OpenAI research

OpenAI Watermarks & Content Provenance

What OpenAI currently documents about C2PA, SynthID, supported ChatGPT, Codex and API images, public verification, and what remains unknown for text.

Published: August 15, 2026
Last updated: August 17, 2026
Review: Reviewed against primary sources

Supported OpenAI-generated images use layered provenance rather than one single visible watermark. This hub is the OpenAI research profile. The OpenAI image watermark pillar answers the generic “OpenAI image watermark” query. The ChatGPT image watermark page covers consumer ChatGPT image behaviour.

Image provenance

Current OpenAI Help Center documentation states that supported images generated with ChatGPT, Codex, and the OpenAI API include both C2PA Content Credentials and SynthID watermarks. C2PA is signed metadata that can be stripped. SynthID is described as a signal in the media that may survive some edits. Generic C2PA is not, by itself, an OpenAI authorship verdict. Missing C2PA does not mean the file is not OpenAI-generated. Current Help Center documentation covers supported images generated with ChatGPT, Codex, and the OpenAI API. Coverage may vary by product, model, export path, file type, and generation date. Do not assume every historical OpenAI image contains both C2PA and SynthID.

C2PA, SynthID, and visible disclosure

Those layers are independent. OpenAI also documents that a visible disclosure can be requested in a prompt. That optional graphic is not SynthID, is not Content Credentials, and is not documented as a default logo on every ChatGPT image.

Audio

On 31 July 2026 OpenAI said supported audio generated through OpenAI tools, including ChatGPT and the OpenAI API, includes SynthID watermarking, and that public verification covers supported audio as well as images. This site has no audio inspector. Audio is mentioned here as supporting provenance context, not as a separate SEO cluster.

Public verification and API

OpenAI publishes a web verifier for supported image and audio files, and documents a Content Provenance API that can return C2PA and SynthID results for supported images (SynthID for audio). Those are OpenAI products. OpenAI documents a Content Provenance API for supported image and audio files. AI Watermark Center does not integrate that API and does not upload user files to OpenAI. AI Watermark Center does not currently detect SynthID.

Official OpenAI verification (opens in a new tab)External provider tool

Text

Current reviewed Help Center tables list images and audio. This database does not infer a general ChatGPT or OpenAI text watermark from image or audio documentation. Claim status for an OpenAI text watermark is unknown. That is not the same as “OpenAI does not watermark text.” OpenAI has said it intends eventually to expand provenance to all modalities including text; an intention is not a current text-watermark claim.

Historical rollout

OpenAI states that it began adding C2PA metadata to DALL·E 3 images in ChatGPT and the API in 2024, and later documented layered C2PA plus SynthID for supported current image generation. Do not read that chronology as if later files all still carry the same signals. GPT-4o-era launch posts are useful as history; current Help Center and API provenance docs are the source of current state.

Sora (adjacent, not this cluster)

OpenAI’s Sora safety documentation describes visible and invisible provenance and C2PA on Sora video outputs. That is part of OpenAI’s broader provenance strategy. It is not a current Help Center image/audio table row and is not a Phase 7 video cluster.

AI Watermark Center support

Embedded C2PA on JPEG/PNG/WebP: available
SynthID detection: not available
Audio inspection: not available
OpenAI Verify and Content Provenance API: not integrated

Cluster

Open OpenAI image watermark research

Sources

Provenance signals (Content Credentials, SynthID) in OpenAI-generated content — OpenAI Help Center
Accessed August 17, 2026.
Primary source for current OpenAI image C2PA + SynthID on supported ChatGPT, Codex, and API outputs; audio SynthID; prompt-requested visible disclosure; and openai.com/verify. Visible marks are separate from embedded signals.
Advancing content provenance for a safer, more transparent AI ecosystem — OpenAI
Published July 31, 2026. Accessed August 17, 2026.
31 July 2026 update: supported audio SynthID, audio on the public verifier, and verification API access. Historical context for layered C2PA + SynthID; current product scope follows the Help Center.
Understanding the source of what we see and hear online — OpenAI
Accessed August 17, 2026.
Historical: OpenAI began adding C2PA metadata to DALL·E 3 images in ChatGPT and the API in 2024. Chronology only — not current product-scope documentation. Also listed as OPENAI-IMAGE-HISTORY-001 in Phase 7 source notes.
Verify — OpenAI
Accessed August 17, 2026.
Public verification surface referenced by OpenAI Help Center for supported image and audio provenance signals (C2PA and/or SynthID). External provider tool — not an AI Watermark Center feature.
Content provenance — OpenAI API documentation
Accessed August 17, 2026.
Content Provenance API (also referenced as OPENAI-API-PROV-001 in Phase 7 notes). Images: C2PA + SynthID. Audio: SynthID. Documents not_detected semantics. Public/documented; not integrated by AI Watermark Center.
Creating with Sora safely — OpenAI
Accessed August 17, 2026.
Adjacent Sora video provenance (visible marks, C2PA). Not a Phase 7 cluster and not a current Help Center image/audio table row. Do not use as current ChatGPT image-scope evidence.