Eminence Growth Solutions, home
Menu

EmTrace reference

What AI tools leave in their files

Each major AI tool decides what to record in the images, video, audio and documents it makes. Here is what each maker says it embeds, by file type, quoted from its own pages and dated.

Last verified Platforms and AI content

Five kinds of trace

Content Credentials
A signed record (C2PA) of how the file was made, sealed so any change shows.
Invisible watermark
A signal woven into the pixels, frames, audio or wording, meant to survive copying.
AI label in metadata
A plain metadata field (IPTC Digital Source Type) saying AI made the content. Anyone can add or remove it.
Visible mark
A mark you can see, such as a logo or a sparkle in a corner.
Tool named in metadata
The tool or model written into ordinary file fields, such as Software or Creator.

By tool and file type

Tool Content Credentials Invisible watermark AI label in metadata Visible mark Tool named in metadata Source documents
Adobe Content AuthenticityAdobeLast verified User-applied1 Yes, user-applied12 Not stated Not stated Not stated
  1. Adobe Content Authenticity (Beta)
  2. Impact of applying Content Credentials
Every page read
Adobe FireflyAdobeLast verified Yes123456 Not stated Not stated No67 Issuer, date, tool16
  1. Content Credentials overview (Adobe Firefly Web Help)
  2. Adobe Firefly FAQ
  3. Adobe Firefly product page
  4. Content Credentials overview (Adobe Content Authenticity help)
  5. Partner models to generate images (Firefly Boards)
  6. Generative AI content labeling FAQ
  7. Watermarks for Generative AI content FAQ
Every page read
Illustrator generative featuresAdobeLast verified Adobe apps in general12 Not stated Not stated No13 App, model, ID1
  1. Generative AI content labeling FAQ
  2. Adobe Generative AI User Guidelines
  3. Watermarks for Generative AI content FAQ
Every page read
Photoshop generative featuresAdobeLast verified Adobe apps in general12 Not stated Not stated No13 App, model, ID1
  1. Generative AI content labeling FAQ
  2. Adobe Generative AI User Guidelines
  3. Watermarks for Generative AI content FAQ
Every page read
ClaudeAnthropicLast verified Yes12 Not stated Not stated Not stated Issuer, time, action2
  1. How Claude marks AI-generated contentAnthropic (Claude Help Center)
  2. Code execution tool (Claude Platform Docs)Anthropic (Claude Platform Docs)
Every page read
FLUX (API and Playground)Black Forest LabsLast verified Yes, as of June 20261 Not stated Not stated Not stated Not stated
  1. C2PA in FLUX generated Images (Black Forest Labs Knowledge Base)
Every page read
FLUX (open weights)Black Forest LabsLast verified Not stated Optional1 Not stated Not stated Not stated
  1. FLUX.2 README (black-forest-labs GitHub)
Every page read
CapCutByteDanceLast verified Yes (Seedream)12 Yes (Seedream)12 Not stated Yes (Seedream)1 Not stated
  1. Giving creators more control with Dreamina Seedance 2.5 and Dola Seedream 5.0 ProByteDance (CapCut Newsroom)
  2. Identify AI-Generated Content (CapCut detection tool)ByteDance (CapCut)
Every page read
DreaminaByteDanceLast verified Detector checks for it12 Yes12 Not stated On download, share2 Not stated
  1. Identify AI-Generated Content (CapCut detection tool)ByteDance (CapCut)
  2. Dreamina User Safety GuideByteDance (CapCut / Dreamina)
Every page read
Jimeng AIByteDanceLast verified Not stated Not stated Not stated Not stated Not stated Every page read
Seedream and Seedance (BytePlus API)ByteDanceLast verified Not stated Not stated Not stated On by default1 Not stated
  1. Image generation API (BytePlus ModelArk)ByteDance (BytePlus)
Every page read
Seedream and Seedance (Volcano Engine API)ByteDanceLast verified Not stated Opt-in metadata label1 Not stated Not stated Label, producer IDs1
