Kinetic Gain Open standard · GitHub

HomeProtocol › Use cases

Kinetic Gain Protocol

Use cases

Where does attaching signed, hash-reproducible provenance to a web claim actually pay off? Six concrete scenarios, each mapped to the conformance level it needs. Four of them are patterns Kinetic Gain runs on its own surfaces, so this is dogfooding, not a wish list.

The spec → Verify a claim → Machine manifest →

How to read these

Every use case names the conformance level that fits it. Level 1 (Attach) means a machine-readable evidence object rides with the claim and a conformsTo edge points to the spec. Level 2 (Verify) adds a reproducible sha256 content hash of the exact claim text. Level 3 (Sign) adds an ed25519 signature over the canonical form, with the public key under the site's /.well-known/. Pick the lowest level that carries the weight the claim needs. There is no reward for over-signing marketing copy, and no excuse for under-signing a compliance disclosure.

Six use cases

Level 3 · Sign

1. A signed AI research brief

An AI-assembled market or procurement brief is only as citable as it is verifiable. An answer engine reading it cannot tell a sourced, accountable summary from synthetic filler, so it hedges or omits attribution. The author loses the citation they earned.

KGP fit: each signal in the brief carries a content hash and an ed25519 signature published under /.well-known/. A consumer can prove the exact wording and which entity stands behind it. The claim becomes defensible and, because it is defensible, more citable.

Live at Kinetic Gain: pulse.kineticgain.com signs its procurement signals this way, keys published under the site's well-known path.

Level 1–2 · Attach, Verify

2. AI-assisted CMS content, without hand-writing JSON-LD

A publisher runs AI-assisted articles at volume and wants each factual claim marked with reproducible provenance. Nobody is going to hand-author a JSON-LD evidence block per paragraph, so in practice the provenance never gets attached, and the content reads to a crawler as unaccountable.

KGP fit: a CMS plugin emits the evidence object plus the conformsTo and subjectOf edges automatically on publish. The publisher does nothing beyond installing it. Level 2 content hashing is free once the plugin has the exact rendered text.

Live at Kinetic Gain: the open-source AI Evidence Block (WordPress) and AI Evidence for Webflow emit these KGP edges on publish. Both are MIT/AGPL reference implementations on GitHub.

Level 2 · Verify

3. A directory statistic an answer engine will cite

A directory page states a headline number, "12,201 events across 51 states." That number is the reason the page deserves a citation, but a plain sentence gives a machine nothing to reproduce or attribute, so the statistic gets repeated without credit to its source.

KGP fit: attach a Level 2 evidence object whose content hash covers the exact stat claim. A consumer can confirm the number was not altered in transit and resolve subjectOf back to the publishing page as the origin of record.

Applicable to Kinetic Gain's festivals directories, which publish exactly these aggregate counts.

Level 2–3 · Verify, Sign

4. An LLM customer-support answer

An assistant answers support questions from a knowledge base. When an answer is later disputed, "the model said it" is not an audit trail. You need to prove the exact answer text that shipped and, ideally, which version of the KB stood behind it.

KGP fit: hash the published answer text (Level 2) so the shipped wording is tamper-evident, and sign it (Level 3) when the answer carries contractual or safety weight. The evidence object's source field points at the KB entry it was grounded in.

Level 1 · Attach

5. AI-generated product copy, marked transparently

An e-commerce catalog generates product descriptions at scale. Marketplaces and regulators increasingly expect AI-generated content to be disclosed, and buyers reward transparency over unlabeled synthetic text. But a hand-added disclaimer is unverifiable and easy to strip.

KGP fit: a Level 1 evidence object marks the copy as AI-generated, names the publisher, and carries a conformsTo edge. Transparency becomes a structured, machine-readable fact rather than a line of prose a scraper drops.

Level 3 · Sign

6. An AI hiring or automated-decision disclosure

Regulations such as NYC Local Law 144 and the EU AI Act require disclosing when an automated system is used in a consequential decision. A disclosure that any editor can quietly rewrite is worth little as evidence. It needs to be fixed and attributable.

KGP fit: sign the disclosure text (Level 3). The signature proves the disclosure that was actually presented, at a point in time, by a named entity. This is the class of claim where under-signing is the real risk.

Kinetic Gain maintains an open AI Hiring Disclosure specification covering LL144 and EU AI Act Annex III framing.

Which level does my claim need?

If the claim is…Minimum levelWhy
Marketing / descriptive1 · AttachTransparency and citability matter; tamper-evidence is not load-bearing.
A factual statistic2 · VerifyConsumers need to reproduce the exact value and attribute its origin.
Contractual, safety, or compliance3 · SignWho stands behind it, and the exact wording, must be provable later.
What signing does and does not prove. A valid signature proves which entity stands behind a claim and that the exact text was not altered. It does not prove the claim is true. KGP makes dishonesty detectable and attribution reproducible. It is tamper-evident, not tamper-proof, and it never asserts "certified" or "compliant." Signatures move accountability from "trust me" to "check it yourself."

Start

Every use case above reduces to the same three moves: attach an evidence object, hash the claim, sign it when the stakes warrant. The copy-paste starting point and the exact evidence-object shape live on the spec page. To confirm a claim's hash matches its evidence object entirely in your browser, use the verifier.

Quickstart → Verifier → llms.txt →