Do not present an AI-generated before-and-after transformation as evidence that a real product, service or intervention achieved that result. Separate authentic observations from illustrative material, and review the implication of the complete sequence rather than relying only on a small disclaimer.
Identify the result viewers may infer
Before approving a comparison, write one sentence describing its apparent claim: “This treatment produced this change”, “This renovation was completed” or “This cleaning product removed these marks”. Then ask what evidence supports that exact claim.
For Indian creators preparing multilingual campaigns, review the visual sequence alongside every language version. A cautious English caption does not answer what another headline or voiceover implies. Treat this as an editorial accuracy check, not a substitute for relevant professional review.
Create an evidence-origin register
Use structured records instead of a table. For every before and after image, enter:
- Panel ID and intended placement.
- Origin: recorded observation, edited recording, composite or generated illustration.
- Subject: fictional scene or consented real subject.
- Capture date and conditions, where applicable.
- Alterations affecting the apparent result.
- Evidence supporting the claimed change.
- Proposed disclosure and unresolved questions.
Keep the original evidence separate from promotional edits. Never invent a capture date or describe a generated panel as a customer result. Do not use customer photographs or private records for exploratory generation.
Hypothetical disclosure decision exercise
Suppose a fictional cleaning advertisement uses two self-created kitchen drawings. The second illustration removes stains and adds a brighter surface. No cleaning experiment has occurred.
Evaluate two possible presentations. “Actual results in five minutes” has no support in these inputs and should stop. “Illustrated cleaning scenario” more accurately describes their origin, but still needs review if the sequence implies a specific product can achieve the depicted outcome.
The next step is either to redesign the piece as a clearly fictional demonstration without unsupported performance claims, or commission an appropriate documented evaluation before making a results claim. Generated imagery cannot fill a missing evidence record.
Do not confuse moderation with validation
Runway says its usage policy works with its terms to mitigate harm and that it uses automated systems and internal human review to detect and block harmful inputs and outputs. This does not establish that an allowed image is factually substantiated.
If sourcing production access, check the AI tools catalog for access format and saved terms before buying. Marketplace account, code or token offers are not proof of official subscription equivalence, and offer descriptions do not validate advertising evidence.
Sources and scope
Source material checked 29 October 2026; features/policies may change. Runway usage policy establishes the bounded harm-mitigation and review statements above. The evidence register, disclosure exercise and stop decision are editorial guidance, not legal advice or a claim that moderation guarantees accuracy.