  1. 为 AI 生成产物添加隐式水印 (Add implicit watermarks to AI-generated output), Volcano Engine ArkByteDance (Volcano Engine / 火山引擎)
Every page read
Canva Dream LabCanvaLast verified Not stated Not stated Not stated Yes1 Not stated
  1. Generate images with Dream Lab (Canva Help Center)
Every page read
Canva Magic MediaCanvaLast verified Not stated Not stated Not stated Not stated Not stated Every page read
Gemini appGoogleLast verified Yes12 SynthID123 Not stated Yes, can turn off1 Not stated
  1. Manage watermark settings in Gemini AppsGoogle (Gemini Apps Help)
  2. How we’re bringing AI image verification to the Gemini appGoogle (The Keyword blog)
  3. Google SynthIDGoogle DeepMind
Every page read
Gemini NotebookGoogleLast verified Not stated SynthID1 Not stated Yes, can turn off1 Not stated
  1. Learn about Gemini Notebook’s plansGoogle (Gemini Notebook Help, formerly NotebookLM Help)
Every page read
Google FlowGoogleLast verified Not stated SynthID1 Not stated Toggle; some regions1 Not stated
  1. Get started with Google FlowGoogle (Google Flow Help)
Every page read
ImagenGoogleLast verified Not stated SynthID (2025)12 Not stated Not stated Not stated
  1. SynthID Detector — a new portal to help identify AI-generated contentGoogle (The Keyword blog)
  2. Generate images using Imagen, Gemini APIGoogle AI for Developers
Every page read
Nano Banana (API)GoogleLast verified Yes (Google Cloud)12 SynthID3 Not stated Not stated Signer, app name1
  1. Content Credentials, Gemini Enterprise Agent PlatformGoogle Cloud Documentation
  2. How we’re bringing AI image verification to the Gemini appGoogle (The Keyword blog)
  3. Image generation (Nano Banana), Gemini APIGoogle AI for Developers
Every page read
IdeogramIdeogramLast verified Not stated Not stated Not stated Not stated Not stated Every page read
Kling AIKuaishou Technology (Kling AI)Last verified Not stated Yes1 Not stated Not stated Yes, fields unlisted1
  1. AI Content Detection (Kling AI Documentation)
Every page read
Leonardo.AiLeonardo.Ai (Canva)Last verified Not stated Not stated Not stated Not stated Not stated Every page read
Luma Dream MachineLuma AILast verified Not stated Not stated Not stated Free, Lite (dated)12 Not stated
  1. Dream Machine Guide: Payments & Subscriptions (Luma Learning Hub)Luma AI, Inc.
  2. Plans & Pricing (Luma)Luma AI, Inc.
Every page read
Meta AIMetaLast verified Not stated Content Seal1234 IPTC metadata (2024)4 Label in 20244 Yes, in 20244
  1. Generate images, edit photos, restyle rooms, and more for free with Meta AIMeta (AI at Meta)
  2. Introducing Muse Image and Muse VideoMeta (AI at Meta blog, Meta Superintelligence Labs)
  3. Content Seal READMEMeta (facebookresearch on GitHub)
  4. Labeling AI-Generated Images on Facebook, Instagram and ThreadsMeta (Newsroom)
Every page read
Azure OpenAI in FoundryMicrosoftLast verified Selected models12 Not stated Not stated Not stated Label, agent, time2
  1. Content provenance for Foundry modelsMicrosoft (Microsoft Learn)
  2. Content Credentials in Azure OpenAI (classic) – Microsoft Foundry (classic) portalMicrosoft (Microsoft Learn)
Every page read
Bing Image CreatorMicrosoftLast verified Yes (2023 statement)12 Not stated Not stated Not stated Not stated
  1. Bing Preview Release Notes: New experiences powered by Bing Image CreatorMicrosoft (Microsoft Bing Blog)
  2. Bing Image Creator and Bing Video Creator – Terms of Use
Every page read
Microsoft 365 CopilotMicrosoftLast verified Yes (PowerPoint)1 Not stated Not stated Optional23 Model, app, time (may)23
  1. Copilot in PowerPoint image suggestions FAQMicrosoft (Microsoft Support)
  2. Add watermarks to content generated or altered by using AI in Microsoft 365Microsoft (Microsoft Learn)
  3. Include a watermark when content from Microsoft 365 is AI-generatedMicrosoft (Microsoft Support)
Every page read
