Best Legal AI for MENA and Arabic Law Firms (2026)

Most "best legal AI" lists were written for BigLaw in New York and London. They rank tools on tasks that assume English-language common law, they price for firms with a procurement department, and the words Arabic, MENA, UAE and Saudi Arabia do not appear in them at all. This list is the other way round: it starts from a practice in Riyadh, Dubai, Cairo or Beirut, names the regional tools alongside the global ones, publishes what each vendor actually says on its own site with the date we checked, and says plainly where HAQQ loses.

Key facts

• None of the global shortlist publishes Arabic support. On 22 July 2026 we checked the public product pages of Harvey, Legora, Spellbook, Lexis+ AI and CoCounsel. None of the five mentioned Arabic, right-to-left support, or any MENA jurisdiction. That is a statement about what they publish, not proof of what their models can or cannot do.
• None of those five publishes a price either. All five route to a demo or a custom quote. Spellbook's pricing page says pricing "is determined by the number of team members on a license" and gives no figure (checked 22 July 2026).
• The regional field is real, and mostly single-country. Adel, Shwra, Qaanoon and Malakah are Saudi-focused; Qanooni, realLaw and Oqood are UAE-centred; LegalMind is Egyptian. We mapped the whole field in the Arabic AI lawyer app teardown.
• Practice-management vendors serve the region better than the AI-first vendors do. Lexzur ships an Arabic interface alongside English, French and Spanish and markets to MENA institutions on its home page (checked 22 July 2026). Any comparison that treats "no Arabic" as universal in this market is wrong.
• Gulf data protection is not one regime. A UAE onshore entity, a DIFC entity, an ADGM entity and a Saudi entity sit under different instruments. Work out which applies to you before you sign any AI vendor — see the residency section below.

How we ranked this, and who paid for it

HAQQ Research ran the benchmark cited below, and HAQQ finishes first in it. That is a conflict of interest, full stop. We are not going to pretend otherwise, because pretending is what the rest of the category does. Treat the scores as our internal testing, disclosed as internal testing — not as an independent, audited or peer-reviewed evaluation. The full result, including the categories where HAQQ loses, is published so you can argue with it.

The benchmark is 50 tasks graded out of 50, covering Sharia and statute awareness, clause completeness, forum selection, hallucination resistance, formatting, brevity, partner-readiness and source linking, spread across 11 legal work types. Everything else on this page — Arabic support, jurisdictions marketed, price — is not scored by us at all. It is copied from each vendor's own public pages, with the date we looked. Where a vendor says nothing, the cell reads "not published", because absence of a marketing claim is not evidence of absence of a capability.

Table 1 — built for, or genuinely usable in, MENA

Two things to take from that table. First, HAQQ is not the only Arabic legal AI and we are not going to claim it is — Adel has traction we respect, and Arabic.ai has a sovereign Arabic model we do not claim to beat on raw language modelling. Second, almost every regional tool stops at one border. If your commercial practice touches more than one Gulf state, a single-country tool covers part of your desk.

Table 2 — the global tools MENA firms shortlist, and what does not transfer

The honest framing matters here. We are not saying Harvey, Legora, Spellbook, Lexis+ AI or CoCounsel cannot process Arabic text. We are saying that on 22 July 2026 none of them published Arabic capability, right-to-left handling, or a MENA jurisdiction as something they support, and that their marketed databases and reference customers are US, UK and European. If Arabic-language filings are your deliverable, ask each vendor for that in writing before you buy — the marketing page will not answer it for you.

1. HAQQ — best all-round for MENA and Arabic firms

HAQQ, powered by its Justinian engine, tops our benchmark in all 11 categories: general legal reasoning (49/50), NDA drafting (49), employment agreements (48), contract drafting (47), legal research (47), shareholder agreements (47), commercial agreements (47), law explanation (46), license agreements (46), consultancy agreements (45), and memo drafting (44). More important for this list, it is the only tool in the global shortlist that publishes native Arabic with full right-to-left support and treats civil-law and Sharia-influenced systems as the default rather than an afterthought. It is not the only Arabic legal AI in existence — table 1 above names the regional ones — but it is the only one there covering more than one country's law alongside a practice-management layer.

It is also the only option in this list that covers both halves of the job. HAQQ runs the Legal AI side (Free, Starter, Pro and Business plans) for research, drafting and review, plus eFirm (Boutique, Purple and Enterprise) for matter workspaces, hearings, KYC screening and double-entry firm accounting. Chat is on mobile as well as web. Published prices, matching the pricing page exactly and in USD: Chat is free, then $30, $100 or $300 per month, with Enterprise on quote; eFirm is $25 per user per month on Boutique or $60 on Purple, both cheaper billed annually ($250 and $600 per user per year). On data location, HAQQ's security page publishes a choice of EU, US and Middle East regions — ask us to confirm in the contract which one applies to your firm rather than relying on this paragraph. Where does it lose? On pure contract drafting it scores 44 and sits third behind Spellbook. On legal research it scores 43, a top-three result led by Lexis+ AI for US common law, though HAQQ leads our Arabic and civil-law research tasks. And on plain-language law explanation it scores 42, behind general chat models. We would rather you know that than find out later.

