Reader trust collapses one bad photo at a time. This newsroom photo verification checklist turns a fuzzy editorial judgement into a repeatable seven-step routine your desk can run on every submitted image — in under ten minutes.
The 7-step checklist at a glance
- 1. Capture provenance
- 2. Inspect EXIF metadata
- 3. Reverse image search
- 4. AI-generation detection
- 5. Tamper / splice analysis
- 6. Location cross-check
- 7. Document the verdict
1. Capture provenance first
Before any tool, log: who sent the photo, on what date, through which channel, and what they claim it shows. Provenance failures cause more retractions than forensic mistakes.
2. Inspect EXIF metadata
Open the file in the EXIF Viewer. Confirm:
- Capture device matches the source's claimed equipment.
- Timestamp matches the claimed event window.
- GPS coordinates, when present, plausibly match the claimed location.
Missing EXIF is common after social uploads — treat it as neutral, not damning.
3. Reverse image search
Run the image through the Reverse Image Search tool. An older hit means the photo is recycled. A matching stock library means it is illustrative, not journalistic.
4. AI detection pass
Score the image with the AI Image Detector. A high synthetic score does not prove fakery on its own, but combined with a stock-source hit or implausible EXIF, it becomes decisive.
5. Tamper analysis
Use the Tamper Heatmap to look for regions with inconsistent JPEG quantisation. Splices, cloned crowds, and pasted objects usually leave visible heat.
6. Location cross-check
If the claim depends on where the photo was taken, run it through AI Location Search. A mismatch between the claimed and inferred location is a hard stop until resolved.
7. Document the verdict
Save scores, screenshots, and your reasoning into a shared verification log. This is non-negotiable: it protects the masthead if the story is later challenged, and it lets newer staff learn from edge cases.
Operationalising the checklist
- Pin it. Add the seven steps as a Slack pinned message or CMS pre-publish checklist.
- Define the second-eyes rule. Any conflicting signal triggers a second editor review.
- Set a "do-not-publish" line. Three or more concerning signals = hold the image until cleared.
The point is not perfection — it is creating an auditable process that makes accidental publication of fakes far less likely, and easy to defend when something does slip through.