Microsoft CopilotMicrosoftLast verified Yes123 Not stated Not stated Not stated Not stated
  1. Transparency Note for Microsoft Copilot (for individuals)Microsoft (Microsoft Support)
  2. Frequently asked questions about Create in the Microsoft Copilot appMicrosoft (Microsoft Support)
  3. Fighting deepfakes with more transparency about AIMicrosoft (Microsoft Source)
Every page read
Microsoft DesignerMicrosoftLast verified Yes12 Not stated Not stated Optional34 Source metadata134
  1. Frequently asked questions about Microsoft DesignerMicrosoft (Microsoft Support)
  2. Fighting deepfakes with more transparency about AIMicrosoft (Microsoft Source)
  3. Add watermarks to content generated or altered by using AI in Microsoft 365Microsoft (Microsoft Learn)
  4. Include a watermark when content from Microsoft 365 is AI-generatedMicrosoft (Microsoft Support)
Every page read
MidjourneyMidjourneyLast verified Not stated Not stated Not stated Not stated Hidden ID tag12
  1. Content Authenticity (Midjourney Documentation, Midjourney Policies)
  2. Image and video verification (midjourney.com/verify)
Every page read
ChatGPTOpenAILast verified Yes1234 SynthID124 Not stated Not stated Edits, tool (2024)5
  1. Provenance signals in OpenAI-generated contentOpenAI (OpenAI Help Center)
  2. Advancing content provenance for a safer, more transparent AI ecosystem
  3. Introducing 4o Image Generation
  4. Verify OpenAI-generated content
  5. Understanding the source of what we see and hear online
Every page read
CodexOpenAILast verified Yes1 SynthID12 Not stated Not stated Not stated
  1. Provenance signals in OpenAI-generated contentOpenAI (OpenAI Help Center)
  2. Advancing content provenance for a safer, more transparent AI ecosystem
Every page read
DALL-E 3OpenAILast verified Yes, since 2024123 Not stated Not stated Not stated Edits, tool (2024)2
  1. Advancing content provenance for a safer, more transparent AI ecosystem
  2. Understanding the source of what we see and hear online
  3. Deprecations (OpenAI API)OpenAI (OpenAI API docs)
Every page read
OpenAI APIOpenAILast verified Yes, with exceptions123 SynthID, exceptions1 Not stated Not stated Issuer, model2
  1. Provenance signals in OpenAI-generated contentOpenAI (OpenAI Help Center)
  2. Content provenance (OpenAI API guide)OpenAI (OpenAI API docs)
  3. Deprecations (OpenAI API)OpenAI (OpenAI API docs)
Every page read
RunwayRunwayLast verified Yes12 Yes (2024 statement)1 Not stated Not stated Not stated
  1. Foundations for Safe Generative MediaRunway AI, Inc.
  2. Safety at RunwayRunway AI, Inc.
Every page read
Snapchat generative AISnap Inc.Last verified Not stated Not stated Not stated May be added12 Not stated
  1. Generative AI on SnapchatSnap Inc. (Snapchat Support)
  2. AI on Snapchat: Improved Transparency, Safety, and PoliciesSnap Inc. (Snap Values)
Every page read
Stability AI Platform APIStability AILast verified Release notes 2024-2512 Listed in 202413 Not stated Not stated Not stated
  1. Release Notes (Stability AI API Platform)
  2. Stable Safety
  3. Artificial Intelligence and the Content Ecosystem
Every page read
Stable Diffusion open modelsStability AILast verified Not stated Yes (reference code)12 Not stated Not stated Not stated
  1. generative-models README (Stability-AI GitHub)
  2. Artificial Intelligence and the Content Ecosystem
Every page read
Grok image generation on XxAI (site branded SpaceXSI)Last verified Not stated Not stated Not stated Not stated Not stated Every page read
Grok ImaginexAI (site branded SpaceXSI)Last verified Not stated Yes (type unstated)1 Not stated Not stated Not stated
  1. FAQ – Grok Website / AppsxAI (SpaceXSI Docs)
Every page read
xAI API (Imagine)xAI (site branded SpaceXSI)Last verified Not stated Not stated Not stated Not stated Not stated Every page read

