Tax research and legal drafting tools, compared for 2026

AI in tax and legal work has settled into three recognisable categories: research, drafting, and planning or prediction. Most products sit firmly in one. A smaller number try to bridge them.

That distinction matters more than it looks. A research tool that cannot draft, or a drafting tool that cannot cite a statute, just reintroduces the manual step you bought the tool to remove. For work across European jurisdictions, the gap between categories is where the hours go.

How the market splits

Research tools answer questions about law, regulation or guidance. The quality signal is citation accuracy: does the answer point at the actual provision, or does it paraphrase and hope?

Drafting tools take research, client facts or instructions and produce a document — a memo, a client email, a clause, an opinion. The quality signal is whether the output is usable or needs rebuilding before it goes anywhere near a client.

Planning and prediction tools model outcomes, score risk and run scenarios. Different job, different evaluation.

Regulators and professional bodies are paying closer attention to how AI-assisted work product is produced and documented. Whatever form that takes in your jurisdiction, it pushes in one direction: citation quality and auditability become requirements rather than nice-to-haves — which is an argument for tools that can show what they read and check their own output, not just tools that sound confident.

azta

azta is built for tax and legal professionals, and not for tax alone — the researcher works across the whole statute book, employment, company and commercial, contract and civil, and regulatory duties included. Ask in plain language, get an answer cited to the statute, directive or official guidance. It reads client files with real OCR and passage-level highlighting, connects to QUiCK for live financial data, and turns the research into a summary, advisory memo or client email in Canvas, exported to Word or PDF.

The differentiator for European work is coverage and citation architecture. Across 24 European jurisdictions, all indexed daily, 17 run on a tree-law engine: statutes ingested as documents, units, attributes and references, researched by walking that structure rather than chunking flat text, so cross-references resolve deterministically and a citation carries a stable legal-unit identity instead of a page offset. Ask about the VAT treatment of a cross-border digital service between Germany and Poland and the answer cites the relevant article and the national implementing provision, not a general description of the rule.

Two mechanisms sit on top of that. A verifier audits every answer before publication — citation coverage, internal contradictions, temporal risk from recent amendments, and whether the answer matches the scope of the question — and can send the agent back for more research. And on a multi-country question a dispatch-coverage gate refuses to answer until every jurisdiction that surfaced a candidate law has had that law actually researched, so a thin jurisdiction cannot ride along on a well-covered one.

The remaining seven jurisdictions run the older research flow. That is a real difference in depth, and any honest evaluation should start by asking which tier your own markets sit in.

The workflow is one motion. Research flows into drafting without a handoff, so a client question becomes a finished memo in the same session — which is what makes several jurisdictions in one engagement manageable. Pro is €27 a month or €270 a year, self-serve; the free tier is five questions a month; and the Statute Finder, a browser and reader over the statute book itself, is free to every signed-in user. Firms needing their own tenant run a dedicated single-tenant deployment with its own domain, database, SAML SSO and branding.

Best for: tax advisors and lawyers working across European jurisdictions who need research and drafting in one place.

TaxGPT

An AI assistant for tax questions with a visible presence in the US market and a clean, accessible interface. On publicly available information its focus is US federal and state tax; cross-border European research, multi-country VAT questions and European directive interpretation are not the core use case. It also sits more firmly in the research category, with a lighter drafting path than tools that produce finished documents from the research.

Best for: US-focused tax professionals who want quick answers on domestic questions.

Blue J

A prediction and risk-scoring platform. Its core capability is assessing how a position is likely to be treated, based on case law patterns and statutory interpretation, and it is well regarded for that in Canada and the US. For European cross-border work the coverage is narrower, and the strength is prediction rather than producing client-ready documents. Pricing is not publicly listed and is disclosed on request, which tends to keep smaller practices out of the evaluation.

Best for: tax teams at larger North American firms focused on risk scoring and outcome prediction.

CoCounsel Tax

Thomson Reuters’ AI layer over Checkpoint. The integration is both the strength and the constraint. Research quality benefits from a deep content library, particularly for US federal tax and established practice areas, and the drafting side is built for professional use.

For smaller practices the entry path is the issue: it is sold as part of the Thomson Reuters ecosystem rather than self-serve. And for European cross-border work, what it can reach depends on which modules the firm subscribes to, so the breadth of national law that European advisors need is not uniformly available.

Best for: large firms and enterprise legal departments already running Checkpoint.

What to weigh

Citation architecture beats speed. A fast answer that cites a summary rather than the article itself is not professional work product. You have to be able to trace it and stand behind it.

Integrated drafting reduces risk. When research and drafting live in separate tools, the translation between them is where errors enter. One workflow means the memo reflects the research rather than a recollection of it.

European coverage is not uniform. "Supports European tax" can mean a handful of VAT summaries or full national implementations. Ask about the specific jurisdictions you work in, and whether citations resolve to the actual legal text.

Accessibility matters for smaller practices. If you cannot evaluate a tool without an enterprise subscription, that is a structural barrier, not a pricing detail.

The drafting gap

Even among tools that claim drafting, quality varies. Producing a document is not the same as producing a document you can send. The difference is whether it reflects the facts of this client’s situation, cites the right authorities and reads the way your practice communicates.

azta reads the client file — with OCR on scanned PDFs, and the cited passage highlighted inside the original document — pulls the relevant figures and produces output you edit rather than rebuild. That is still not common across the category.

Competitor descriptions here reflect publicly available positioning and have not been independently verified.