Research preview — all forecasts and metrics shown are illustrative
SCOTUS Seer

Computational legal forecasting

A clearer view of
what the Court may do.

Nine justice-modeled AI agents analyze briefs, precedent, oral arguments, and judicial history to produce transparent, probability-based forecasts of Supreme Court rulings.

Explore forecasts How it works

Independent research project. Not affiliated with the Supreme Court of the United States.

Featured forecast

The question before the model

Sample · updated Jun 12, 2026
No. 25–418Fourth Amendment

United States v. Mercer

Whether long-term collection of location data from connected vehicles constitutes a search under the Fourth Amendment.

View forecast rationale

Predicted disposition

Affirm68%
Affirm 68%Reverse 32%
Model confidenceModerate

Projected vote

6–3to affirm
JRCTSANGBKKB SSEKAC

Initials are fictional placeholders in this research preview—not representations of real justices.

On the horizon

Upcoming cases

Administrative lawHigh confidence

Calder Energy v. EPA

Agency authority to regulate emerging industrial emissions under an older statutory framework.

Likely outcomeAgency prevails
77%
First AmendmentModerate

Dawson v. North Elmore

Whether a public-school social media policy impermissibly burdens off-campus student speech.

Likely outcomePetitioner prevails
61%
Federal courtsLow confidence

Wren County v. Patel

Standing and redressability where a local policy has been suspended but not formally repealed.

Likely outcomeDismissed
54%

Case names, facts, vote projections, and probabilities on this preview are fictional and presented solely to demonstrate the product.

Methodology

Nine perspectives.
One calibrated forecast.

SCOTUS Seer is designed to make uncertainty visible—not to dress a guess as a verdict. Each modeled agent produces an independent analysis before a synthesis layer tests the result.

  1. 01

    Build the record

    Briefs, opinions, oral-argument transcripts, and relevant doctrine are retrieved with source provenance.

  2. 02

    Model each perspective

    Nine AI agents, informed by publicly documented jurisprudential patterns, reason independently.

  3. 03

    Stress-test the vote

    Repeated simulations vary uncertain facts and legal assumptions to expose fragile conclusions.

  4. 04

    Publish probabilities

    A forecast, rationale, vote distribution, and uncertainty notes are released before the decision.

Track record

Accountability starts before the opinion.

Once live forecasting begins, every prediction will be timestamped and preserved. Results will include both simple accuracy and calibration—whether 70% forecasts actually happen about 70% of the time.

Public scorecard coming after the first completed term
—Cases scored
—Outcome accuracy
—Brier score

Metrics intentionally blank until real, preregistered forecasts can be evaluated.

Research commitments

Built for scrutiny

◇

Transparent uncertainty

Probabilities and confidence ranges accompany every forecast.

◎

Auditable sources

Legal claims connect back to public primary materials.

⌁

No synthetic authority

Modeled agents are research tools. They are not the justices and do not speak for them.