Claude for Word went into public beta this spring, and it landed on exactly the trajectory GC AI CEO Cecilia Ziniti had described on CZ and Friends, GC AI’s podcast:
“as AI gets more powerful, Claude is adding the ability to actually do things. We’re adding that as well in GC AI, where it’s like, okay, don’t just tell me what to type in my email, send the dang email.”
That ability arrived in the exact place lawyers work. Anthropic’s official add-in went into public beta on April 10, 2026, and it puts Claude inside Microsoft Word with edits that land as native tracked changes. For in-house counsel the question is concrete: how much legal work does Claude for Word cover, and what does a legal team still have to add? The answer runs in both directions.
What Is Claude for Word?
Claude for Word is Anthropic’s official Microsoft Word add-in, released in public beta on April 10, 2026 and installed through Microsoft AppSource. It runs as a sidebar that reads your open document, answers questions with clickable section citations, and proposes edits as revisions in Word’s native review pane. The add-in is free to install and requires a paid Claude plan; per Anthropic’s documentation, it is available on Pro, Max, Team, and Enterprise plans, with Pro at $20 per month as of August 2026.
It completes Anthropic’s Microsoft Office lineup alongside the Excel and PowerPoint add-ins, and the three share conversation context across apps. It is also one piece of Anthropic’s broader legal AI push, the piece built directly into the tool where contracts get drafted. Platform support covers Word on the web, Windows (Microsoft 365 version 2205 or later), and Mac (16.61 or later), working on .docx files.
What Claude for Word Does Well
The add-in gets the mechanics of document editing right, and several capabilities matter for legal work specifically.
Edits land as native tracked changes: Every proposed edit arrives as a revision you accept or reject in Word’s review pane, with deletions and insertions marked the way a human reviewer’s would be. This is the single feature that makes AI editing auditable.
It reads the whole document. Text, comments, tracked changes, footnotes, tables, and bookmarks are all in context, and answers come back with clickable citations to the section they rely on.
Style-preserving edits. Selected-text edits keep the surrounding numbering, heading styles, and formatting, which anyone who has pasted AI output into a numbered contract will appreciate.
Comment-thread workflow. Claude can work through a comment thread, edit the anchored text, and reply in the thread, which fits how redlines move between reviewers.
Template filling. It drafts content into templates while inheriting the document’s styles, a real assist on first drafts of routine paper.
For general drafting, a board memo, a policy rewrite, a first-pass summary, this is a strong, well-built layer. The gaps show up when the document is a contract and the work is legal.
What Claude for Word Misses for Legal Work
Four gaps decide whether the add-in can carry contract work, and Anthropic’s own documentation describes each one.
It Reviews Against Its Judgment, Your Playbook Lives Elsewhere
The add-in ships general document skills. Anthropic’s legal-specific layer, the Claude for Legal plugins with clause-by-clause review commands, runs in Claude Cowork as a separate surface from the Word add-in as of August 2026. Inside Word, a contract review runs on Claude’s general judgment of what looks standard, with your positions supplied prompt by prompt.
A legal team needs the review to run against its own standards: approved fallback language, walk-away terms, and per-clause guidance that applies the same way on the fortieth NDA as the first. That is playbook infrastructure, and GC AI for Word carries it as Playbooks, pre-built for NDAs, DPAs, and MSAs, plus custom playbooks built from your own paper.
One Document at a Time
Anthropic’s documentation states it plainly: Claude for Word “can only access the document you have open in Word.” Legal review rarely works on one document. The MSA reads against the order form, the amendment against the original, the DPA against your subprocessor list.
A matter-aware platform holds that context: GC AI’s Files keeps permanent document collections available across every chat, analyzing up to 1,500 pages at once, and Projects carries matter context between sessions.
Verification Stops at the Document Edge
Clickable citations to sections of the open document are genuinely useful, and they answer a different question than legal verification. When the output cites a statute, a case, or a market position, an in-house lawyer needs character-level proof from source documents and research grounded in authoritative databases. That is the job Exact Quote and Research exist for, and it is the difference the AI hallucination case tracker keeps documenting when lawyers ship unverified AI citations.
