The Honest Review Picture: One Voice, Lots of Vendor Numbers

Before you spend $400+ a month on AI legal research, you should know something: almost nobody has independently tested it. The most substantive hands-on review of Westlaw CoCounsel comes from a single credentialed practitioner, and everything else on the vendor's pages is marketing.

Appellate specialist Jeffrey I. Ehrlich spent 90 days with CoCounsel, uploaded 50,000 PDFs, and published his verdict in Plaintiff Magazine in November 2023. His conclusion was balanced: "CoCounsel is impressive. I can see use cases for some law firms that would justify the investment. But after using it for 90 days, I concluded that it was not worth $500 per month, on top of my Westlaw subscription." He found real strengths — a deposition-outline tool that "seemed to cover most of the important issues," and transcript summaries generated in roughly 8 minutes — but also a hard 50-result cap on database searches and citator checking that was flat-out unreliable.

That's one voice. Plaintiff-side, pre-dating the Westlaw integration, but honest about limitations. For an ethics-sensitive defense audience, it's the only review in the corpus you can actually weigh.

Everything else is Thomson Reuters talking about Thomson Reuters. The product page cites a Forrester study commissioned by the vendor claiming 76% of users saw improved research and drafting quality and a 33% average reduction in time spent on document review, research, and drafting. Commissioned research isn't fabrication, but it's not independent either — the client chooses what gets published.

Weaker still, the vendor's criminal-defense blog claims a 63% reduction in discovery-review time and 79% faster timeline building, attributed only to anonymous "attorney reports" with no methodology, no sample size, and no named practitioners. Treat those as positioning, not evidence.

• Independent hands-on review — Ehrlich's 90-day test: detailed methodology, named author, explicit caveats. Highest credibility.
• Vendor-commissioned study — Forrester's 76%/33% figures: real survey firm, paid client, unpublished raw data.
• Unsourced vendor blog stats — the 63%/79% defense claims: anonymous, unverifiable, promotional.

Notably absent: any independent defense practitioner reviewing CoCounsel at all, and any source — vendor or otherwise — describing CoCounsel reviewing bodycam video, interrogation footage, or jail calls with citations to the exact moment. Its documented skills center on documents, research memos, and deposition questions. That evidence-media gap matters for defenders comparing tools: if your review question is "does the police report match the bodycam," a research assistant isn't the tool being reviewed. Evidence Copilot exists for that other half of the workflow — grounded, cited evidence review — while CoCounsel's honest review picture remains "impressive speed, real limitations," on the strength of one careful reviewer's word.

What Reviewers Say CoCounsel Does Well

The most detailed independent review of CoCounsel comes from appellate specialist Jeffrey I. Ehrlich, who tested it for 90 days and uploaded 50,000 PDFs before rendering a verdict. His conclusion on capability was blunt: "CoCounsel is impressive." The question for a defense practice is where that speed actually holds up.

Transcript summaries are the standout. Ehrlich found CoCounsel generated a summary of a normal volume of trial testimony in roughly 8 minutes — fast enough to change how you triage a preliminary-hearing prep file, per his first-hand review. The caveat: good for gist, not good enough to cite in a brief without checking the transcript yourself.

The corroborated strengths from reviewers and vendor documentation cluster around four things:

• Fast transcript summaries suitable for triage and gist-level review
• Deposition-outline generation that "seemed to cover most of the important issues" — a time-saver for repetitive case types
• Research memos at junior-associate speed: a patent lawyer's test found summer clerks' memos "substantially superior," but they took 3–4 days versus CoCounsel's under 10 minutes
• RAG-grounded citations drawn from Westlaw content, which institutional guidance says "significantly reduces the risk of hallucinations" compared with general-purpose LLMs

The privacy posture is a real differentiator worth noting. Westlaw states its contracts prohibit using user-provided documents to train its models — a meaningful line for defenders whose files contain client-privileged material. That's not a claim most general AI tools can make.

