Clearbrief

What it is

Clearbrief is a Microsoft Word add-in that checks whether the record supports the sentence you just wrote. It is not a research assistant and it does not draft the brief for you. Point it at a draft and its Analyze feature reads each factual assertion, searches the deposition transcripts, exhibits, and records you loaded, and returns the passage that supports it — or reports that it found none. The patented Add Fact-Cite button drops the citation in, and every citation it writes is hyperlinked inside the Word document, so a reviewing partner or a judge clicks the cite and lands on the source page. Clearbrief was founded by Jacqueline Schafer, a former appellate attorney who served as an Assistant Attorney General in Washington and Alaska, and has raised roughly $8M across a $1.2M seed and a $4M round in June 2024 that included a fund affiliated with Wilson Sonsini.

Around that core sit the tools a litigator uses on the same corpus: hyperlinked Timelines generated as a Word table with a record cite on every row, Table Builder, narrative or tabular summaries of deposition and hearing transcripts, automatic Tables of Authorities, exhibit and filing hyperlinking, citation-format conversion between the Bluebook and state style manuals, and a cross-examination outline built while testimony is running. The vendor reports that more than 3.2 million pleadings have been drafted or checked on the platform since its 2021 launch.

Why it shows up in legal-ops stacks

• It targets the defect class a citator cannot see. The AI Hallucination Cases database maintained by Damien Charlotin holds 1,934 decisions worldwide in which a court found a party relied on hallucinated material. Across the cases with tagged findings it records 5,748 individual defects, of which 3,238 are fabrications. The remaining 2,475 — misrepresented authorities, false quotes, superseded law — are real citations that KeyCite and Shepard’s return clean, because the authority exists and is good law; it simply does not say what the brief says it says. Clearbrief’s Analyze checks that last question against the record and the cited text, which is the gap the Westlaw and Lexis citators structurally leave open.
• The output is a review artifact, not a chat transcript. Verification lands as hyperlinks inside the filed document, so the check survives the handoff to whoever reviews it. That is the difference between a control an associate ran and a control a partner can audit in ten seconds.
• Institutional buyers have already run the procurement. The AAA-ICDR rolled Clearbrief out to its panel of 5,500 arbitrators and mediators in January 2025 after a six-month pilot. The vendor lists adoption across the AmLaw 200, state and federal courts, government agencies, and Microsoft’s own legal department. It is SOC 2 Type 2 certified, states that customer documents are never used to train language models, and offers a bring-your-own-storage option that keeps documents in the firm’s cloud.

Pricing reality

The Solo plan is $300 per user per month, published on the pricing page and checked 2026-08-20. That is the only number Clearbrief prints. Enterprise Unlimited is quote-based with volume discounts, bundles 1:1 training, and carries no published floor; the vendor directs Solo buyers who want a lower rate to email about multi-year agreements. There is no self-serve free trial on the pricing page — the enterprise path is a demo followed by a pilot, which is how the AAA-ICDR deal ran.

Budget it as an add-on, not a replacement. Clearbrief carries no primary-law corpus, so the firm still pays for research: Paxton runs $499 per user per month billed monthly or about $250 effective on an annual commit, and a Westlaw Advantage seat bundled with CoCounsel Essentials lands near $639 per user per month. A litigator running Clearbrief plus one research subscription is therefore at roughly $550 to $940 per user per month all-in. The comparison that clears procurement is not tool-versus-tool but seat-versus-hours: a $300 seat is under two billable hours a month at mid-market rates, against a paralegal manually walking a 20-deposition record.

Best for

A litigation or appellate team whose deliverable is a filed document with a record cite behind every factual assertion — product liability, personal injury, employment, civil rights, appellate practice, and arbitration neutrals writing awards. The return is highest where the record is large and the drafting window is short, which puts the sweet spot at 1-to-50-lawyer litigation practices where the paralegal record-check is the throughput ceiling.

Skip it if the practice is transactional. Clearbrief checks assertions against a record, and a contract has no record — Definely, Spellbook, and DraftWise are the Word-native tools for that work. Skip it if the team drafts anywhere other than Word, because there is no Google Docs surface. And skip it if you want the check to run as an automated gate: there is no public API and no MCP server, so it cannot be wired into a pre-filing pipeline.

Versus the alternatives

The two incumbents are Thomson Reuters CoCounsel and LexisNexis Protégé. Pick CoCounsel when the question is whether the brief characterizes the law correctly — Westlaw Advantage’s Litigation Document Analyzer, built on the Quick Check engine, reviews an uploaded draft against KeyCite, flags mischaracterizations of law, and surfaces authority the brief missed, none of which Clearbrief does. Pick Protégé for the same job in a Lexis shop, where Brief Analysis runs a full Shepard’s pass on the draft. Pick Clearbrief when the risk you are managing sits in the facts rather than the law, and when the verification has to live in the document instead of a separate report. The three are complements rather than substitutes: Clearbrief’s own case-citation check resolves through a LexisNexis integration rather than an owned corpus.

Litera is the alternative when the requirement is firm-wide document tech across practice groups — proofreading, comparison, TOA generation — rather than litigation record verification; its coverage is wider and its fact-checking is not. Harvey is the fastest-growing entrant in this segment and the pick at enterprise scale, where one platform has to cover research, drafting, and matter work, now that it reaches primary law through its LexisNexis alliance.

If none of them fit — a small team that cannot justify another seat — the fallback is to run the record check yourself. The citation verification Claude Skill does the same third-question check at zero license cost, at the price of assembling the corpus by hand for every brief.

Watch-outs

• Silence is not a clean bill of health. Analyze finds support only in the documents you loaded. If an exhibit never made it into the corpus, the tool reports no supporting text — which reads identically to an unsupported assertion. Guard: assemble the full record set through the iManage, Relativity, or NetDocuments connector before the first run, and treat every “no supporting text found” as an open item assigned to a named person rather than a failed check.
• The case-law hallucination check rides on a Lexis subscription. Clearbrief verifies case citations by linking them into Lexis and Lexis+, so the depth of that check and the usefulness of the resulting links track what the firm’s Lexis contract covers. Guard: during the pilot, run a brief containing ten citations drawn from the courts you actually practice in, and confirm which of them resolve on the firm’s own subscription before you price the rollout.
• No API and no MCP server means it cannot be a gate. Every check is a human deciding to click Analyze, and a step that can be skipped is not a control. Guard: write the run into the filing SOP as a named step with a sign-off line on the filing checklist, and audit a sample of filings each quarter rather than assuming coverage.
• Word is the only surface. A practice group drafting in Google Docs sits outside the control entirely. Guard: standardize filed documents on Word before you count those lawyers in the seat total, or exclude them from both the license and the compliance claim.
• Above Solo, the pricing is dark. $300 is the only published figure, and Enterprise Unlimited has no floor, no published per-seat rate, and no public definition of what “unlimited” bounds. Guard: make the vendor quote in dollars per user per year, get the unlimited definition and the training entitlement written into the order form, and benchmark the total against the Solo rate times headcount before signing a multi-year term.