Memory and Privilege Run on Consumer Rails
Two lines from Anthropic’s own support page carry the most weight for legal buyers. First, chat history is “stored locally in your browser using IndexedDB” and does not sync across devices, so your review history lives in a browser cache. Second, Anthropic recommends against using the beta for “documents containing highly sensitive or privileged data without proper controls,” with backend inputs and outputs deleted within 30 days by default. Take the vendor at its word. ABA Formal Opinion 512 treats that vetting as part of the job: a lawyer’s duty of confidentiality covers evaluating a generative AI tool’s data handling before client information goes in. After the Heppner ruling, where AI outputs generated outside counsel direction lost privilege protection, the setup around the model decides as much as the model itself: who directed the work, where the chat record lives, and what retention terms govern it. GC AI is SOC 2 Type II and SOC 3 certified, GDPR compliant, with zero data retention agreements with OpenAI and Anthropic, and AES-256 encryption.
How Claude Scores on In-House Legal Work
The In-House Legal Bench, GC AI’s May 2026 benchmark of AI assistants across 100 real in-house tasks scored against 1,200+ attorney-developed criteria, tested Claude directly:
GC AI: 86.8%
ChatGPT (GPT-5.5): 79.8%
Claude (Opus 4.7): 68.4%
Gemini (3.1 Pro): 57.5%
The gap concentrates where in-house work is most specialized: regulatory tracking, legal research, and checklist-driven review. Claude’s underlying model is excellent, which is exactly why GC AI runs frontier models from Anthropic and OpenAI under a legal-specific layer: the system prompt, playbooks, and verification tooling are what convert a strong model into work a lawyer can sign.
Claude for Word vs GC AI for Word
Both products live inside Microsoft Word. They are built for different jobs.
Capability | Claude for Word | GC AI for Word |
Built for | General document editing | In-house legal work |
Tracked-changes redlining | Yes | Yes |
Playbook review against your standards | No | Yes, pre-built and custom Playbooks |
Character-level source citation | Document-section citations | Exact Quote across source documents |
Legal research with authoritative sources | No | Yes, Research with citations |
Matter context beyond the open document | No, single open document | Yes, Files and Projects |
Web research from inside Word | No | Yes, Chat2 in the Word Add-in |
Status | Public beta | Generally available |
Table facts verified against Anthropic’s Claude for Word documentation and GC AI for Word, August 2026.
Trisha Mauer, VP of Legal at Tonal, ran this comparison across general-purpose AI and reached the conclusion the table points to:
“I go straight to GC AI for everything from research requests to litigation responses. I’ve compared against ChatGPT, GC AI gives more comprehensive responses appropriate for a lawyer to use. After six months of use, I’m sure I’ve saved hundreds of hours.”
Her test is the right one: run the same matter through both and read the outputs like a lawyer. For the two platforms measured feature by feature, beyond the Word add-ins, see GC AI vs Claude.
Is Claude for Word Right for You?
Claude for Word is right for you if your team wants a well-built general drafting assistant inside Word for memos, summaries, and first-pass edits, and your confidential-document policy can live with beta-stage data handling. It does that job well. If the job is contract review against your own standards, research a lawyer can verify, and privileged work product handled on enterprise rails, that calls for a legal AI platform: start a 14-day GC AI trial and run it on the next redline that crosses your desk. For the category-level view of AI inside Word, the guide to having AI review a Word document walks the full workflow.
Frequently Asked Questions
Is There an Official Claude Add-In for Microsoft Word?
Yes. Anthropic released Claude for Word in public beta on April 10, 2026, installed through Microsoft AppSource. It runs as a sidebar in Word on the web, Windows (Microsoft 365 version 2205 or later), and Mac (16.61 or later), and works on .docx files.
Is Claude for Word Free?
The add-in itself is free to install, and using it requires a paid Claude plan: Pro, Max, Team, or Enterprise, with Pro at $20 per month as of August 2026. There is no standalone purchase for the add-in.
Is Claude for Word Safe for Confidential Legal Documents?
Anthropic’s own documentation recommends against using the beta for documents containing highly sensitive or privileged data without proper controls, and chat history is stored locally in the browser. In-house teams handling privileged work product should apply their AI use policy before loading contracts into any beta tool, and weigh platforms with enterprise security postures and zero data retention agreements for that class of work.
Can Claude Redline a Contract in Word?
Yes, mechanically. Claude for Word proposes edits as native tracked changes you accept or reject, which works well for language cleanup and first-pass drafting. Review against your negotiating positions still depends on the standards you feed it prompt by prompt, which is the job dedicated legal playbook features automate.
Can Claude for Word Replace a Legal AI Platform?
Not for contract work. Claude for Word handles general drafting and editing well, and its gaps are structural: review runs on Claude’s general judgment with your standards supplied prompt by prompt, context stops at the one open document, and the beta’s data handling draws Anthropic’s own caution for privileged work. A legal AI platform closes those gaps with playbook review, matter-wide document context, verified research, and enterprise-grade security.
