Sample Image Authenticity Reports

Adobe Firefly AI-generated image sample report

The origin record identifies AI-generated content.

What FauxScan found: The file's signed C2PA Content Credentials passed validation. The validated record declares AI-generated content. It lists gemini-flash as the model. The visual detector favors a camera photo, but this estimate does not override the signed declaration.
C2PA and provenance: C2PA was present, embedded, and valid. The manifest names Adobe Firefly as generator and producer, identifies Adobe Inc. as issuer, and includes an AI model reference for gemini-flash.
Metadata: No camera EXIF metadata was present, which means the file did not carry normal camera make, model, capture, or GPS context. The PNG was analyzed with native pixel-domain residual checks instead of JPEG ELA.
Why it matters: This sample shows how a visual detector can conflict with a signed origin record. The validated record declares AI generation, while the detector favors a camera photo. The detector estimate does not override that signed declaration.
Limitations: Validation supports the signed record; it does not establish scene truth or signer trust. The report does not prove where the image was first published or whether it was used deceptively. Residual analysis is a supporting signal, not standalone proof of manipulation.

Open interactive sample report

Google Gemini AI watermark sample report

The origin record identifies AI-generated content.

What FauxScan found: The file's signed C2PA Content Credentials passed validation. The validated record declares AI-generated content. The visual detector favors a camera photo, but this estimate does not override the signed declaration.
C2PA and provenance: C2PA was present, embedded, and valid. The manifest identifies Google C2PA Core Generator Library metadata and Google Media Processing Services as the signer.
Metadata: The image did not include camera EXIF metadata or GPS context. Because the source is PNG, FauxScan uses native pixel residual analysis instead of JPEG ELA.
Why it matters: This sample demonstrates a high-confidence provenance case: the validated record declares AI generation, while the visual detector favors a camera photo. That estimate does not override the signed declaration.
Limitations: The validated record supports the signed AI declaration; it does not establish scene truth or signer trust. The report does not identify the uploader, original prompt, publication path, or user intent behind the image.

Open interactive sample report

Adobe Firefly AI-modified image sample report

The origin record identifies AI modifications.

What FauxScan found: The file's signed C2PA Content Credentials passed validation. The validated record declares AI modifications to source imagery. It lists gemini-flash as the model. The visual detector favors a camera photo, which is compatible with the record's declaration of AI involvement.
C2PA and provenance: C2PA was present, embedded, and valid. The provenance record names Adobe Firefly, Adobe Inc., and a trained-algorithmic-media source type.
Metadata: No EXIF camera metadata was present. Without camera make, model, or capture context, the provenance record becomes the main source of origin evidence.
Why it matters: This sample helps reviewers distinguish AI-modified content from purely AI-generated content and from ordinary camera-origin photos that have simply been exported or recompressed.
Limitations: Validation supports the signed record; it does not establish scene truth or signer trust. The report does not show the exact edited region or prove the edit's purpose. Residual and frequency signals support review, but the signed provenance record is the strongest evidence.

Open interactive sample report

Likely authentic iPhone photo sample report

The scan findings suggest this image originated as a photograph captured by a physical camera. This does not establish whether the photo is unmodified.

What FauxScan found: Camera metadata identifies Apple iPhone 15 Pro as the capture device. The visual detector also favors camera origin. Compression variation can result from brightness or contrast adjustments, camera processing, or JPEG saving. This finding alone does not establish localized editing.
C2PA and provenance: No C2PA Content Credentials were found. Camera metadata and the visual detector support camera origin. Compression variation was also found; it does not establish localized editing or confirm that the photo is unmodified.
Metadata: Camera make, camera model, software version, and GPS coordinates were present. These are useful source-context signals, though they still need chain-of-custody review in high-stakes settings.
Why it matters: This sample shows why camera origin and editing are separate questions. Camera metadata and the visual detector support photographic origin, while compression variation can reflect adjustments, camera processing, or JPEG saving. Review the processing findings if possible editing would affect your assessment.
Limitations: Likely authentic does not mean legally proven original. Metadata can be copied, stripped, or altered, and location context should be confirmed against the source file and surrounding evidence.

Open interactive sample report

Likely inauthentic image sample report

The image's origin remains unverified.

What FauxScan found: The scan did not establish a validated origin record or enough supporting evidence to confirm the source. Camera details are missing; sharing or exporting can remove them. This is not evidence of AI generation.
C2PA and provenance: No C2PA Content Credentials were found. That absence does not prove manipulation, but it removes an important source of signed origin evidence.
Metadata: Camera details are missing; sharing or exporting can remove them. This is not evidence of AI generation.
Why it matters: This example retains the Likely Inauthentic display label. Its latest scan returned Authenticity Unverified: the available evidence does not establish the image’s origin. Missing camera details alone do not prove AI generation or manipulation. Check the capture and editing history to resolve the uncertainty.
Limitations: The report does not prove that the image is fake or identify a pasted region. It flags weak authenticity support and recommends manual review before reliance.

Open interactive sample report

Authenticity unverified image sample report

The scan found processing patterns, but could not determine the image's origin or editing method.

What FauxScan found: Texture patterns can result from AI generation, resizing, or other image processing. The visual detector favors a camera photo, but that estimate does not resolve the image's origin or explain the processing patterns.
C2PA and provenance: No C2PA Content Credentials were found. Without signed provenance, the report cannot tie the file to a verified creator, capture device, or generation tool.
Metadata: The image lacked camera and GPS metadata. Because the source is PNG, FauxScan uses native pixel residual analysis and does not claim JPEG compression-history evidence.
Why it matters: This sample shows why processing patterns do not establish a particular origin or editing method. Texture patterns can result from AI generation, resizing, or other processing. The camera-origin detector estimate does not explain those patterns, so the result remains unverified.
Limitations: The result does not prove AI generation or manipulation. The scan cannot determine what caused the processing patterns. Review the forensic findings and the image’s capture and editing history to assess whether possible changes affect the details you need to rely on.

Open interactive sample report

Integrity compromised image sample report

The scan found signs of localized editing that compromise confidence in the image's integrity.

What FauxScan found: Some edges show a strong compression difference that suggests possible localized editing. Compression-error patterns suggest a localized region may have been edited. The visual detector favors camera origin. Camera origin does not establish that the image is unmodified.
C2PA and provenance: No C2PA Content Credentials were found. The integrity concern comes from image-forensic signals rather than signed provenance.
Metadata: No EXIF metadata was present, so the report could not use camera make, model, capture time, or GPS data to support source verification.
Why it matters: This sample shows how FauxScan can separate AI-origin questions from integrity questions. A human-origin detector result can coexist with localized evidence that the image may have been altered.
Limitations: The report does not prove exactly what was changed. Boundary and ELA findings should guide manual inspection, comparison to originals, and source-context review.

Open interactive sample report