How to Impeach a Medical Expert Witness: 8 Proven Techniques

In medical malpractice and personal injury cases, the expert witness is often the most important witness the jury hears. A strong cross-examination of the opposing expert can reshape the entire case. Weak cross puts the expert's credibility on autopilot — and lets them run the narrative.

This guide covers eight impeachment techniques that work consistently, with examples of how to phrase each question and how AI deposition analysis tools surface openings in real time.

1. Prior inconsistent testimony

The gold standard. If you have the expert's prior deposition or trial transcripts from other cases, read the exact passage, then ask the current question. When the answers don't match, you have impeachment gold for closing argument.

"In the Smith v. Mercy Hospital case on [date], at page 47 of your deposition, you testified that informed consent is required for any drug with a 1% or greater risk of serious adverse events. Do you remember that?"

This works because juries intuitively distrust experts who change their story case to case. AI deposition tools make this technique dramatically more effective — you upload prior depositions in advance, and the tool flags contradictions in real time with exact quotes from both.

2. Bias from compensation

Rule of thumb: ask directly and specifically. Not "Are you being paid?" but:

"What is your hourly rate as an expert witness?"
"How many hours have you billed on this case to date?"
"What portion of your annual income comes from expert witness work versus clinical practice?"
"In the past three years, how many cases have you been retained on — and of those, what percentage were for plaintiffs versus defendants?"

A defense expert who makes $200K/year from insurance companies — out of $300K total income — is not a neutral opinion. Jurors see this immediately when it's framed concretely.

3. Scope of review

What the expert didn't review is often more valuable than what they did. Get a complete list of materials reviewed, then ask about what's missing:

"Did you review the post-incident MRI?"
"Did you review Dr. Jones's deposition?"
"Did you review the nursing flow sheets from the night shift on [date]?"

Every "no" is a Rule 702(b) data-sufficiency attack that simultaneously undermines the expert's credibility with the jury.

4. Literature contradictions

Work with your own medical expert to identify peer-reviewed literature that contradicts the opposing expert's position. Have the articles ready. Under learned treatise doctrine (Fed. R. Evid. 803(18)), you can use authoritative medical literature on cross.

"Are you familiar with the [Year] NIH Consensus Statement on [topic]?"
"Would you agree that the [Journal] publication from [Author] is authoritative in this field?"
"Let me read you the conclusion of that paper: '[Quote].' Does that conclusion differ from your opinion in this case?"

5. Qualifications scoping

If the expert is testifying beyond their narrow specialty, expose the gap. An orthopedic surgeon opining on ER triage decisions; a general surgeon opining on vascular reconstruction; an internist opining on obstetric complications — these scope-overreach attacks can be devastating.

"Are you board-certified in emergency medicine?"
"How many times in the past five years have you personally triaged a patient in an emergency room?"
"Is it fair to say your clinical practice is primarily [their actual specialty], not ER medicine?"

6. Methodology absence

When the expert says "based on my clinical experience," press for the methodology:

"When you say 'based on clinical experience,' can you describe the specific methodology you used to arrive at this opinion?"
"What peer-reviewed literature did you consult?"
"What differential diagnosis process did you follow, and what alternatives did you rule out?"

Experts who can't articulate a methodology beyond intuition get excluded under Rule 702(c) reliability analysis.

7. Non-responsive answers and evasion

Track when the expert answers a question that wasn't asked — often in minute 45 when they're tiring. Each deflection is a credibility wound the jury notices. Call it out:

"Doctor, the question was [repeat question]. Not what you just answered. Can you answer my question?"

Do this once or twice. After that, let the pattern speak for itself at trial.

8. Internal contradictions within the current deposition

The hardest to catch live. The expert defines "standard of care" narrowly at minute 20, then applies it broadly at minute 90. Or concedes a fact early, then denies it later. Or gives a precise number at the start and rounds it up at the end.

This is where AI deposition analysis earns its keep. Running live alongside the deposition, AI tools cross-reference every statement against every prior statement in the same transcript. When the expert contradicts themselves, the AI flags it with both quotes shown side-by-side — in time for you to use it in the same deposition.

Example flagged by Courtroom AI in a recent cardiothoracic surgeon deposition:

Minute 27: "I always review the full operative note before opining on surgical technique."
Minute 78: "I didn't see the full operative note in this case — I only had the discharge summary."

That's a Rule 702(b) data-sufficiency admission AND a credibility impeachment. Jury hears both at trial.

The order of attack

Don't open with your strongest impeachment. Save it for the middle third of the deposition, after you've:

1. Locked in the expert's qualifications (so they can't narrow them later)
2. Gotten commitments on methodology
3. Walked through what records they DID and DIDN'T review
4. Established the literature landscape

Then pivot. Use the prior inconsistent testimony. Use the literature contradiction. Use the qualifications overreach.

Save the weakest impeachment for last — because by then the expert is tired, flustered, and more likely to concede on cross.

Related reading

• How to Daubert Challenge an Expert Witness in 2026
• How AI Analyzes a Two-Hour Deposition in Real Time
• How AI Catches Expert Witness Inconsistencies