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.
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. OpenAI images carry Content Credentials and SynthID, and Google adds SynthID to Gemini output, so a manifest can point to a generator. A missing record proves little.
- Detector. Score the original with a checker and treat the result as one signal. 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 overview of detector limits 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?
No. A detector returns a probability and produces false positives on edited or compressed photos. Use it as one signal, alongside the source, the original file, and provenance.
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.
