How to Verify AI Images in News and Local Media
For a reader photo, the file and the source carry more weight than any tool. Ask where and when the picture was taken, get the original rather than a forwarded copy, and run a reverse image search before you look at provenance or a detector. Local desks often publish images they cannot visit in person, so a repeatable check beats a one-off judgment call.
Reader submissions and community reports
Community photos arrive with good intent and thin context. The image may be genuine but from another place or year, or it may be generated. Those are different problems, and a reverse image search separates them fast. If the same photo appears in an unrelated story or an old post, you have your answer without opening a detector. That matters on deadline, and it is also why the reader conversation comes first: the answers to “where” and “when” shape everything else you check.
Five questions for the person who sent the photo
- Did you take this photo? If someone forwarded it, who took it?
- Where were you standing, and when was it taken? Ask for a nearby street or landmark you can verify.
- Can you send the original file and an adjacent photo from the same sequence?
- Was anything added, removed, generated, or substantially edited?
- How should we credit you, and do we have permission to publish it?
For a claimed local flood photo, compare buildings, road signs, and the stated time with independent local reporting. A genuine photograph of a different town is a context problem even if every detector gives it a low AI score. Record the reader’s answers with the file, then pass both to the editor.
A workflow that fits a local desk
- Source. Who sent it, did they take it themselves, and does their account or history support that? Ask for the original file.
- Reverse search. Google Lens or TinEye shows earlier appearances and different captions. A hit is stronger than any score.
- Provenance. Inspect the file for Content Credentials at contentcredentials.org/verify. A readable provenance record can identify the software or service used to create or edit the file. A missing record proves little.
- Detector. Score the original with a checker and review its AI confidence score. Compression and edits move the number.
- Corroborate. Look for a second source, an on-the-ground detail, or a matching original. If it does not hold, say so.
What to tell your audience
Be specific about what you know. If you verified the location and time, say so. If you could not, say that too. “We have not verified this image” is a complete sentence, and it protects both the report and the reader who sent it. Avoid presenting a detector score as proof, and link to the evidence you do have. Our guide to AI image detection is in Do AI Image Detectors Actually Work?.
The broader newsroom workflow, including deadline pressure and how to describe a result in copy, is in the guide for journalists. The general visual method is in the seven-step image check.
Common questions
What is different about local news and reader photos?
Local desks often publish photos from readers they cannot visit in person. That makes the source and the original file the main evidence, and it makes a careful, repeatable check more valuable than any single tool.
What should I ask a reader who sends a photo?
Ask where and when it was taken, for the original file rather than a forwarded copy, and whether they took it themselves. Cross-check the location and time against the story you are reporting.
Can a detector settle whether a reader photo is real?
Use the AI confidence score to document your image review. The editorial team should confirm the source and context before publishing.
How should I describe an unverified image to my audience?
Say what you could and could not confirm. "We have not verified this image" is a complete, honest sentence, and it is better than a firm claim you later have to correct.
A fast check that does not store files
AI Detector scores an image in seconds and deletes it immediately after processing.
