Expert Opinion Challenge
Independent review of an opposing software expert opinion: whether its sources exist, whether its methodology produces its conclusion, and whether AI was used to draft it.
Who this is for
- Counsel holding an opposing opinion that reads as AI-drafted. The prose is too smooth, the confidence too even, and the sources look fine until you try to open them.
- Counsel whose opposing opinion simply does not hold up technically. The conclusion sounds plausible, but the route to it is undocumented, not reproducible, or skips a step.
- Counsel preparing to cross-examine an expert. You need the weak points ranked by strength, and the questions already written.
The four layers
- Source audit. We check every citation, ruling, standard, paper and number in the opinion one by one: does the source exist, does it say what the opinion attributes to it, and is the cited version the relevant one. This is the layer that produces evidence. A finding that a source does not exist is a fact you can put in front of the judge, not an assessment.
- Internal technical consistency. We check whether the methodology the opinion describes actually produces the conclusion it reaches. We reproduce the analysis from the raw material and locate logical jumps, unstated assumptions, and numbers that do not follow from the data shown.
- Discovery. Where there is real reason to suspect AI use, there is a procedural route: demand the prompts, run logs, model version, and the chain of custody for inputs and outputs. In May 2026 a federal magistrate judge in the District of Connecticut, in Conservation Law Foundation v. Shell Oil, ordered the prompts an expert fed to OpenAI models to be produced to the opposing side. Reportedly this was the first time a US federal court has so ordered. We wrote about the case and what it opens up in a separate analysis. We define exactly what to demand on the technical side; the motion itself is the lawyer's work.
- Statistical indication. Text detectors and phrasing signatures have one job here: they tell us where to dig first in layers 1 and 2. A detector score does not enter a submitted opinion and is never presented as evidence. Text detectors are known to be unreliable, and in our assessment an opinion resting on such a score is badly exposed under cross-examination.
Two packages
Both are independent. Either can be commissioned directly, and there is no requirement to start with the memo.
| Findings memo | Counter-opinion | |
|---|---|---|
| Filed with the court | No | Yes |
| Exposure | Preparation work, stays with counsel | Disclosed, including cross-examination of the expert |
| Deliverable | Ranked findings plus ready cross-examination questions | Written opinion, technical appendix, independence statement |
| Timing | Initial response 48 hours; delivery scoped to the volume of material | Two weeks from receiving the complete material |
What the service does not do
- It does not give legal advice. What to do with the findings, how to work them into the argument, and when to raise them are counsel decisions.
- It does not declare a text AI-written based on a detector score. If there is no testable finding in the sources or the methodology, the answer is that there is no finding. We will not sell you a suspicion.
- It does not draft the discovery motion. We define the technical list of what to demand; the drafting and the filing stay with you.
Pricing
Scoped to the volume of material and the software question, stated in the engagement letter after a no-cost intro call. If what you need is a primary opinion rather than an attack on an existing one, the right track is expert witness.
Frequently asked
Can you prove an opinion was written with AI?
Not on a text-detector score: a detector score is not evidence, and it is not what this service rests on. What is provable: that a cited source does not exist, that it does not say what the opinion claims, or that the stated methodology does not produce the stated conclusion. Those findings are testable and reproducible, and those are what enter the opinion.
What is the difference between a findings memo and a counter-opinion?
A findings memo is preparation work that is never filed: a list of findings plus ready cross-examination questions. A counter-opinion is filed, carries a technical appendix and an independence statement, and exposes the expert to cross-examination. The two are independent, and either can be commissioned directly.
How long does it take?
Initial response within 48 hours. The findings-memo delivery date is scoped to the volume of material and stated in the engagement letter. A counter-opinion is delivered within two weeks of receiving the complete material.
Does the service include the discovery motion?
Only the technical half of it. We define exactly what to demand: the prompts, run logs, model version, and the chain of custody for inputs and outputs. Drafting and filing the motion is the lawyer's work.