2. Harvey — best for large firms doing English cross-border work

Harvey is the enterprise standard for global BigLaw, and it is genuinely strong on complex, high-value English-language work: it is a close second on shareholder agreements (44/50) and commercial agreements (43). If you run a large MENA firm handling cross-border deals under DIFC or ADGM common-law rules, in English, with a real software budget, Harvey belongs on your list.

The catch for most MENA firms is fit. Harvey's site publishes no Arabic or right-to-left capability and names no MENA jurisdiction (checked 22 July 2026), the whole product surface is presented in English, there is no published price and no self-serve tier — the call to action is "Request a Demo" — and there is no practice-management or billing layer. On general legal reasoning it scores 38 on our internal benchmark to HAQQ's 49. It is a specialist bought by large firms, not a starting point for a boutique practice.

3. Legora — collaborative review for mid-to-large teams

Legora, built in Europe, is good at collaborative and tabular document review across a team, and it drafts contracts competently: 42/50 on contract drafting, behind Spellbook and HAQQ, and 40 on employment agreements. Teams that want several lawyers working over the same set of documents at once will find it capable.

Like Harvey, Legora publishes no Arabic or right-to-left capability and names no MENA jurisdiction (checked 22 July 2026), and it publishes no price. Its European civil-law heritage is closer to MENA systems than a US-centric tool, which is a genuine point in its favour, but the language half of the problem is left to you to verify with their sales team. Read the longer Legora vs HAQQ comparison if a team review workflow is your priority.

4. Spellbook — best pure contract drafting, if you live in Word

Spellbook wins one category outright on our benchmark: contract drafting, at 46/50, the top score in the whole set for that task. It lives as a Microsoft Word add-in, so transactional lawyers who spend their day redlining in Word get suggestions exactly where they work, and there is a free trial. It does not publish a rate: its pricing page says pricing "is determined by the number of team members on a license" and gives no figure (checked 22 July 2026). An earlier version of this post carried a per-seat estimate; we removed it rather than restate a number the vendor does not publish.

It is deliberately narrow. Spellbook scores 18/50 on legal research on our benchmark, publishes no multi-jurisdiction research depth, has no practice-management layer, and mentions no Arabic or right-to-left support on its public pages (checked 22 July 2026). Outside the drafting-in-Word workflow it falls away fast. For a MENA firm that also researches, explains and manages matters, it solves one slice of the job and leaves the rest.

5. LexisNexis (Lexis+ AI) — best for common-law research and citations

LexisNexis leads legal research at 46/50, ahead of everything else, on the strength of its case-law database and citation reliability. If your work is common-law research where a verifiable citation is the whole deliverable, Lexis+ AI is the reference standard and has been for decades.

That strength is anchored in US and UK common law. The Lexis+ AI product page names no Arabic capability and no MENA content set, and states that pricing "varies based on factors such as the size of your organization, specific capabilities required, and the scope of content access" rather than publishing a figure (checked 22 July 2026). On our internal benchmark it scores 36 on general legal reasoning, and HAQQ leads our Arabic and civil-law research tasks with a top-three overall research score of 43. Lexis+ AI is a research tool, not an all-in-one practice platform.

The five at a glance

The Arabic test: a protocol you can run yourself

Every vendor in this category will tell you its model "supports over 100 languages". That claim is close to meaningless for legal work, because a model that can translate a menu is not a model that can hold a defined-terms section together across a right-to-left rewrite. So here is the test, published in full, with the prompts, so you can run it on any tool you are evaluating during its trial. Five tasks, each graded pass, partial or fail:

• 1. Arabic clause extraction. Upload a commercial contract drafted in Arabic. Prompt: "استخرج بنود إنهاء العقد والتعويض وحدد الجهة القضائية المختصة، مع الإشارة إلى رقم كل بند." Pass means the right clauses, correctly numbered, in Arabic. Fail means it silently answers about a different clause, or answers in English when asked in Arabic.
• 2. EN to AR translation of a defined-terms section. Give it a definitions block with ten defined terms and ask for an Arabic version that keeps every defined term used consistently and flags any term it could not render. Partial is the common result: the translation reads well and the defined terms drift.
• 3. Arabic statute citation. Ask a question of statute in Arabic that requires naming the instrument and article. Pass means it names the instrument and article and says when it is unsure. Fail means a fluent, confident, invented citation — the failure mode that matters most and the one a fluency-based review will not catch.
• 4. Right-to-left document formatting. Ask for a two-page Arabic memo with numbered headings, a table and an English defined term inline. Check what happens to the numbering, the table column order and the bidirectional run around the Latin text. This is where tools built for English quietly break, and it is invisible until a client sees it.
• 5. Mixed AR/EN contract review. Give it a contract with an Arabic body and an English annex and ask which language governs and where the two conflict. Pass means it notices the conflict clause and the conflict.