But even the vendor's own guidance pulls the punch. Westlaw's materials acknowledge that AI summaries "can still misrepresent case holdings or statutory language" and should "never be used to advise a client, write a brief or motion for a court, or otherwise be relied on without doing further research." Ehrlich's testing bore that out: CoCounsel once cited a California appellate opinion that had been reversed by the state Supreme Court without flagging it.

That verification burden is the honest bottom line. The speed is real; the reliability requires a lawyer in the loop. It's the same principle we build Evidence Copilot on for evidence review — every AI output cites its source moment and stays marked pending human review, because no tool earns trust without it.

The Weaknesses Reviewers Actually Hit

The praise for CoCounsel comes with caveats that matter a lot more for a criminal-defense practice than the marketing suggests. The most substantive independent review — a 90-day hands-on test by appellate specialist Jeffrey I. Ehrlich — found real speed alongside limitations serious enough that he canceled his subscription rather than pay $500/month on top of Westlaw.

The most concrete complaint: database searches are hard-capped at 50 results per query, no matter how many hits exist. Ehrlich contacted support, who confirmed the cap is by design. For a solo defender drowning in a 3,000-page discovery dump, that ceiling is the difference between a research tool and a research toy.

Then there's the citator problem. Ehrlich found CoCounsel either did not run citator checks at all, or did so unreliably — at one point citing a California appellate opinion that had been granted review and reversed by the state Supreme Court without flagging it. Westlaw's own instructional materials concede that AI summaries "can still misrepresent case holdings or statutory language" and should never be relied on without further research, per a Penn State law library guide. His memos also tended to list broadly relevant cases rather than on-point ones, and he noted that prompting skill — knowing how to formulate queries — is itself a learned competency.

But the limitation that should stop defense attorneys cold is buried in Thomson Reuters' own documentation. The Review Documents skill reads each file independently and "cannot complete requests that require CoCounsel to compare or contrast information across separate files", per the vendor's product documentation. That is precisely the cross-file task at the heart of defense work: does the officer's report match the bodycam, does the interrogation-room footage match the Miranda timeline, does the jail call contradict the statement in the police report. A tool that cannot compare across files cannot do report-vs-bodycam impeachment work.

The remaining complaints cluster around access and cost:

• Westlaw lock-in: CoCounsel is tied to Westlaw subscribers, with premium pricing bundled into the subscription
• The reviewer's verdict that the product was impressive but not worth $500/month as a standalone add-on
• Document-review requests limited to roughly 30 questions per session

None of the review corpus — vendor or independent — describes CoCounsel handling bodycam video, interrogation footage, or 911 audio with second-level citations at all. Its documented skills center on documents, research memos, and deposition questions. For defenders evaluating vendors, that gap is the honest takeaway: CoCounsel is a research and memo tool with real limits, not an evidence-review workspace. If your bottleneck is comparing the report against the footage, a grounded evidence-review platform like Evidence Copilot — built specifically for cross-file inconsistency work with citations to the exact video moment — addresses the task CoCounsel's own documentation says it can't.

What No Review Covers: Your Bodycam, Jail Calls, and 911 Audio

Here's the thing nobody tells you before you buy: every review of CoCounsel — including the vendor's own materials — is about documents, research, and depositions. Not one source describes it reviewing bodycam video, interrogation footage, jail calls, or 911 audio with citations down to the exact moment.

Scan the documented skills and the pattern is clear. CoCounsel's vendor documentation covers document review, research memos, and deposition questions. Thomson Reuters's defense-focused blog touts discovery review, timelines, and trial prep outlines. The Penn State law library guide to Westlaw's AI tools describes research features only. Media evidence — the bulk of a criminal discovery dump — never comes up.

That matters for defense work because your case lives in the media. The officer's report says consent to search; the bodycam says otherwise. The jail call contradicts the statement. Finding those conflicts is the job, and it requires tooling built for evidence, not briefs.

