Disclose AI-generated concept media accurately

Label the artifact for what it is

Start the AI-generated concept disclosure workflow before generation, not at upload. Give every concept an internal state: concept, simulation, or final evidence. A concept explores a direction that has not been approved or produced. A simulation depicts a product, person, place, event, result, or interaction that was not photographed or observed as shown. Final evidence documents something that actually exists or happened and has passed the organization’s normal approval and substantiation checks. AI-assisted polish does not move an asset from concept to evidence.

For advertising, draft a qualifier against the impression created by the complete ad rather than treating an AI label as universal corrective language. The FTC’s digital-advertising guidance says to consider the ad as a whole, including its depictions; required qualifying information must be clear and conspicuous; and a disclosure may qualify a claim but cannot cure a false claim or contradict a material claim. In this workflow, candidate wording such as “AI-generated product concept; product, controls, and availability are not final” is an editorial starting point. The claim owner still has to compare that wording with the actual visual, copy, evidence, medium, and applicable rules.

Decide from the viewer’s likely impression

Ask three questions at the review gate. First, what reasonable interpretations arise from the complete combination of visual and copy? FTC guidance says each reasonable interpretation of an objective advertising claim needs substantiation. Second, does the destination’s current rule cover this asset? If qualifying realistic content falls within YouTube’s rule, use its AI-use field; if it falls within Meta’s or TikTok’s stated scope, use the disclosure mechanism that policy specifies. Third, does an objective statement or depiction lack prior support? If a disclosure would contradict a material claim, the FTC guidance says the claim itself must be modified; do not treat the disclosure as a cure.

Jurisdiction and medium still matter. The European Commission says Article 50 of the EU AI Act applies from 2 August 2026 and requires deployers to clearly label deepfakes—AI-generated or manipulated image, audio, or video resembling existing subjects or events that could falsely appear authentic. Its summary distinguishes the visible deployer label from providers’ machine-readable marking duties. Treat that as a legal review trigger, not as universal legal advice: identify audience, territory, media type, subject, and publication date, then have the responsible legal or policy owner decide which obligations apply.

Never use disclosure to manufacture proof

A label can explain that media is synthetic; it cannot turn invented evidence into substantiation. Do not depict a nonexistent testimonial, product feature, before-and-after result, certification, news event, laboratory setup, customer interaction, or performance comparison and assume “AI-generated” makes the implied claim safe. Separate the visual’s production history from the truth of what the advertisement communicates.

The FTC’s advertising-substantiation policy says advertisers and agencies need a reasonable basis for objective express and implied claims before dissemination. Its digital-disclosure guidance adds that advertisers must consider the ad as a whole, including depictions, and that a disclosure may qualify a claim but cannot cure a false one or contradict a material claim. The operating rule is therefore simple: connect every objective claim to evidence already held by the claim owner; if that evidence is absent or weaker than the visual impression, remove the claim or redesign the scene. Record “not substantiated” as a stop condition, not a copywriting task.

Match the disclosure to the destination

For YouTube, use the upload workflow’s AI-use setting when realistic content is generated or meaningfully altered. YouTube identifies examples such as making a real person appear to do something they did not, altering footage of a real event or place, or generating a realistic scene that did not occur. Selecting the field produces a viewer-facing label. That platform declaration does not establish support for a separate product or performance claim; evaluate those claims under the substantiation rule for the complete ad.

For Meta organic content, current policy requires its AI-disclosure tool for photorealistic video or realistic-sounding audio that was digitally created or altered, and Meta may label certain high-risk digitally altered content involving matters of public importance. TikTok’s current help page requires creators to label AI-generated content containing realistic images, audio, or video, describes a creator-applied label, and says TikTok may automatically label content it identifies as generated or significantly edited with AI. Platform rules differ in scope, terminology, and enforcement, so recheck the exact destination policy on every release rather than copying one platform’s answer to another.

Photoreal person or event

Disclose at first exposureName the synthetic or simulated fact and use the destination’s native declaration.

Stylized concept

Explain commercial statusState that the product, feature, package, or scene is conceptual when that status could be misunderstood.

Objective outcome

Require prior evidenceDo not publish a depicted result, comparison, testimonial, or demonstration without the matching substantiation.

Use a release gate that survives reposting

Package the final candidate with a disclosure record. Save the asset ID and version; generation or editing method; creator and approver; concept, simulation, or evidence state; exact on-media label; exact caption disclosure; destination-specific toggle selected; claims shown or implied; evidence owner and evidence location; territories; approval timestamp; and the exported file’s provenance status. The approver should inspect the actual crop, thumbnail, autoplay frame, caption truncation, audio opening, and repost format—not only the design master.

For EU-facing review, the European Commission’s optional AI-content icons provide a current placement reference rather than a compliance shortcut. Its guidance says a label should be clear at first exposure, avoid intervening overlays, remain visible when content is reshared or downloaded, and use plain language and accessible presentation. The Commission also states that using an icon does not establish legal compliance by itself. Apply those principles to the exported placement, then send scope and applicability to the responsible legal owner.

Content Credentials can carry cryptographically verifiable provenance assertions and signed claims, but they are a supporting signal rather than a complete communication plan. The C2PA security guidance says manifests can be removed from assets even though unauthorized tampering can be detected. Preserve credentials when the production chain supports them, verify that the export still contains them, and keep the visible disclosure meaningful if metadata disappears during screenshots, transcoding, cropping, or reposting.

Release only when the visible label, caption, platform declaration, claim evidence, and approval record agree. Reject if the label is missing from a derived crop, if a caption calls an image a concept while the ad claims the product is available, if the native toggle is unanswered, or if an approver cannot locate the evidence behind a depicted outcome.

Choose the next guide from the failed gate

Return to the creative-workflow hub when the concept needs a different production path. Use the thumbnail safe-zone guide if a disclosure or qualifying line disappears at small sizes or behind interface chrome. Use the camera-motion guide when an approved still concept is moving into video and the simulated action needs to remain unambiguous. These are production handoffs, not substitutes for disclosure review.

If the failure is missing evidence, stop creative iteration and send the objective claim back to its owner. If the failure is ambiguity, rewrite the label to name what is conceptual or simulated. If the failure is destination policy, obtain the current platform answer before scheduling. Do not solve any of those failures by making the disclaimer longer while leaving the misleading visual or unsupported claim intact.

Keep the page inside its evidence boundary

This guide owns the narrow workflow for distinguishing concept, simulation, and final evidence; writing a specific disclosure; applying destination controls; blocking unsupported claims; and preserving human approval. It does not decide whether a disclosure is legally sufficient in every jurisdiction, guarantee that a platform will retain metadata or labels, certify an asset as authentic, or authorize a product claim.

Recent discussions and disclosure examples did not provide a complete, reproducible concept-media release workflow or evidence of disclosure effectiveness. This page therefore makes no claim about common creator practice, engagement effects, consumer understanding, or comparative performance. Platform rules and EU guidance can change. Recheck monthly, keep the page in researched-draft status, and return it to placeholder research if the cited current rules can no longer be verified.