What we are not doing here is publishing a scored grid across every tool on this page. Harvey, Legora and Lexis+ AI are sales-gated: we could not access them to run these prompts, and scoring a tool we could not run — or inferring a fail from a silent marketing page — would be exactly the thing we criticise the rest of the category for. So the grid says "not tested — no access" for those three. If a vendor wants a score, the prompts are above and the offer is open.

Jurisdiction by jurisdiction

"MENA" is not a jurisdiction, and treating it as one is the mistake that makes most tool comparisons useless here. What follows is the legal system, the working language of filings, and where to check the regulator's position for each market. On whether your bar or regulator has said anything about AI use, do not take our word for it: our country-by-country tracker is the page we keep current, and the regulator's own site beats both.

United Arab Emirates — onshore vs DIFC vs ADGM

Three legal environments in one country, and a tool that fits one may not fit the others. Onshore UAE is a civil-law system with Arabic as the language of the courts, so Arabic drafting and RTL output are functional requirements, not preferences. The DIFC and ADGM are separate common-law jurisdictions with their own courts operating in English, which is precisely where an English-first tool like Harvey is a reasonable fit. Most UAE commercial practices touch both. Deeper detail is in the UAE guide.

Saudi Arabia

A Sharia-based system with a growing body of codified law, and Arabic throughout. It also has the deepest homegrown legal AI cluster in the region — Adel, Shwra, Qaanoon, Malakah and others — every one of them Saudi-only. If your matters stop at the Kingdom's border, a Saudi-specific tool with in-Kingdom corpora is a serious option and we say so. If they do not, you need multi-jurisdiction coverage as well. See the Saudi Arabia guide.

Egypt

A French-influenced civil-law system, Arabic-language filings, and a large, price-sensitive bar. This is the market where sales-gated enterprise pricing is not a negotiation problem but a non-starter, and where self-serve, USD-priced tools with a free tier are the only realistic entry point for most firms. LegalMind is the notable local player.

Qatar and Kuwait

Both civil-law, both Arabic for court work. Qatar adds a wrinkle familiar from the UAE: the Qatar Financial Centre operates on a separate, English-language footing from the onshore courts, so the same firm can need two different tools for two matters. Neither market has the vendor density Saudi Arabia or the UAE has — the regional landscape post maps who is actually building where.

Lebanon and Jordan

Lebanon runs a French-derived civil-law system where practice moves between Arabic and French, which puts a hard requirement on genuine multilingual handling rather than English-plus-translation. Jordan is civil-law and Arabic, and is where Arabic.ai is building sovereign Arabic language models with Qistas — an approach aimed at government and enterprise deployment rather than a firm buying seats.

Morocco and the Maghreb

Civil-law systems shaped by French law, with Arabic as the official language of proceedings and French heavily used in commercial practice. The bilingual reality here is the same as Lebanon's and it is the single most under-served requirement in the whole category: no global vendor markets an Arabic-and-French legal workflow, and the regional vendors are almost all Gulf-focused. The MENA landscape post covers where the gaps sit.

Data residency, PDPL, DIFC and ADGM

This is the section a general counsel checks first and no competitor list has at all. It is also the section where we are going to be deliberately unhelpful about specifics, because paraphrasing a data-protection obligation from memory is how firms get bad advice. Four instruments govern most of the region's legal AI procurement, and which one applies depends on where your entity is established, not where your client is:

• UAE onshore — Federal Decree-Law No. 45 of 2021 on the Protection of Personal Data. Start from the UAE government's data protection page and read the current consolidated text and any implementing regulations before relying on a summary.
• Saudi Arabia — the Personal Data Protection Law, administered by SDAIA, which publishes the law, its implementing regulations and its guidance. The regulations have been amended since the law was first issued, so check the version date on anything you read.
• DIFC — Data Protection Law No. 5 of 2020, a separate regime with its own Commissioner, published at difc.com. A DIFC entity is not covered by the onshore federal law.
• ADGM — its own Data Protection Regulations, published at adgm.com, again with a separate Commissioner. Same point: separate jurisdiction, separate rules.

What we will say is what to ask a vendor, because that part is jurisdiction-neutral. Where is the data physically stored, and can you name the region in the contract? Where is it processed, which is often a different answer? Are sub-processors listed, and does that list include a model provider? Is client data used for training, by the vendor or by any third-party model in the stack? Can you export everything and delete everything on request? What happens to matter data if the contract ends? Get those in the agreement, not the marketing page.