Stated The maker says it does this. No The maker says it does not. Not stated We read the maker’s pages and found nothing either way.

A number beside a value links to the maker’s own page that says so; the same pages are listed under Source documents, and Every page read opens the full list for that tool. Last verified is the date each of those pages was last read and the quoted passage found on it.

Terms, as the standard-setters define them

C2PA Manifest

The C2PA Technical Specification 2.4 defines a C2PA Manifest as the set of provenance information about an asset made of one or more assertions (including content bindings), a single claim and a claim signature, held in a C2PA Manifest Store.

Content Credentials: C2PA Technical Specification, version 2.4Coalition for Content Provenance and Authenticity (C2PA)

Quoted passage 2.3.4 C2PA Manifest; 2.3.5 C2PA Manifest Store
The set of information about the provenance of an asset based on the combination of one or more assertions (including content bindings ), a single claim , and a claim signature . A C2PA Manifest is part of a C2PA Manifest Store .
Content Credentials

The C2PA Technical Specification 2.4 says Content Credential is the preferred non-technical term for a C2PA Manifest, that an asset's Manifest Store represents its Content Credentials, and that the term also refers to the overall C2PA technology.

Content Credentials: C2PA Technical Specification, version 2.4Coalition for Content Provenance and Authenticity (C2PA)

Quoted passage 2.3.6 Content Credential
This is the preferred non-technical term for a C2PA Manifest . The C2PA Manifest Store therefore represents the Content Credentials of an asset. Content Credentials also refers to the overall C2PA technology, and is therefore essentially treated as a plural noun.
C2PA Trust List

The C2PA Technical Specification 2.4 defines the C2PA Trust List as a C2PA-managed list of X.509 certificate trust anchors that issue certificates to hardware and software signers, and the C2PA TSA Trust List as the equivalent list for time-stamping authorities.

Content Credentials: C2PA Technical Specification, version 2.4Coalition for Content Provenance and Authenticity (C2PA)

Quoted passage 2.3.15 C2PA Trust List; 2.3.16 C2PA TSA Trust List
A C2PA-managed list of X.509 certificate trust anchors that issue certificates to hardware & software signers that use them to sign claims . 2.3.16. C2PA TSA Trust List A C2PA-managed list of X.509 certificate trust anchors that issue certificates to Time Stamping Authorities (TSAs).

Content Credentials: C2PA Technical Specification, version 2.4Coalition for Content Provenance and Authenticity (C2PA)

Quoted passage 14.4.1 C2PA Signers
For the c2pa-kp-claimSigning (1.3.6.1.4.1.62558.2.1) EKU, the list of trust anchor configurations shall include, but need not be limited to, the signer trust anchor configurations provided by C2PA (i.e., the C2PA Trust List).

ConformanceCoalition for Content Provenance and Authenticity (C2PA)

Quoted passage C2PA Trust List; Plan and Timeline
As part of the maturation of the C2PA ecosystem, the C2PA Conformance Program and the official C2PA Trust List (C2PA TL) were launched in mid-2025. The Interim Trust List (ITL) has been retired, since it was a temporary measure for early C2PA implementations. […] The ITL was frozen. No new entries will be added, and no updates will be made. Existing certificates will remain valid for legacy support, but no future refreshes or additions will occur. Eventually, those certificates will expire and no longer be usable for signing. However, if content was signed during the ITL certificate’s validity period, the content will always be considered valid against the legacy trust model.

