Designers now work with AI images every day, sometimes on purpose and sometimes without knowing it. A stock photo that looks a little too clean, a client-supplied hero image, a “photo” pulled from a mood board: any of them might have come out of a generator. That matters more than it used to. Clients ask for disclosure, some platforms label AI imagery, and a few stock and editorial outlets reject it outright.
Here is a practical checklist for checking an image before it goes into a layout, a campaign or a portfolio.
1. Ask where it came from
The fastest check is provenance. Who made it, with what, and can they send the original file? A photographer can send a raw file or the camera original. A stock library can show you the contributor and the license. If nobody can tell you where an image came from, treat that as a finding in its own right.
2. Read the file, not just the pixels
Original files carry metadata. Camera images usually include EXIF fields for the body, lens and exposure. Many AI tools write their own name into the file, either in the software field or in a Content Credentials (C2PA) manifest. Adobe Firefly, for example, attaches Content Credentials to its output, and several other generators leave identifiable tags.
The catch: social platforms, messaging apps and most export workflows strip that data. A screenshot or a re-saved copy carries nothing useful. Always check the earliest version of the file you can get.
3. Look for the classic visual tells
- Text inside the image. Signs, labels and logos that look like lettering but do not spell anything are still one of the most reliable giveaways.
- Hands, teeth and ears. Count fingers, check that teeth are separate and that earrings match.
- Repeating patterns that drift. Fabric, tiles and brickwork that change pattern halfway across the surface.
- Light that disagrees with itself. Shadows falling in two directions, or reflections that do not match the scene.
- Backgrounds that dissolve. Railings, window frames and furniture that merge into each other.
Newer models have fixed many of these, so a clean result on this list does not prove an image is real.
4. Run it through a detector
For anything you are about to publish or license, a second opinion from software is worth the thirty seconds. One simple way to check if an image is AI generated is a free web checker that gives a whole-image verdict with an AI score, plus a separate source check that reads the file’s metadata and Content Credentials and names the generator when the file still carries that evidence. It also scores partial edits, which matters for designers, because a real product shot with an AI-filled background is a different problem from a fully generated image.
Treat the score as evidence rather than proof, especially for borderline results, and combine it with the checks above.
5. Decide what you will disclose
AI imagery is not automatically a problem. It is a problem when a client, an audience or a platform assumes an image is a photograph and it is not. Agree a simple rule with clients up front: generated or heavily AI-edited visuals are labelled, and the team keeps a note of the tool used.
6. Keep a record
For campaign and editorial work, save the original file, the source and the result of any check in the project folder. If a question comes up six months later, you will have an answer instead of a guess.
Why this belongs in the workflow
Checking images used to be a job for photo editors and fact-checkers. Now it sits with anyone who chooses visuals for a living. A two-minute routine before an image goes live protects the client, the brand and your own reputation, and it is far cheaper than a public correction.
- PicCollage Logo: How Colour, Shape and Simplicity Express Creativity - 5 November 2026
- Before You Publish That Image: A Designer’s Checklist for AI-Generated Visuals - 5 November 2026
- How to Unlock Canva Page and Locked Elements - 4 November 2026