There's also a structural limitation worth knowing. Vendor documentation states CoCounsel reads each file independently and "cannot complete requests that require CoCounsel to compare or contrast information across separate files." Report-vs-bodycam impeachment is exactly that kind of cross-file comparison.

Evidence Copilot exists to fill this gap — as the alternative for evidence-media review, not a replacement for CoCounsel's research strengths. The workflow looks like this:

• Diarized, timestamped transcripts of bodycam, interrogation footage, jail calls, and 911 audio, with word-level timestamps and one-click speaker renaming.
• OCR'd police reports and scanned discovery, so every page becomes searchable, citable text.
• Auto-signals on ingest: each new file is compared against earlier evidence in the case, surfacing candidate inconsistencies — report vs. bodycam, statement vs. call audio — before anyone asks.
• Every AI answer cites the exact video second or PDF page it came from, marked AI-generated and pending human review, with originals kept immutable.

That last point is the honest dividing line. CoCounsel's citator-checking reliability drew criticism from its most detailed independent reviewer, who found it sometimes failed to flag reversed authority. When the stakes are a suppressed statement or a cross-examination moment, you need citations that point to the source itself — the frame, the second, the page — not just a linked case.

So the fair summary for a defense practice: CoCounsel remains a legitimate research and drafting tool, and the reviews reflect that. But if your bottleneck is 40 hours of bodycam and a police report that doesn't quite match, no review of CoCounsel covers that workflow — because the product wasn't built for it. That's the gap a grounded evidence-review workspace is designed to close, with human review on every output and published pricing you can verify before you talk to anyone.

How to Evaluate CoCounsel (or Any Legal AI) for a Defense Practice

The review evidence on CoCounsel teaches a broader lesson: whatever legal AI tool you're considering for a defense practice, test the specific things that break under real caseload before you trust the marketing. The hands-on review from appellate specialist Jeffrey Ehrlich — who used the product for 90 days and uploaded 50,000 PDFs — is the most useful template for how to do that.

Start with citation verification. Ehrlich found that CoCounsel "either did not 'citate' or 'Shepardize' the cases it cited, or that it tried, but did so unreliably" — including citing a California appellate opinion that had been granted review and reversed by the state Supreme Court without flagging it. Don't accept a vendor's "linked citations" claim at face value. Run a test query with cases you already know, then check the citations yourself.

Second, check search and result caps against your actual discovery volumes. Ehrlich found search results hard-capped at 50 per query regardless of total hits, and support confirmed this was by design. He called the feature "essentially worthless" for e-discovery-scale review. If your typical case involves thousands of pages of discovery, ask the vendor directly what the caps are.

Third, ask whether the tool compares across files or reads them independently. Thomson Reuters's own documentation states CoCounsel reads each file independently and "cannot complete requests that require CoCounsel to compare or contrast information across separate files," with requests limited to roughly 30 questions. That matters enormously for defense work — spotting a police report that contradicts bodycam footage, or a jail call that conflicts with a client's statement, is exactly cross-file comparison.

Finally, check pricing access before you commit. Ehrlich canceled at $500/month on top of his Westlaw subscription, concluding it wasn't worth the price as a standalone add-on. A comparison guide argues Thomson Reuters "locked CoCounsel to Westlaw subscribers" with premium bundled pricing — though that source is promotional itself, so weigh it accordingly. Either way, you should be able to see what you'll pay without sitting through a sales call.

That last standard is worth holding every vendor to. Evidence Copilot publishes its pricing openly — Solo at $39/user/month, Firm at $99/user/month, and a PD Office tier at $750/attorney/year — with no feature tiers and no sales wall. And because defenders live in evidence media, not case law, every AI answer in the workspace cites the exact second of video or page of PDF it came from, always marked for human review.

If you want to see how that grounded workflow handles a real discovery dump, you can book a 20-minute walkthrough — no pressure, no urgency tactics.