ConformanceCoalition for Content Provenance and Authenticity (C2PA)

Quoted passage Conformance, paragraph 2
This Conformance Program is a risk-based, transparent and unbiased governance process intended to hold generator products, validator products and certification authorities accountable to the Content Credentials specification, the Certificate Policy and the Security Requirements . Conforming products are placed on a publicly-accessible list.
C2PA validation states (Well-Formed, Valid, Trusted)

The C2PA Technical Specification 2.4 says a C2PA Manifest may be Well-Formed, Valid or Trusted, and that any Trusted manifest is also Valid and any Valid manifest is also Well-Formed.

Content Credentials: C2PA Technical Specification, version 2.4Coalition for Content Provenance and Authenticity (C2PA)

Quoted passage 14.3.2 Manifest States
Based on these statements, a C2PA Manifest may be one of the following: Well-Formed Valid Trusted Any Trusted manifest is also Valid, and any Valid manifest is also Well-Formed.

Content Credentials: C2PA Technical Specification, version 2.4Coalition for Content Provenance and Authenticity (C2PA)

Quoted passage 14.3.4 Well-Formed Manifest
A C2PA Manifest is Well-Formed if validation determines that each of the following is true: The manifest’s contents abide by the normative requirements of this specification, that are validated via the validation process . Only those assertions allowed for the specific type of the manifest are present

Content Credentials: C2PA Technical Specification, version 2.4Coalition for Content Provenance and Authenticity (C2PA)

Quoted passage 14.3.5 Valid Manifest
A C2PA Manifest is Valid if validation determines that each of the following is true: The manifest is Well-Formed [ Section 14.3.4, “Well-Formed Manifest” ]. The manifest has not been modified since the manifest was signed [ Section 13.2.6, “Cryptographic validation” ]. The claim signature receives a success code of claimSignature.validated [ Section 15.7, “Validate the Signature” ]. Validation of the claim signature validity period receives the success code of claimSignature.insideValidity [ Section 15.8, “Validate the Time-Stamp” ]. The credential of the signer of the C2PA Manifest is not rejected with a failure code of signingCredential.ocsp.revoked [ Section 15.9, “Validate the Credential Revocation Information” ]. If a C2PA Manifest is Valid, then the manifest’s claim can be attributed to the claim generator which is identified by the claim_generator_info field of the claim [ Section 10.2.3, “Claim Generator Info” ].

Content Credentials: C2PA Technical Specification, version 2.4Coalition for Content Provenance and Authenticity (C2PA)

Quoted passage 14.3.6 Trusted Manifest
A C2PA Manifest is Trusted if validation determines that each of the following is true: The manifest is Valid [ Section 14.3.5, “Valid Manifest” ]. The signing credential of the C2PA Manifest receives the success code of signingCredential.trusted [ Section 15.7, “Validate the Signature” ].
Invisible watermark and visible watermark (C2PA definitions)

The C2PA Technical Specification 2.4 defines an invisible watermark as information incorporated in a substantially human-imperceptible way into an asset's digital content, for example to identify the asset or store a reference to a C2PA Manifest, and a visible watermark as a perceptible component carrying provenance information.

Content Credentials: C2PA Technical Specification, version 2.4Coalition for Content Provenance and Authenticity (C2PA)

Quoted passage 2.4.3 Invisible Watermark; 2.4.4 Visible Watermark
Information incorporated in a substantially human imperceptible way into the digital content of an asset which can be used, for example, to uniquely identify the asset or to store a reference to a C2PA Manifest . 2.4.4. Visible Watermark A perceptible component of the digital content carrying some human consumable information about the provenance of the asset .
C2PA Technical Specification version in force

Version 2.4 of the C2PA Technical Specification is dated April 2026.

Content Credentials: C2PA Technical Specification, version 2.4Coalition for Content Provenance and Authenticity (C2PA)

Quoted passage 5.3.1 2.4 – April 2026
5.3.1. 2.4 – April 2026
IPTC Digital Source Type

IPTC defines the Digital Source Type vocabulary as indicating from which source a digital image was created.

