ChatGPT disclosure guide
Does content created with ChatGPT need AI disclosure?
Direct answer
Not automatically. Treat ChatGPT-written text and ChatGPT-generated images separately. The content, publication purpose, human editorial control and destination platform determine the disclosure answer. Ordinary AI-assisted copy is not automatically covered, while realistic generated images and unreviewed public-interest text can require separate analysis. Record the responsible human editor and final publication context.
Applies when
- Photoreal image of an event that never happened — Disclosure required.
- Generated visual used as an ad background — Disclosure recommended.
Doesn't apply when
- AI draft rewritten and approved by a marketing editor — No disclosure identified.
Worked decision
Example
AI draft rewritten and approved by a marketing editor
Decision: No disclosure identified
Assess the final output rather than the tool name alone.
No public wording identified; record AI-assisted drafting and the editor.
Next steps
Required action
- Record exactly how ChatGPT was used and which component of the final asset it affected.
- Verify the final export instead of assuming its provenance status: C2PA and SynthID documented for compatible generated images.
- Apply the destination platform’s native label and precise audience-facing wording where the resulting content triggers them.
Publishing layer
Platform differences
Detection and native labels
Platforms may read image provenance or respond to publisher self-disclosure.
Policy threshold
AI-written text usually has no equivalent universal native platform label.
Publishing workflow
Platform detection is not a substitute for documenting the actual ChatGPT workflow.
Evidence
Source
Freshness and accountability
Last verified
Last reviewed 2026-09-12 · Current
Run this exact scenario
ChatGPT and a representative content scenario are prefilled. Confirm the actual AI operation, affected component and destination.
Tool → type of AI usage
- Drafting and editing text
- Generating and editing images
- Transforming uploaded material
- Ideation that may leave no AI-generated material in the final asset
Does ChatGPT generate C2PA or other provenance?
documented
C2PA and SynthID documented for compatible generated images
OpenAI states that compatible images generated with ChatGPT, Codex and its API include C2PA metadata and SynthID. Screenshots and later processing can remove metadata, so a missing credential does not prove that an image was not generated with ChatGPT.
EU implications
- Ordinary AI-assisted marketing copy is not automatically subject to Article 50 public disclosure.
- Public-interest text without qualifying human editorial responsibility requires separate Article 50 analysis.
- A realistic ChatGPT-generated image can require disclosure when it meets the deepfake definition.
More examples and exact wording
| Use of ChatGPT | Decision | Exact wording |
|---|---|---|
| Photoreal image of an event that never happened | Disclosure required | AI-generated image. The depicted event is fictional. |
| Generated visual used as an ad background | Disclosure recommended | Background generated with AI. Product photographed. |
Tool capability, EU law and platform policy are separate evidence layers. This page uses first-party product documentation and primary regulatory sources; it is not legal advice.