Aewita vs. Legora: two different regions, two different bets

Every few weeks a U.S. attorney asks us the same question. "How does Aewita compare to Legora?" It is a fair question. Both products sell legal AI. Both sit in the same category on analyst charts. Both get compared in the same RFPs.

But the more time you spend with each, the more you notice the seams. Legora was built in Stockholm. It grew up on Nordic and EU case law. Aewita was built for U.S. practice, trained on every U.S. case from 1665 to today. The two products are not really competing for the same desk.

This piece is a straight comparison. No hit piece. If you need EU coverage, the honest answer is Legora. If you need U.S. coverage with transparent pricing and a self-hosted model, the honest answer is Aewita. The rest explains why.

The geographic fault line

Legora, formerly known as Leya, is a Swedish legal AI company. It raised roughly $35 million across its seed and Series A, grew fast in the Nordics, and expanded into the broader EU. Its marketing, case studies, and early adopters skew heavily toward European firms. That is where the product is strongest.

Aewita was built in the United States, for U.S. attorneys. The training corpus reflects that. Every federal case, every state supreme court opinion, every intermediate appellate decision, every statutory title across all 50 states and D.C. Federal regulations. Federal and state rules of civil and criminal procedure. 792 document types across 22 practice areas.

The two products overlap in the middle: a brief template, a clause comparison, a deposition outline. Ask either tool to redline an NDA and you will get a competent answer. But ask for the controlling authority on a narrow question of Texas contract law, or the split among circuits on a federal pleading standard, and the gap widens fast.

Training data tells the story

Legal AI is downstream of what the model has read. A tool trained on Swedish labor law will reason well about Swedish labor law. A tool trained on U.S. common law will reason well about U.S. common law. There is no free lunch.

Aewita's corpus covers the U.S. from the colonial era forward. That matters more than it sounds. A lot of property, admiralty, and common-law reasoning still points to 18th and 19th century cases. When an attorney cites Pennoyer v. Neff or Swift v. Tyson, the model should know the case, the holding, the later treatment, and whether the authority still stands. Corpus gaps show up as hallucinated citations.

Legora's public material emphasizes EU and U.K. coverage. The product is strong there. It is less strong on U.S. case law, U.S. statutory interpretation, and U.S. procedural rules. That is not a knock on the team. It is what they built.

A legal AI tool is only as good as the law it has actually read. Everything else is packaging.

Architecture: wrapper vs. self-hosted

Most legal AI products in this market run on third-party LLMs. They call out to OpenAI, Anthropic, or Google, add a retrieval layer, and ship the result. Legora, from what its public documentation describes, follows that pattern. The model that generates the answer is operated by an upstream vendor.

Aewita is different. The frontier reasoning model is self-hosted on infrastructure we operate. Zero API calls to OpenAI, Anthropic, or Google. Zero client text routed to third-party model providers. We hold patents on the retrieval system, the inference pipeline, and the citation verifier.

Why this matters for a U.S. firm:

• ABA Model Rule 1.6. Confidentiality is not a policy commitment. It is the default when the data never leaves.
• Model Rule 5.3. Supervision of non-lawyer assistance. Easier when the assistant is one system, not a chain of subprocessors.
• Model Rule 1.1. Competence. Easier to diagnose and correct errors on a model you control.

You can read the full architectural posture on the security page.

Pricing: published vs. negotiated

Legora is enterprise-focused. Pricing is not public. Deals are negotiated. Procurement cycles run long. That model works for large firms with a dedicated legal ops function and the budget to absorb six-figure annual contracts.

Aewita publishes its price. $99 per month, or $720 per year (a 39% discount for annual). 14-day free trial. No seat minimum. Cancel anytime. A solo practitioner can swipe a credit card and start using the product the same afternoon. A two-partner firm can outfit both desks for under $200 a month.

This is a deliberate choice, not a growth hack. The U.S. legal market is 1.3 million attorneys, and the long tail (solos, small firms, in-house counsel at mid-market companies) is badly underserved by enterprise-priced tools. Transparent pricing is the point.

Coverage at a glance

Dimension	Aewita	Legora
Primary market	United States	EU, Nordics, U.K.
Case law depth	Every U.S. case, 1665–today	Strong EU / Nordic coverage
Statutory coverage	All 50 states + D.C. + federal	EU directives, national codes
Model architecture	Self-hosted, zero third-party calls	Third-party LLM underneath
Pricing	$99/mo or $720/yr, published	Enterprise, custom
Trial	14 days, self-serve	Sales-led pilot
Compliance anchor	ABA Model Rules 1.1 / 1.6 / 5.3	EU data protection (GDPR)

Hallucination posture

We publish a measured hallucination rate: under 0.3% at 95% confidence interval, measured across 800 queries. That is a specific number tied to a specific methodology. It is not a marketing line.

Most competitors, Legora included, do not publish a comparable figure. That is not an accusation. Benchmarking legal AI honestly is hard, and most vendors have decided the risk of publishing a number outweighs the benefit. We disagree. If a tool is going to write briefs that get filed, the error rate is the most important number on the page.

A citation verifier sits between the model and the page the attorney reads. Every cited case, statute, and regulation gets checked against the source corpus before the answer is returned. If a citation cannot be verified, the system says so rather than inventing one. That design choice is what pulls the measured rate below 0.3%. It is not a prompt trick.

Procurement and onboarding

The buying experience differs as much as the product. Legora, like most enterprise software, runs through procurement: security questionnaire, data processing agreement, pilot scoping, commercial negotiation. That cycle can run three to six months from first meeting to first seat provisioned. For a global firm, the structure makes sense. For a two-partner shop in Austin, it is a non-starter.

Aewita's path is different. A credit card, an email, and a free trial. If the firm wants an MSA and a DPA for compliance, we sign. If the firm wants to start this afternoon, that works too. Most of our users go from signup to first real research query in under ten minutes. That speed is a deliberate choice: it removes the cost of trying.

Who Aewita is for

Aewita fits three profiles well:

• U.S. solo and small-firm attorneys who want a frontier research and drafting tool without enterprise pricing.
• U.S. mid-market firms who want to consolidate research, drafting, and citation verification on one platform without adding three subprocessors.
• In-house counsel at U.S. companies who need ABA-aligned controls and a clear audit trail.

If any of those describe you, start with a 14-day free trial. You will get the full product, not a watered-down demo version.

Who Legora is for

Legora fits a different profile, and it fits it well:

• EU and Nordic firms with strong regional case law needs.
• Multinationals that want a European-grounded legal AI layer for their European offices.
• Enterprise buyers with procurement cycles, dedicated legal ops, and budget for negotiated contracts.

If that is your firm, Legora is a credible choice. The team is serious. The investors are serious. The regional focus is a feature, not a limitation.

The honest summary

Aewita and Legora are not really substitutes. They are two companies that drew different maps.

Legora drew Europe. Aewita drew the United States. Within each region, each product is strong. Across regions, neither one is a direct replacement for the other. If you run a Stockholm firm with a Frankfurt office, Legora. If you run a Dallas firm with a San Diego office, Aewita.

The rest of the decision is about architecture, pricing, and trust. On those, we have published our posture openly: self-hosted model, zero third-party inference, $99/mo published price, under 0.3% measured hallucination rate, ABA Rules compliance by design. You can compare our posture with any competitor on the compare page, read the pricing details on the pricing page, or dig into the research product at /product/research.

If you are a U.S. attorney weighing the two, the honest answer is: try Aewita for 14 days. If it does not fit, you will know fast. If it does, you will stop asking which tool to pick.