Use Case · Verified Prediction Archive
Verified Prediction Archive
Publish a tamper-evident archive of AI-generated forecasts that any reader can verify. Pre-outcome timestamps, signed receipts, and a daily manifest that proves the published set is complete.
Credibility for AI forecasts comes from one thing: the record cannot be revised after the outcome. A verified archive makes that property cryptographic, not editorial.
Why forecast credibility breaks
Publishing predictions in a CMS is the default — and the reason most forecast archives cannot be trusted by an outside reader.
Forecast pages can be edited
A platform that publishes predictions in a CMS can edit, delete, or backdate them after the outcome. Readers cannot prove what was claimed in advance.
Selective disclosure looks complete
Publishing only the winners makes an archive look favorable without making it false. Without a daily manifest, no outside party can verify the published archive is the complete set.
Screenshot evidence is weak
Screenshots prove a prediction was displayed at a time. They do not prove the timestamp is real or that the prediction was not modified between display and dispute.
Trust is asymmetric without proof
Readers cannot easily distinguish forecasters who never revise from those who silently edit. A verified archive flips the asymmetry — readers can prove the record themselves.
How a verified archive works
Your platform produces an AI-generated prediction with the inputs, model version, and timestamp captured at generation time.
Construct the prediction_receipt.v1 JSON payload with source, prediction_id, model_version, predicted_at, and any context summary.
POST the canonicalized payload to the issuance endpoint. Receive a certificate_id and Ed25519 signature. Store both in your database alongside the prediction.
Render a 'Verify this prediction' link on each forecast page pointing to /verify/<certificate_id>. Readers can verify without leaving your platform — or independently if they prefer.
At the end of each day (or whatever period your archive uses), gather every prediction hash and submit a manifest certificate. Display the manifest ID prominently — this is what proves the archive is complete.
Who builds a verified archive
Independent forecasters who publish weekly or daily picks. A verifiable archive proves the predictions were not revised after the outcome.
Consumer-facing apps that publish AI-generated forecasts. A signed archive differentiates the product from competitors that can silently edit results.
Analytics products that publish predictive models for outcomes. Framing is analytics-first, not gambling — verification adds credibility with media partners and institutional users.
Research teams publishing public forecasts. A verified archive supports replicability and citation in academic and institutional work.
Individuals publishing forecasts on personal sites or social platforms. A verified archive elevates the published record above unverified claims.
Research projects evaluating AI trading agents. Verified prediction archives are the canonical data source for benchmark replicability.
Reference implementation
BetFire AI is the published reference integration for verified prediction archives in a high-trust-deficit category. The integration pattern documents the canonical JSON snapshot, SHA-256 fingerprint, signed receipt, certificate ID storage, public verification page, and daily manifest layer.
The same pattern applies to any platform publishing AI-generated predictions: the storage layer changes, the schema fields differ, but the integrity model is identical.
Machine-readable summary
{
"concept": "Verified prediction archive",
"concept_type": "use-case",
"canonical_url": "https://certifieddata.io/use-cases/verified-prediction-archive",
"parent_concept": "Prediction Integrity",
"related_concepts": [
"Certified predictions",
"Daily prediction manifest",
"BetFire AI reference integration"
],
"target_audience": [
"forecasting newsletters",
"AI prediction apps",
"sports analytics platforms",
"financial research labs",
"public forecasters",
"trading-agent benchmark projects"
],
"pattern_layers": [
"per-prediction certificate",
"per-period manifest",
"public verification link",
"model artifact certificate reference"
],
"signing_algorithm": "Ed25519",
"hash_algorithm": "SHA-256 (RFC 8785 canonicalized)",
"positioning": "CertifiedData proves the prediction record. DecisionLedger proves the prediction process."
}Build a verified prediction archive
The certification flow is the same one CertifiedData uses for every artifact type. No new infrastructure required.