is that third-party frontier models sit in the inference path, alongside Thomson Reuters’s own infrastructure.

Aewita. We built and host our own frontier reasoning model. There is no external AI provider in the data path. The prompt does not leave the boundary to be processed by a third party’s model. The retrieval, the generation, and the citation verification all terminate inside Aewita. This is architectural privilege by design.

None of these architectural choices is inherently wrong. A firm with a client roster that permits third-party LLM subprocessors under outside-counsel guidelines has a full set of options. A firm whose largest clients have started pushing back on named AI subprocessors has a narrower set.

Pricing: seat licenses vs. a published list price

Westlaw and LexisNexis both use variations of traditional legal-research seat licensing, with pricing that varies by firm size, included practice areas, feature modules, and negotiated term. Neither publishes a single list price for its AI offering in the way a SaaS vendor does — the quote depends on the firm. This is not a criticism; it is how the two incumbents have sold for decades. It does mean that a small firm comparing line items is often quoted a different price than a mid-size firm, and a mid-size firm a different price than a BigLaw buyer.

Aewita is $99 per month or $720 per year. No seat minimum. No annual commit required. 14-day free trial. Cancel anytime. That is the whole quote. A two-attorney firm pays the same per-seat price as a forty-attorney firm. The procurement conversation is a credit card.

For enterprise firms already committed to a Westlaw or Lexis contract for other reasons — treatises, practice-area modules, or decades of institutional workflow built into the product — Aewita is not trying to replace the relationship overnight. For solo and small-firm attorneys who have been priced out of enterprise AI entirely, the difference is not a percentage; it is access.

Accuracy: a published number versus a public silence

This is the part of the comparison most vendors would rather not dwell on. Aewita publishes an audited hallucination figure with methodology, sample size, and confidence framing. Our core statement is specific:

In internal testing, Aewita observed zero hallucinated outputs across 800 consecutive queries — statistically, a rough upper bound under 0.3% at 95% confidence.

The full methodology and statistical framing are published, including how we define a hallucination (fabricated citations, misattributed propositions, and materially misstated holdings all count) and how we arrived at the rule-of-three upper bound.

Neither Westlaw AI nor Lexis+ AI has published a comparable audited rate with sample size and confidence interval. Both companies describe their accuracy work in qualitative terms on product pages and in white papers. That is not the same thing as a measured, disclosed number a firm can compare.

We take that fact seriously. If either vendor publishes an audited rate with sample size and methodology in the future, we will read it, and if the number is better than ours we will say so publicly. Until then, the only number in the category comes from the newcomer.

Coverage: genuine parity, not a differentiator

This is the section of the comparison I think competitors get wrong in their own posts, so let me be direct. Westlaw, LexisNexis, and Aewita all cover the full U.S. case-law corpus — federal and all 50 states, back through the history of American jurisprudence. Aewita specifically covers every U.S. court opinion from 1700 to today, 792 document types, and 22 practice areas. Westlaw and Lexis have their own proprietary headnote systems, editorial enhancements, and treatise libraries that Aewita does not replicate. We are not trying to.

For most attorneys most of the time, coverage parity means you can stop worrying about which platform has the case. All three have the case. The question is what happens to your prompt in the two minutes between the question and the answer.

What Aewita does not replace

I want to be candid about the limits, because a buyer’s guide that pretends the newcomer wins on every axis is not useful to anyone.

If your firm has a forty-year Westlaw relationship, a library of institutional treatises you rely on daily, a knowledge-management practice built around Westlaw’s editorial features, or a training pipeline for new associates that assumes Westlaw fluency, Aewita does not unseat all of that overnight. Many firms run Aewita alongside Westlaw or Lexis as a pilot — using Aewita for AI-assisted research and drafting, and keeping their incumbent for the features built up over decades of institutional use. That is the right starting posture. A pilot that shows value earns a conversation about consolidation; a pilot that does not, fails cheaply.

The same is true for firms with deep Lexis+ investment in analytics, Shepard’s, or practice-specific treatise collections. The newcomer’s job is to earn its place, not to demand a rip-and-replace.

Who should pick what: scenario-based recommendations

Large firm with an existing enterprise contract and clients that allow third-party LLM subprocessors. Westlaw AI or Lexis+ AI remain defensible, well-integrated choices. If you are already locked into the platform for other reasons, the AI layer is additive. Pilot Aewita alongside on a handful of seats to benchmark accuracy and see whether the published hallucination rate holds up on your work.

Firm with regulated-enterprise clients pushing back on named AI subprocessors. This is the scenario Aewita was built for. If your outside-counsel guidelines make it difficult to add a specific external AI provider as a subprocessor, an architecture with no third-party LLM in the data path turns a multi-month security conversation into a one-page answer.

Solo attorney or small firm priced out of enterprise AI. Aewita’s $99 per month, no-minimum model is explicitly the one built for you. Pair it with whatever primary-source subscription you already trust, run your hardest queries, and evaluate accuracy on actual work.

Firm that values a published accuracy number over marketing claims. Aewita is the only vendor in the category that has published a measured hallucination rate with sample size and confidence interval. If that is a hard requirement for your procurement team, the list of candidates is currently one.

Those are the scenarios. None of them requires the vendor to lose for another vendor to win.

Three good tools. One of them doesn’t send your work to someone else’s AI.

The bottom line

Westlaw AI, Lexis+ AI, and Aewita are all comprehensive U.S. legal research platforms. Coverage is a tie. Where they separate is architecture, pricing transparency, and published accuracy — and on all three of those axes, the newer vendor has spent the last two years turning what would normally be marketing claims into things you can ask a vendor directly and expect an answer to.

If you want to see whether the published number holds up on your own work, the 14-day trial is a real trial with real product access. Bring a research question from a live matter, bring your hardest jurisdiction, and grade the output against what you already know is correct. If it clears the bar, you have a simpler stack. If it does not, you walk away with better evaluation data for your next conversation with Thomson Reuters or LexisNexis.

That is the deal I am comfortable offering. Three good tools. Run the comparison on your work.