See how AI medical-record review links every fact to the exact Bates page that proves it — click any citation and jump straight to the record.
See the 60-second demo →Every medical malpractice plaintiff attorney has had the post-deposition realization. You walk out of a four-hour expert witness deposition, drive back to the office, and two days later your LNC reads the transcript and says "you know, on page 143 the defense cardiologist admitted he hasn't read the peer-reviewed literature on this specific technique in eight years." You read it. You confirm it. And you spend the next week trying to remember what else you missed.
The Daubert motion you could have filed the next morning — with a clean, on-the-record admission of a methodology problem — is now a two-week reconstruction project. Sometimes it still works. Sometimes you've already passed the motion deadline.
The value proposition of AI during an expert deposition is precisely this problem. A tool that flags potential Daubert issues while the witness is still in the room, while you can still ask the follow-up question on the record, is a different deposition than one where the insights come after the fact.
This post walks through what the four Daubert factors are, how an AI system like Courtroom AI actually flags them in real time, three concrete examples of AI-flagged moments from recent depositions, and a checklist for what to do when the AI flags something during live testimony.
Daubert v. Merrell Dow Pharmaceuticals, 509 U.S. 579 (1993) established that federal trial judges serve as gatekeepers of expert testimony. The Court identified four non-exclusive factors judges should consider when evaluating the reliability of an expert's methodology:
Kumho Tire Co. v. Carmichael, 526 U.S. 137 (1999) extended the Daubert framework to all expert testimony, not just scientific testimony. And the 2023 amendments to Federal Rule of Evidence 702 clarified that the proponent of expert testimony must demonstrate to the court by a preponderance of the evidence that the testimony is reliable — not just that it's "from an expert."
The operative text lives at Cornell's Legal Information Institute Rule 702 page. That's the standard any expert-witness deposition is implicitly tested against, and the standard an AI analysis layer is checking each batched witness statement against in real time.
Courtroom AI runs four analysis lenses in parallel against every batched witness statement during a deposition. One of the four lenses is specifically trained on Daubert vulnerability detection. The analysis prompt encodes the four Daubert factors plus the post-Kumho extension and the 2023 Rule 702 amendments.
For each witness statement, the AI produces a vulnerability classification (HIGH, MEDIUM, LOW, or NONE) along with the specific basis for the flag. Typical flag categories include:
The flag surfaces on the attorney's screen within 5-15 seconds of the witness finishing the answer. That's fast enough to ask the follow-up while the witness is still on the clock.
Cardiothoracic surgeon deposition, defense expert, medical malpractice case involving a TAVR (transcatheter aortic valve replacement). Witness testifies:
"In my experience, this procedure has a very low complication rate when performed correctly. Most patients do well."
AI flag (HIGH): "No methodology cited. 'Very low complication rate' is quantitative claim without source. No peer-reviewed literature referenced. No error rate disclosed. Factor 3 (known error rate) vulnerability."
Generated cross-exam question: "Doctor, when you say TAVR has a 'very low complication rate when performed correctly,' what specific complication rate are you testifying to? Can you identify the published study or registry data that supports that figure?"
This is the question that, on the record, forces the expert to either produce a source or concede they don't have one. Either outcome is good for the plaintiff on a future Daubert motion.
Emergency medicine physician, deposition in a delayed-stroke-diagnosis case. Witness testifies:
"I don't have the EKG in front of me, but based on the way the presenting symptoms were described, I'd say this was a classic TIA presentation, not an evolving stroke."
AI flag (HIGH): "Evidence disclaimer with testimony. Witness disclaims access to primary data (EKG) then testifies to differential diagnosis. Factor 1 (testability) and foundation both compromised. Rule 702(b) data-reliability vulnerability."
Generated cross-exam question: "Doctor, you testified this presentation was 'classic TIA, not evolving stroke,' but you also testified you don't have the EKG in front of you. What specific clinical data have you actually reviewed to support that diagnostic conclusion?"
If the witness is going to testify about a diagnosis, they need to have reviewed the data. An admission of non-review, on the record, paired with a conclusion about that same data is the textbook predicate for a Rule 702(b) challenge.
Deposition of a defense-retained expert in a birth injury case. The expert was designated to testify on standard of care for labor and delivery nursing. In response to a question, the witness drifts:
"Based on my review of the fetal monitor strips, the Category II pattern did not meet the threshold for emergent intervention under the ACOG guidelines."
AI flag (MEDIUM): "Possible scope creep. Witness designated as L&D nursing expert but testifying to obstetric interpretation of Category II fetal heart rate patterns — typically OB/MFM scope. Factor 4 (general acceptance within witness's own specialty) and Rule 702(a) qualification vulnerability."
Generated cross-exam question: "Nurse, you were designated to testify on labor and delivery nursing standards. Are you now offering opinions on the obstetric interpretation of Category II fetal heart rate tracings under ACOG guidelines? Is fetal heart rate interpretation within the scope of your nursing practice or within the scope of an obstetrician's practice?"
A good nurse witness can interpret a strip. Whether she can offer expert testimony on ACOG guideline thresholds for intervention is a different question, and one that can be raised in a post-deposition motion to strike or limit her testimony.
A few honest limits to avoid overselling the tool:
A plaintiff med-mal firm that previously had to choose between "depose the defense expert thoroughly" and "stay inside budget" no longer has to make that tradeoff the same way. The cost of running an AI analysis layer during the deposition is a rounding error (see our cost breakdown). The cost of missing a Daubert-viable admission at deposition and losing a motion-in-limine opportunity at trial is much larger.
And because the AI runs alongside (not instead of) the certified court reporter — see our companion post on whether AI can replace a court reporter — there's no friction with the existing deposition workflow. The reporter still produces the certified transcript you'll quote in your motion. The AI just made sure you knew what to quote.
For firms modeling total Daubert-practice ROI across a multi-case docket, the Courtroom AI ROI page walks through the case-lifecycle numbers with a specific focus on motion-practice leverage. A feature-by-feature look at the Daubert lens, prior-testimony cross-check, and other live analysis layers lives on the Courtroom AI product page, with a side-by-side comparison against other live deposition tools.
The 2023 amendments to Rule 702 matter. The amendment made explicit that the proponent of expert testimony bears the burden of establishing, by a preponderance of the evidence, that the expert's opinion reflects a reliable application of the principles and methods to the facts of the case. That's a real shift from the pre-amendment standard in some circuits.
Courtroom AI's Daubert lens was updated in Q4 2025 to incorporate the post-amendment standard. Specifically, the lens now flags instances where the expert testifies to a conclusion that goes beyond what their methodology actually supports — the "overreach" problem the amendment was designed to address.
If you're drafting a Daubert motion in 2026 and you haven't read the current text of Rule 702 with the committee notes, stop and do that first. The old case law still applies, but the pleading burden has shifted in ways that catch practitioners off guard if they're citing pre-2023 authority.
COURTROOM50 at checkout = 50% off 3 months. Essentials becomes $49.50/mo, Pro becomes $149.50/mo. Capped at 10 redemptions. Details on the Courtroom AI overview or calculate your ROI.
Daubert motions are won or lost at deposition, not at the motion hearing. The admission that shows up on page 143 of the certified transcript is either there or it isn't. An AI analysis layer that flags Daubert-viable statements in real time — while you can still ask the follow-up question — shifts the deposition from an information-gathering exercise into a record-building exercise.
That's a real change in what an expert deposition is for. And it's available for the price of a deli sandwich per depo.
Questions on Daubert strategy or pilot access? Email [email protected]. We're happy to run a live demo against one of your actual expert deposition recordings before you commit to anything.