Find the original
Trace a quotation to the full speech, a vote to its exact motion, and a policy promise to the operative text.
THE POLITICAL EVIDENCE DESK
Go beyond the claim.
Follow the record.
Examine claims against original records, with context and limitations.
An AI-powered evidence-auditing engine in development for newsrooms, journalists, researchers, and documentary producers.
Try the narrow original-record retrieval pilot or inspect the prepared example. General AI auditing is not connected.
EVERY LINK SHOULD BE EXAMINABLE
- 01Claim
- 02Original source
- 03Statement
- 04Bill
- 05Amendment
- 06Vote
- 07Funding
- 08Implementation
- 09Measurable outcome
- 10Counterevidence
- 11Confidence assessment
BUILT AROUND THE REPORTING QUESTION
A claim is the beginning.
The evidence is the work.
Clips lose context. Votes get detached from amendments. Spending promises get confused with money spent. EvidenceDesk is designed to connect those records while keeping the gaps visible.
Challenge the finding
Surface counterevidence and alternative explanations with the same care given to supporting material.
Hand over a source packet
Give a researcher or editor the citations, limitations, context, and unanswered questions needed for an independent decision.
NEWSROOM MODE
Your political evidence desk.
Your editorial judgment.
Designed for political desks, investigative reporting, research teams, documentary production, and political media creators. Broader voter tools are a later phase.
Request Newsroom Beta Access →- 1Paste the claim
- 2Retrieve evidence and full context
- 3Inspect counterevidence
- 4Trace policy and measurable outcomes
- 5Generate the source packet
- 6Editor reviews, then decides whether to publish
THE PROPOSED RESEARCH SYSTEM
Specialized checks.
One accountable review.
These are planned research roles, not active agents in this preview. Each finding stays connected to evidence. Deterministic release rules hold high-risk packets for editor approval.
Research Agent
Locate original records and document missing material.
Legislation Agent
Track text, amendments, motions, and roll calls.
Quote Verification Agent
Retain full passages, dates, exclusions, and disclaimers.
Policy Implementation Agent
Separate authorization, appropriation, obligation, and spending.
Outcome Agent
Inspect baselines, denominators, timeframes, and data quality.
Causality Agent
Evaluate attribution and plausible alternative explanations.
Counterevidence Agent
Seek the strongest material that challenges a conclusion.
Bias / Adversarial Review Agent
Apply the same test regardless of party or subject.
Citation Validation Agent
Check source identity, passages, versions, and link integrity.
Human Editor CheckpointReview held Tier C packets and sampled releases; authorize publication separately.
METHODOLOGY BEFORE VERDICTS
One standard.
Every political subject.
Republican, Democratic, independent, national, state, or local: the evidence standard stays the same. Neutrality means consistent scrutiny, not equal conclusions.
Read the methodology →Outcomes need more than a timeline.
A change after a policy is not proof that the policy caused it. The proposed workflow checks comparison groups, economic conditions, implementation, measurement changes, and competing causes.
Show what remains unknown.
Missing records are recorded as gaps. A lack of evidence does not automatically establish that a claim is false.
FOR MEDIA ORGANIZATIONS
Bring a real reporting problem.
Request Newsroom Beta Access to discuss your evidence workflow, editorial requirements, and source-packet needs.
This is an inquiry, not instant access or a subscription. Do not include confidential sources or unpublished sensitive reporting.