isGenAI research · Original measurements
Do news sites serve images with Content Credentials?
Result on 3 November 2026: 0 of 122 images carried a C2PA manifest. We fetched the homepages of 50 news organisations once, downloaded up to five article images from each of the 29 that returned images, and read every file as served. 25 files from 7 sites kept an XMP packet, 1 kept EXIF camera data, and none carried GPS or an IPTC AI source label.
Results by record
| Record | Files (122) | Sites (29) |
|---|---|---|
| C2PA manifest | 0 | 0 |
| IPTC DigitalSourceType (AI) in XMP | 0 | 0 |
| Any XMP packet | 25 | 7 |
| EXIF camera make/model | 1 | 1 |
| EXIF GPS coordinates | 0 | 0 |
| Saved prompt (A1111 parameters) | 0 | 0 |
| ComfyUI workflow graph | 0 | 0 |
BBC is listed as a member of the Content Authenticity Initiative and returned images; none of those images carried a manifest. The other member sites in our list did not return images to a plain request. A manifest can be stripped by image CDNs and resizing, so this measures what readers receive, not what newsrooms attach in their own systems.
Results by site
| Site | Images read | Records found |
|---|---|---|
| BBC | 5 | None |
| The New York Times | No images returned | |
| Reuters | No images returned | |
| Associated Press | No images returned | |
| The Washington Post | No images returned | |
| The Guardian | No images returned | |
| CNN | 5 | None |
| Al Jazeera | 5 | Any XMP packet |
| NPR | No images returned | |
| The Wall Street Journal | No images returned | |
| Financial Times | No images returned | |
| Bloomberg | No images returned | |
| USA Today | 5 | None |
| NBC News | No images returned | |
| CBS News | 5 | None |
| ABC News | 1 | None |
| Fox News | 5 | None |
| Los Angeles Times | 5 | None |
| The Independent | No images returned | |
| The Telegraph | No images returned | |
| Deutsche Welle | No images returned | |
| France 24 | 5 | None |
| Le Monde | 1 | None |
| Der Spiegel | 5 | Any XMP packet |
| Die Zeit | No images returned | |
| El País | No images returned | |
| Corriere della Sera | 5 | None |
| la Repubblica | 5 | None |
| NRC | 5 | Any XMP packet |
| ABC Australia | 1 | None |
| CBC | No images returned | |
| The Hindu | 1 | None |
| The Indian Express | 2 | Any XMP packet |
| South China Morning Post | 5 | None |
| The Straits Times | 5 | None |
| The Japan Times | No images returned | |
| The Korea Herald | 5 | None |
| AFP | 5 | None |
| Euronews | 1 | None |
| Axios | No images returned | |
| Politico | No images returned | |
| The Atlantic | No images returned | |
| TIME | 5 | None |
| The Economist | No images returned | |
| Newsweek | 5 | Any XMP packet |
| Vox | 5 | None |
| The Verge | 5 | None |
| WIRED | 5 | None |
| TechCrunch | 5 | Any XMP packet |
| Ars Technica | 5 | Any XMP packet; EXIF camera make/model |
Method
- On 3 November 2026 we requested each homepage once with the user agent
isgenai-research/1.0. Sites that blocked the request or returned no article images are listed as such. - From each page we took the og:image and twitter:image, then article photos from
<img>tags, skipping logos, icons, SVG, GIF and files under 10 KB, up to five per site. - We read each downloaded file, unchanged, with the readers behind the C2PA checker, the IPTC/XMP AI source reader, the EXIF viewer and the saved prompt extractor.
Limitations
- One day and one homepage per site. Article pages, apps and syndication feeds may serve different files.
- Records, not origin. A missing manifest does not mean an image is fake or AI-made, and newsrooms may keep credentials on originals that their CDN strips.
- Sites that blocked automated requests are not measured, including several Content Authenticity Initiative members.
Related: what X keeps after an upload and how to check an image's history. Report a problem with a result.