Methodology
This page documents how AI Trace operates: our editorial process, evidence standards, data model, and commitments.
Editorial Process
Entries move through a four-stage lifecycle: intake, review, publication, and ongoing verification. Intake has two channels, and both feed the same human review.
Intake
Anyone can submit a report via the Report a Sighting form, no account required. Alongside community submissions, our automated evidence discovery pipeline scans public sources and drafts candidate entries with citations. Both channels land in the same moderator review queue; neither can publish anything on its own.
Review
A moderator reviews each candidate. They verify the claim against the cited sources, assess each source against the tier taxonomy below, accept or reject each source individually, and check for duplicates.
Publication
Accepted entries become structured records linked to the company profile, with a category, structured fields describing deployment and labor impact, and at least one accepted cited source.
Ongoing Verification
Entries are periodically re-audited: an automated pass re-checks published practices against current sources and proposes corrections, which go through the same moderator review. If a company reverses a practice or contradicting information emerges, entries are updated with full edit history preserved.
Every change to every entry is recorded in the public edit history visible on each company profile page. This append-only log is the primary mechanism against astroturfing.
Evidence Standards
Every claim must be backed by at least one cited source, and every source is classified into a three-tier taxonomy. The tier determines what a source can do on its own.
Tier 1: Primary
Company Disclosure, Regulatory FilingMaterial from the company itself or filed with a government: official statements, press releases, engineering blogs and documentation on official domains, SEC and regulatory filings. A single Tier 1 source can substantiate a claim.
Tier 2: Reputable Press and Research
News Article, Academic PaperReporting from established outlets with editorial standards and named authorship, and peer-reviewed academic research. Reliable, though it can lag or compress technical detail.
Tier 3: Secondary and Community
Social Media Post, Community ReportSocial media posts, community reports, and forums. Never sufficient alone: a Tier 3 source can open an entry as Reported, but the practice needs at least one Tier 1 or Tier 2 source to stand. Screenshots are required for social media sources.
Three sourcing rules apply. We distinguish which are enforced by code and which by editorial review, because the difference matters to anyone auditing us:
At least one accepted cited source per published practice.
Enforced in code: the automated publishing path refuses to create a practice whose sources were all rejected in review.
At least one accepted Tier 1 or Tier 2 source per practice.
Enforced in code: the automated publishing path refuses to publish a practice whose accepted sources are all Tier 3. Tier assignment is deterministic (government and regulatory domains and the company's own registered domain are Tier 1; a maintained allowlist of press with editorial standards is Tier 2; everything else is Tier 3), and a reviewer can override an individual source's tier only with a recorded reason.
Claims that AI automates a prior human task carry a stricter, two-sided standard: one side of the evidence must document the prior human process at the specific company, the other must confirm AI now performs that task.
Enforced in code at every write path, backed by a database constraint. Where only one side is confirmed, the claim is classified as unclear rather than published as automation.
We reject claims based solely on speculation, rumors, or sources that cannot be independently verified. When evidence is disputed, the disputed status is shown on the entry. Where sources do not support a specific value, fields are honestly marked unknown rather than filled with plausible guesses.
Neutrality Commitment
We do not tell you how to feel about AI. We document what companies do and let you draw your own conclusions.
The UI does not use red/green good/bad framing. Status badges are factual, not emotional. “Confirmed AI Use” is descriptive, not a judgment. We track AI use; we do not rate it.
We do not accept payments from companies to influence their entries. Companies cannot submit corrections directly; they must go through the same community process as everyone else. All edits are logged and public.
Labor impact classifications describe real-world effects, not moral judgments. A practice classified as automating a prior human task is factual, not an accusation, and it carries the strictest evidence standard on the platform.
Data Model
Every AI practice entry is structured with the following attributes:
Dispute Process
If you believe an entry is inaccurate, you can file a dispute directly from the practice detail page. Disputes are reviewed by moderators and resolved transparently.
Dispute types are: wrong or fabricated information, source inaccuracy, miscategorization, outdated information, and biased framing. Every reviewed dispute's outcome, whether accepted or rejected, is published on the entry's page with a resolution summary written by the reviewer, and an accepted dispute's changes additionally appear in the public edit history.
The dispute submission itself is never republished. What appears publicly is our own text: the dispute category, the dates, the verdict, and the reviewer's summary. Submissions that are not genuine disputes, such as spam or abuse, are dismissed without a public record; the transparency promise applies to disputes reviewed on their merits.
We do not silently remove contested information. If an entry is disputed, it carries a “Disputed” badge until the dispute is resolved.
Corrections follow the same transparency rule. When an error is verified, the entry is corrected and the change is recorded in the public edit history. An entry that turns out to be materially wrong is marked Retracted rather than deleted: the record of what we got wrong stays public, because a transparency platform does not get to hide its own mistakes.
How to Contribute
The most valuable thing you can do is submit evidence. If you notice a company using AI, report it via the Report a Sighting form. It takes under 90 seconds. No account required.
You can also help by sharing news articles about corporate AI practices, linking to company profiles in relevant discussions, and filing disputes when you spot inaccuracies.
Our Use of AI
AI Trace tracks corporate AI use. We hold ourselves to the same standard. Our own AI practices are documented on our company profile at /company/trace-foundation.
Intelligent Submission Processing
ActiveWhen a community report arrives, a language model pre-screens it for the moderator: which company it likely refers to, whether it duplicates an existing entry, and which category fits. Suggestions appear as pre-filled, editable fields in the review interface. Moderators always have the final say.
Semantic Search
ActiveVector embeddings power semantic search alongside keyword matching, so a search for a concept finds relevant entries even when they use different terminology.
Automated Evidence Discovery (Scanner)
ActiveA multi-stage discovery pipeline built on commercial frontier language models searches public sources for evidence of corporate AI use and drafts candidate entries as structured records, with every claim citing the sources it came from. Candidates enter a moderator review queue where each source is individually accepted or rejected and the entry's text can be edited before publication. The guarantees on this path are enforced in code, not policy: a practice cannot be created if all of its sources were rejected, and the corroboration guard described under Evidence Standards applies to every write path.
Automated Re-Verification (Auditor)
ActiveA companion pipeline periodically re-examines published entries against current public sources and proposes field-level corrections, each with citations. Proposals pass through the same review queue, and accepted changes are recorded in the public edit history.
Continuous Company Monitoring (Watchers)
ActiveWatchers poll the newsrooms and SEC filings of tracked companies around the clock. When something new appears, it is screened for relevance and flagged to our reviewers, typically within minutes of publication. An alert is a lead, never evidence: it can prompt a scan or a re-audit, and everything passes through the same human review and sourcing standards before publication.
Every entry here starts with automated discovery and ends with human judgment. No human team can continuously track the public record of hundreds of companies, so our automated research pipeline, built on commercial AI models, does the finding and the first draft: it watches company newsrooms and regulatory filings for new developments, searches public sources, and drafts candidate entries with citations. Human reviewers decide what is accurate enough to publish, and every published entry records who reviewed it and when, field by field, on the entry itself. This disclosure does not depend on watermarking or AI-detection technology; we describe how our system works whether or not tools exist that could detect it.
Nothing AI-assisted is published without human moderator approval. The specific commercial models in use are documented on our company profile and updated as they change, so this page describes the architecture and its guarantees rather than any one vendor.
For information about the team and how to support our work, see the About page.