Decision Desk · Issue 06 · complete public learning edition

Can You Prove That AI-Written Claim?

Decide whether a defined public-claim class has evidence, authority, approval, provenance, version control, and a correction path strong enough to publish.

Begin the issue Run the Substantiation Gate

Learning time90 minutes

Teaching12 cumulative lessons

InstrumentSubstantiation Gate

DefaultNo evidence · no publication

Claim-control chain

Claim → evidence → inference → authority → approval → correction.

Polished generated copy cannot skip any link.

Freeze express and implied meaning

Gate

Locate competent supporting evidence

Gate

Bound the inference and qualifications

Gate

Verify permission and responsible owner

Gate

Approve the exact channel version

Gate

Monitor, correct, and withdraw

Gate

Complete issue curriculum

Twelve cumulative lessons.

Use the fictional launch campaign only. Never paste unreleased or confidential campaign copy.

Issue 06

Freeze the exact claim

Open

A topic or draft theme cannot be substantiated. Record the precise words, audience, channel, placement, date, implied message, and material qualification.

Issue 06

Classify what the audience will understand

Open

Distinguish subjective expression from objective, comparative, performance, scientific, sustainability, endorsement, testimonial, and customer-outcome claims.

Issue 06

Assign the evidence owner and approver

Open

The writer, model, evidence owner, subject-matter reviewer, legal/regulatory reviewer, and publication authority may be different people. Name them.

Issue 06

Separate source fact from inference

Open

A source may establish a measured result without supporting a broader causal, universal, future, comparative, or customer-specific conclusion.

Issue 06

Match evidence to claim conditions

Open

Population, product version, dose, setting, methodology, timeframe, sample, comparator, uncertainty, and limitations must travel with the conclusion.

Issue 06

Govern endorsements and testimonials

Open

Verify identity, experience, permission, material connection, representativeness, scope, currentness, and whether generated language changes what the person authorized.

Issue 06

Test comparisons

Open

Name the comparator, metric, conditions, source, timeframe, and meaningful difference. A vague ‘better’ claim can still communicate an objective comparison.

Issue 06

Protect intellectual property and confidentiality

Open

Do not treat generated wording as proof of originality or permission. Check source use, third-party material, trade secrets, and confidential evidence separately.

Issue 06

Record generation and approval provenance

Open

Preserve source, prompt/input boundary, model or tool version where material, human edits, claim version, evidence version, approver, date, and channel.

Issue 06

Control channel and version drift

Open

A claim approved for one page, audience, qualification, or product version may become misleading when shortened, reposted, translated, or detached from context.

Issue 06

Design correction and withdrawal

Open

Name monitoring, challenge intake, correction owner, affected channels, withdrawal authority, notice, archive treatment, and retesting after evidence or product change.

Issue 06

Issue a claim-specific disposition

Open

Do not publish, narrow/qualify, gather evidence, or approve only the governed fictional version. The tool never clears real copy for publication.

Critical stop conditions

Do not publish through these gaps.

the exact express and implied claim is not frozen
competent evidence is missing or materially mismatched
a source fact is stretched into an unsupported inference
an endorsement, comparison, or third-party use lacks authority
the approved qualification disappears in the publication channel
claim, evidence, product, or approval versions cannot be reconstructed
there is no owner who can correct and withdraw the claim

Interactive instrument · device-local

Public-Claim Substantiation Gate

Evaluate fictional launch claims only. Do not enter unreleased copy, confidential evidence, customer data, trade secrets, regulatory correspondence, or active legal matters.

0/30claim gates evaluated

Generated language is not substantiation.
No adequate evidence, authority, approval, or correction path means do not publish.

Fictional claim class	Claim classified	Evidence competent	Source-to-claim fit	Use authorized	Approval recorded	Correction ready
Performance claim	Performance claim: Claim classified	Performance claim: Evidence competent	Performance claim: Source-to-claim fit	Performance claim: Use authorized	Performance claim: Approval recorded	Performance claim: Correction ready
Sustainability claim	Sustainability claim: Claim classified	Sustainability claim: Evidence competent	Sustainability claim: Source-to-claim fit	Sustainability claim: Use authorized	Sustainability claim: Approval recorded	Sustainability claim: Correction ready
Customer outcome	Customer outcome: Claim classified	Customer outcome: Evidence competent	Customer outcome: Source-to-claim fit	Customer outcome: Use authorized	Customer outcome: Approval recorded	Customer outcome: Correction ready
Expert endorsement	Expert endorsement: Claim classified	Expert endorsement: Evidence competent	Expert endorsement: Source-to-claim fit	Expert endorsement: Use authorized	Expert endorsement: Approval recorded	Expert endorsement: Correction ready
Comparative claim	Comparative claim: Claim classified	Comparative claim: Evidence competent	Comparative claim: Source-to-claim fit	Comparative claim: Use authorized	Comparative claim: Approval recorded	Comparative claim: Correction ready

Exact claim, audience, and channelWrite a fictional exact claim. Who will receive it, where will it appear, and what would a reasonable audience understand it to promise?

Evidence and inference boundaryIdentify the source, date, population, conditions, limitations, and the precise inference it can support. What cannot be said?

Approval, version, and provenanceWho owns the evidence and approval? Which claim and source versions were reviewed? Record human versus generated contributions.

Monitoring, correction, and withdrawalWho monitors continued accuracy? How is a claim corrected across channels, withdrawn, and preserved in the change record?

Bounded dispositionDo not publishNarrow or qualify the claimGather evidence and re-reviewApprove only the fictional governed version

Restoring your local gate…

Completion boundary

The educational product and reusable Substantiation Gate are complete.

Real publication still requires current claim- and sector-specific authority, competent evidence, qualified review where needed, exact-channel approval, rights and confidentiality review, monitoring, and organizational authorization.

Return to the 12-month program Previous issue