For HAQQ specifically: the security page publishes a choice of EU, US and Middle East data regions, and our security and certifications write-up sets out the rest of the posture. Ask us to put the region in your contract. On the global tools, only CoCounsel named region-based hosting at all when we checked, and the regions it named were the UK, Australia and Canada — not the Gulf.

What this actually costs a MENA firm

Every price below is either published by the vendor with the date we checked it, or marked not published. We do not estimate, and we do not carry over figures from software directories. HAQQ prices are in USD and match the pricing page; the Egypt exception is handled there rather than restated here.

That table is mostly empty, and the emptiness is the finding. Five of the seven best-known tools in this category will not tell a lawyer in Cairo or Amman what they cost without a sales call. That is not dishonest — enterprise sales motions genuinely price per deal — but it does mean the published-price tools are the only ones a solo or boutique practice can evaluate on its own time.

Where this ranking does not help you

• US or UK common-law research where the citation is the deliverable. Lexis+ AI and CoCounsel sit on case-law databases we do not have and are not trying to replicate. If that is the job, buy the research platform.
• Redlining all day inside Microsoft Word. Spellbook beat us on that exact task on our own benchmark. We published the loss rather than hide it.
• High-volume US eDiscovery and litigation analytics. Not a market we serve.
• Single-country Saudi depth. Adel's Saudi corpus and traction are real. If your practice never leaves the Kingdom, evaluate it seriously against us.
• Sovereign or air-gapped government deployment. Arabic.ai's on-prem, sovereign-LLM approach targets a procurement shape we do not currently sell into.

Which one should a MENA firm actually pick?

• Solo, boutique or mid-size firm working in Arabic, French or English, often from a phone — HAQQ. It publishes native Arabic and RTL, covers research, drafting and practice management, and starts free. See also the boutique and solo firm list.
• Practice never leaves Saudi Arabia, consumer or small-firm work — evaluate Adel and Shwra alongside HAQQ. Local corpus depth is a real advantage and we are not going to pretend otherwise.
• Large firm, real budget, English-language cross-border deals under DIFC or ADGM rules — Harvey is a credible fit for that specific matter type; HAQQ for the Arabic and onshore civil-law side.
• Your day is redlining contracts in Microsoft Word, in English — Spellbook for that one task, HAQQ for everything around it.
• Your deliverable is US or UK common-law research with airtight citations — Lexis+ AI or CoCounsel, with HAQQ for Arabic and civil-law research.
• A team reviewing the same English document set together — Legora.
• Government or sovereign deployment with an on-prem requirement — Arabic.ai with Qistas is the one built for that shape.

The pattern is consistent. For the general shape of a MENA legal practice — drafting, research, review and matter management across Arabic and civil law — HAQQ is the best single tool we know of in 2026, and we are the ones who make it, so weigh that accordingly. Where your work narrows to one English-language specialty, or to one country with a deep local corpus, the specialist that owns that lane can beat us there. An honest list says both.

Changelog

• 22 July 2026 — Re-checked Harvey, Legora, Spellbook, Lexis+ AI, CoCounsel and Lexzur against their own public pages and dated every claim. Removed an unsourced per-seat price estimate for Spellbook. Replaced "does not read Arabic" claims with what each vendor actually publishes. Corrected HAQQ pricing to match the pricing page ($30 / $100 / $300 for Chat, $25 / $60 per user for eFirm). Relabelled the benchmark as HAQQ's internal testing and moved the conflict disclosure above the ranking. Added the regional tool table, the Arabic test protocol, jurisdiction sections for the UAE, Saudi Arabia, Egypt, Qatar, Kuwait, Lebanon, Jordan and Morocco, the PDPL and DIFC/ADGM residency section, the pricing table and the limits section.
• 9 July 2026 — First published, ranking five tools.

• Harvey — product site (checked 22 Jul 2026)
• Legora — product site (checked 22 Jul 2026)
• Spellbook — pricing page (checked 22 Jul 2026)
• Lexis+ AI — product page (checked 22 Jul 2026)
• CoCounsel, Thomson Reuters (checked 22 Jul 2026)
• Lexzur — product site (checked 22 Jul 2026)
• UAE government — data protection laws
• SDAIA — Saudi Personal Data Protection Law
• DIFC — laws and regulations
• ADGM — legal framework
• HAQQ pricing
• HAQQ security and data residency
• The full 50-task benchmark, including where HAQQ loses

HAQQ provides legal information, not regulated legal advice. The benchmark cited here is HAQQ's own internal testing, not an independent or audited evaluation, and scores describe model performance on defined tasks rather than a guarantee of outcome. Data protection references name the governing instruments only; confirm the current text and what it requires of your firm with qualified counsel in your jurisdiction before relying on any of it. Vendor claims are what each vendor published on the date shown and may have changed since.