Artificial Intelligence • Supreme Court Prediction

The most important thing we do is not doing.

Justice Louis D. Brandeis • on the denial of certiorari

Upload a lower-court opinion and receive an AI-powered assessment of its certiorari-worthiness — calibrated against a corpus of over 1,700 historical SCOTUS petitions.

Analyze a Lower-Court Opinion
scroll

About

Predicting the Court’s Docket

Each year the Supreme Court receives more than 7,000 petitions for certiorari and grants fewer than 80 — roughly one percent. Identifying which petitions stand a genuine chance requires expertise in reading circuit splits, vehicle quality, and the legal landscape the Court is willing to enter.

Cert Predictor applies a two-stage pipeline: a semantic embedding model finds the most similar historical petitions in our corpus, and Claude — Anthropic’s large language model — reads the full lower-court opinion to assess legal question importance and vehicle quality independently before combining the signals into a grant probability.

The system is trained on paid-docket petitions from OT2015 through OT2024 and validated blind on OT2025 cases, where it achieves an F1 score of 0.83 at a 20% probability threshold.

1,780+
petitions in corpus
10
terms of training data
0.83
F1 score (OT2025)
The Supreme Court of the United States

Methodology

How It Works

01

Semantic Embedding

Your opinion is converted into a 384-dimensional vector using a sentence-transformer model, then compared against every petition in our historical corpus to identify the twenty closest analogues by legal content.

02

Corpus Baseline

A weighted baseline grant score is computed from the similar cases, boosted by cert outcome and petition type (full grant, argued case, or GVR), giving a prior probability rooted in historical precedent.

03

AI Analysis

Claude reads the full opinion text — preserving dissents in their entirety — and evaluates legal question importance and vehicle quality independently before combining all signals into a final grant probability and predicted outcome.

Petition Analysis

Analyze a Lower-Court Opinion

Upload the lower-court opinion from the petition appendix as a PDF. Text-searchable PDFs work best; scanned image PDFs will not extract correctly.

Drop a PDF here, or click to browse

Lower-court opinion from the petition appendix

Extracting text and embedding the opinion…
Consulting Claude on cert-worthiness…
This usually takes 15–30 seconds.

Predicted Outcome

Grant Probability
—
grant probability
Baseline Score
—
corpus-weighted avg (0–4)
Predicted Score
—

Reasoning

Most Similar Historical Petitions

Docket Case Name Term Lower Court Outcome Similarity

Petition Analysis

Analyze a Petition

Upload the petition for a writ of certiorari (and optionally a brief in opposition) to receive an AI-powered assessment calibrated against a corpus of petitions from OT2019–OT2024.

Petition for Certiorari

Required — drop PDF here or click to browse

Brief in Opposition

Optional — drop PDF here or click to browse