Every way the opposing expert can be broken — written down, before the deposition
Upload the report, the CV, the prior transcripts and the records. Get back a marked-up attack
map: contradictions with the Bates page, an FRE 702 attack for each subsection with the questions to ask,
real peer-reviewed literature that cuts against the opinion, and a public-data check on whether this expert
actually performs the procedure at issue.
A simulator gives you practice. This gives you the list.
Rehearsing against an AI examiner is useful, and we build that too. But the thing counsel
actually carries into the room is a document — one that names the weakness, cites the page, and hands you
the question.
Attack 1
Contradictions, with the page number
Where the report disagrees with the record, or with what this expert said under oath in another case.
Every contradiction quotes both sides and pins each to a true page — and for numbered transcripts, to
the printed page and line.
Attack 2
Methodology, by 702 subsection
A separate attack for each of 702(a) helpfulness, (b) sufficient facts, (c) reliable principles and
(d) reliable application — each with a short script of leading questions rather than a topic to
improvise around.
Attack 3
Qualification, from Medicare data
The CV says twenty-two years and board certified. Medicare says how many times this physician actually
billed the procedure at issue. Runs from an NPI in about a minute, no documents needed.
Attack 4
Literature that cuts against them
We pull the expert's factual assertions, query PubMed against each one, and write the literature
section from records that came back. Real PMIDs, linked, fetched before the brief is drafted.
What it looks like
A short excerpt in the format the tool produces.
This example is constructed, not a real case. The clinical scenario and the quoted testimony are
invented for illustration and involve no real patient, physician or matter. The two things in it that are
real are the ones you can check: the PubMed citation is a genuine record, and the Medicare figures in the
qualification block are live values from the CMS public dataset for the 2024 reporting year.
Attack 1 · Record contradiction
Neurological status on admission
Report: “The patient had no documented neurological deficit at any point prior to
transfer.” [CASE:expert-report.pdf p7]
Significance: The opinion on causation is built on an unremarkable neuro exam.
One nursing note removes the premise, and the witness must either concede they did not review it or
defend having read past it.
Attack 2 · FRE 702(c) — reliable principles
No methodology stated for the standard-of-care conclusion
The report concludes the standard of care was violated without identifying a governing
guideline, a specialty standard, or the method by which the conclusion was reached.
[CASE:expert-report.pdf p11]
1. Doctor, what method did you use to reach that conclusion?
2. Is that method written down anywhere?
3. Can another physician apply it to these records and reach your result?
4. Which published guideline states the standard you applied?
Attack 4 · Contradicting literature (real citation)
“An ICD remains indicated regardless of ejection-fraction recovery”
PMID 33636331
— Baman JR (2021), “Primary Prevention Implantable Cardioverter-Defibrillator Therapy in Heart
Failure with Recovered Ejection Fraction” [J Card Fail]. Directly addresses the population the
witness asserts is settled, and treats the question as open.
Attack 3 · Qualification check — live CMS data
⚠ Specialty mismatch
CMS lists this physician's specialty as Internal Medicine.
72% of physicians who bill 33249 — insertion of implantable defibrillator
system are Clinical Cardiac Electrophysiologists. Unlike the volume figures, this comparison is not
distorted by suppression, which makes it the most defensible signal in the report.
Measure
Value
Physicians nationally billing 33249
3,959
With volume high enough to be reported
886 (78% suppressed)
Median among reporting electrophysiologists
15 (25th: 13 · 75th: 19 · max: 61)
Reliability: question only. Absence of volume is not
proof of inexperience — establish payer mix on the record first.
The number every other vendor would have hidden
Medicare suppresses any physician/procedure combination with ten or fewer beneficiaries. How
much of the physician population that removes depends entirely on the procedure — and nobody publishes it,
so we measured it.
27%
suppressed · upper endoscopy with biopsy
40%
suppressed · knee replacement
78%
suppressed · ICD insertion
100%
suppressed · vaginal delivery
So we refuse to score some procedures at all. In obstetrics the check simply cannot be run, and the
tool says so rather than returning a confident-looking zero. Where suppression passes 90%, the percentile is
withheld instead of printed. Every procedure comes back labelled confirms, question only,
or not usable — because a percentile computed against a truncated distribution is exactly the kind
of number that gets a lawyer embarrassed on the record.
Experts: run it on yourself first
Everything above is public or discoverable. Opposing counsel can assemble most of it. The only
question is whether you see it before they use it.
Stated too strongly
The sentences that hand them the case
“There is only one appropriate approach” invites a single counter-example to collapse the
whole opinion. We quote the line from your own report, name the question you will face, and offer the
same opinion phrased so it survives.
Cannot defend
Opinions to narrow or withdraw
Where an opinion cannot be supported as written, the tool says so and tells you to narrow or withdraw
it before the deposition. It will not coach you to defend an opinion you should not be offering, and it
will not help you evade a fair question.
One thing this cannot do, and neither can anyone else
It cannot go find the expert's prior testimony on its own.
This is a rules problem, not an engineering problem. FRCP 26(a)(2)(B)(v) requires an expert's report
to list every case in which they testified in the previous four years — but FRCP 5(d)(1)(A) forbids filing
that disclosure until it is used in the proceeding, and California CCP § 2034.260 works the same way. The
list is served on opposing counsel and never docketed, so it exists in no public database. We find
prior-testimony contradictions in transcripts you upload, and we say plainly that we cannot find
the transcripts themselves. Treat any vendor promising automatic discovery of an expert's full testimony
history with real skepticism.
How it runs
Upload what you haveExpert report, CV, prior transcripts, medical records. All optional —
with none of them you still get a full 702 framework built from your case theory, and the brief tells you
which sections sharpen with which document.
We extract with true page anchorsEvery page keeps its real number, and numbered transcripts
keep printed page and line. A pin must be copied from a marker beside the quoted text; an item that cannot
cite a source is dropped rather than invented.
We check the claims against PubMedThe expert's factual assertions are pulled out, queried,
and the literature section is written only from records that came back.
We check qualification against MedicareNPI plus the procedure at issue, returning volume,
peer distribution, years since medical school, and a reliability label on every figure.
You get a documentContradictions, 702 attacks with scripts, literature, and cross-exam
questions organized by duty, breach, causation and damages. Export as PDF.
Build the map before you take the deposition
Free to try. Bring one report and one set of records, and see what comes back.
For licensed counsel and retained experts — not legal advice. The Attack Map is work product for a
professional to review, verify and exercise independent judgment on. Verify every pin cite against the
source document before use. Medicare utilization figures are fee-for-service only and exclude Medicare
Advantage, commercial insurance, Medicaid, self-pay, VA and military care; they describe billing, not
competence, and support a question rather than a conclusion.