Digital Source Type (NewsCodes controlled vocabulary)IPTC

Quoted passage Metadata about the scheme
Name (en-GB) : Digital Source Type Definition (en-GB) : Indicates from which source a digital image was created.

IPTC Photo Metadata Standard 2025.1IPTC

Quoted passage 11.13 Digital Source Type
11.13. Digital Source Type Row header Specification Name Digital Source Type Definition The type of the source of this digital image Help Text Select one of the values for identifying the type of the source of the digital image from the controlled vocabulary […] CV Note(s) Required: Digital Source Type NewsCodes History Note(s) Added by v 1.1 (2009), User Note added by v 2022.1 XMP Specs Iptc4xmpExt:DigitalSourceType
trainedAlgorithmicMedia

IPTC's Digital Source Type term trainedAlgorithmicMedia, named 'Created using Generative AI', means digital media created algorithmically using an AI model trained on captured content.

Digital Source Type (NewsCodes controlled vocabulary)IPTC

Quoted passage digsrctype:trainedAlgorithmicMedia
Concept ID (QCode) = digsrctype:trainedAlgorithmicMedia, ID (URI) = http://cv.iptc.org/newscodes/digitalsourcetype/trainedAlgorithmicMedia […] Name (en-GB) : Created using Generative AI Definition (en-GB) : Digital media created algorithmically using an Artificial Intelligence model trained on captured content
compositeWithTrainedAlgorithmicMedia

IPTC's Digital Source Type term compositeWithTrainedAlgorithmicMedia, named 'Edited using Generative AI', means augmentation, correction or enhancement using a generative AI model, such as inpainting or outpainting.

Digital Source Type (NewsCodes controlled vocabulary)IPTC

Quoted passage digsrctype:compositeWithTrainedAlgorithmicMedia
Concept ID (QCode) = digsrctype:compositeWithTrainedAlgorithmicMedia, ID (URI) = http://cv.iptc.org/newscodes/digitalsourcetype/compositeWithTrainedAlgorithmicMedia […] Name (en-GB) : Edited using Generative AI Definition (en-GB) : Augmentation, correction or enhancement using a Generative AI model, such as with inpainting or outpainting operations
algorithmicallyEnhanced

IPTC's Digital Source Type term algorithmicallyEnhanced, named 'Algorithmically-altered media', means modification or correction by algorithm without changing the main content, initiated or configured by a human, such as sharpening or noise reduction.

Digital Source Type (NewsCodes controlled vocabulary)IPTC

Quoted passage digsrctype:algorithmicallyEnhanced
Concept ID (QCode) = digsrctype:algorithmicallyEnhanced, ID (URI) = http://cv.iptc.org/newscodes/digitalsourcetype/algorithmicallyEnhanced […] Name (en-GB) : Algorithmically-altered media Definition (en-GB) : Modification or correction by algorithm without changing the main content of the media, initiated or configured by a human, such as sharpening or applying noise reduction
composite (related IPTC term)

IPTC's Digital Source Type term composite, named 'Composite of elements', means a mix or composite of several elements, any of which may or may not be generative AI.

Digital Source Type (NewsCodes controlled vocabulary)IPTC

Quoted passage digsrctype:composite
Concept ID (QCode) = digsrctype:composite, ID (URI) = http://cv.iptc.org/newscodes/digitalsourcetype/composite […] Name (en-GB) : Composite of elements Definition (en-GB) : Mix or composite of several elements, any of which may or may not be generative AI

How these records are built

Every statement comes from the maker's own help, policy or engineering pages, opened and read by us, and links to the passage it rests on. Each record shows the date its pages were last verified. Third-party reporting is not used. "Not stated" means we read the pages listed and found nothing either way; it is not the same as "no".

These records describe what makers say they do. EmTrace checks files against them using our extensive library of sample files, and a file that lacks a trace proves nothing: most traces can be removed by editing, exporting or re-uploading.