How it works
From a court opinion to a sourced expert dossier, with every step visible and every fact traceable to where it came from.
The workflow
You stay in control of what gets relied on.
Search by name, field, institution or court across every expert in the database. Filters narrow by discipline, by whether a court has ruled on their testimony, and by exposure. Nothing is behind an account, and searches aren’t logged.
Each expert lists every opinion that named them, with the court’s ruling on their testimony: excluded, limited, admitted, or challenged. Where an opinion named the expert but never ruled, the dossier says so instead of guessing.
Name any expert from a brief or disclosure. Uncommon Context pulls their prior rulings, publications, industry payments, and past expert work, and, if you paste in the report’s opinions, compares each one to what the expert has said before. It takes about two to four minutes.
Every finding gets an exhibit sticker (PX for plaintiffs, DX for defense) only if its source page was actually retrieved. Anything that can’t be tied to a source is marked “Unconfirmed” and hidden unless you choose to see it.
Uncommon Context drafts short, leading questions grouped by topic, each tied to the exhibits that support it. If the expert is yours, it drafts the attacks you should prepare for instead.
“Check for new rulings” searches for orders, Rule 702 rulings, and expert disclosures naming an expert you’re tracking since your last check. Any expert named in a new filing can be sent straight to a dossier.
Identity matching
Court opinions say “Dr. Smith.” Journals say “J. A. Smith.” Payment records say “SMITH, JOHN A.” Joining those correctly is the hardest part of expert research, and the part most tools don’t show you.
A shared registry number — a national provider ID, a researcher ID or a payments profile ID — links records immediately. Two different registry numbers mean two different people, always.
A rare name needs less corroboration than “John Smith.” Uncommon Context estimates how many real people share a name, using the national physician registry and its own records, before deciding.
A name match alone never links records automatically. At least two independent facts must agree, such as the same institution and specialty, or the same state and field.
Near-miss first names (Jane versus Janet), conflicting middle initials or suffixes, and cases where two people are about equally likely all go to a human reviewer instead of being guessed.
A name in a court opinion can attach to a person but never adds facts to their profile, so one weak match can’t make the next weak match look stronger.
Each connection records its evidence, its probability, and whether a machine or a reviewer made it. Human decisions are permanent and train the model, and random samples are re-checked to measure accuracy.
Exposure scoring
The exposure rating summarizes what’s confirmed on the public record. It’s a triage signal, not a prediction of how a judge will rule.
| Confirmed finding | Points | Why it matters |
|---|---|---|
| Prior ruling excluding the expert | 3 | A court has already found the testimony unreliable. |
| Professional discipline record | 3 | Goes directly to credibility and qualifications. |
| Prior ruling limiting the testimony | 2 | Part of the opinion was cut back before. |
| Direct conflict with a prior statement | 2 | The expert said the opposite elsewhere. |
| Financial tie to a named defendant | 2 | Bias evidence tied to this case. |
| Partial conflict with a prior statement | 1 | Tension worth probing, not a clean contradiction. |
| Other industry tie | 1 | General bias evidence. |
Under 3 points.
3 to 6 points.
7 or more points.
Only confirmed findings count. Unconfirmed items and roster entries without completed research show “Exposure not assessed yet.”
Limits
Get started
Use a capability on its own, or connect all five so every answer starts from what the others found. Your team works in the site; your AI agents reach the same capabilities over MCP.