Bartolus Appellate intelligence

Oral argument intelligence

Prep reports built on
behavioral data.

From the founder

Oral argument preparation has not changed in fifty years. Attorneys read cases, run moots, and hope for the best. The problem is that judges are not abstractions — they are specific people with patterns, habits, and tells. Those patterns are in the transcripts.

Bartolus reads every transcript, tags every question by purpose, technique, and tone, and turns that record into a prep report. Not a chatbot answer. Not a dashboard. A document you can mark up at your desk the night before argument.

Methodology 3D Purpose · technique · tone
Available for All 13 Federal circuit courts + SCOTUS
Both briefs read 400k Characters extracted, no truncation
§ 01

The case for Bartolus

Why Bartolus

Generic AI predicts what judges ask about. Bartolus predicts how they ask it — and can show you when it was wrong.

Same briefs. Different preparation.

ChatGPT reads doctrine.
Bartolus reads behavior.

We fed the same briefs to both. Here is what each predicted Judge Higginson would ask in FBCC CityPoint v. City of Austin — and what he actually said.

ChatGPT "What is your limiting principle for 'same transaction' under res judicata?"
Bartolus "Compare the two complaints. What overlaps?"
Higginson "In your pleadings, in the two lawsuits, it's a copy paste, practically."
ChatGPT predicted the doctrinal issue. Bartolus predicted the interrogation technique. Higginson used the technique. · FBCC CityPoint, L.P. v. City of Austin
Built to fail honestly.

Every roadblock we hit
is in the report.

We don't know what other tools show clients when they're wrong. We know what we show ours.

01

Wrong metric. Early versions graded on how often our predictions fired — a passing grade even when we missed questions judges actually asked. We rebuilt the backtest to measure the inverse: how many of the judge's actual questions did we cover?

02

Missed the record. A judge questioned an attorney about who represented them at a critical hearing. Neither brief named that person. We now run a Record Vulnerability Audit on every set of briefs — flagging facts and third parties that could become hot-button questions regardless of what either side argued.

03

Thin data. For judges with fewer than five cases in the corpus, every report says so explicitly. Directional, not predictive. You deserve to know when we are guessing.

On Harvey

Neal Katyal's demonstration showed what a well-resourced custom build can do for one argument. Bartolus is the productized version — available to any appellate attorney, not only those with enterprise AI contracts.

Built by a practitioner.

Not a chatbot. Not a dashboard.
A prep report.

The methodology was designed by a working appellate attorney and law professor who has read every transcript in the corpus and quality-checked every tag in the taxonomy.

  • Reads both briefs in full — up to 400 pages each, without silent truncation
  • Three-dimensional tagging: purpose, technique, and tone as separate signals
  • Compounding corpus — every new case sharpens every future report
  • Backtested against actual argument transcripts with falsifiable accuracy claims
  • Delivered as a Word document you can mark up, not a platform to learn
§ 02

Free to explore

OT2025 SCOTUS Analysis

An interactive behavioral profile of all nine justices across the current term's oral arguments. Explore the methodology before committing to a report.

Oct. Term 2025 · All nine justices · 58 cases

The SCOTUS Dashboard

Purpose, technique, and tone profiled across every argument of the current term. Seniority-sorted. Filter by subject matter. Built on official transcripts only.

Open the dashboard →
§ 03

The team

About

Bartolus is a product of Greil Analytics LLC, an Austin, Texas company.

John Greil

Founder · Greil Analytics LLC · Bartolus

John Greil is a law professor and practicing appellate attorney. He has argued before the Fifth Circuit, clerked at the federal appellate level, and holds a J.D. from Harvard Law School. He built Bartolus because he wanted a prep tool that would survive cross-examination by the attorneys who use it — and found that nothing on the market did.

Every transcript in the corpus has been read by a human. Every category in the taxonomy has been tested against real argument. The methodology is transparent by design: if a judge has thin data, the report says so. If a prediction fired or missed in backtesting, the record is in the file.

Harvard Law J.D. Federal appellate clerk Law professor Fifth Circuit practitioner Austin, TX
§ 04

Get started

Request a Prep Report

Available for arguments in any federal circuit and the Supreme Court. Turnaround is typically 48–72 hours once briefs are received.

If you have an argument coming up and want to know whether we have data on your judges, reach out. If we don't have corpus coverage yet, we'll tell you honestly rather than generate a report we can't stand behind.

For firms interested in ongoing access, we are building a subscription product. Mention it in your message.