Subscribe
,

How to check whether a photo or video was made with AI

AIooze Editorial

In one minute

  • You will be able to check where a photo or short video came from, and whether there are signs it was made or edited with AI.
  • You need: the image or clip (or its web address) and a web browser. Every tool here is free to use.
  • It takes about 5 to 10 minutes.
  • No single check can prove a picture is real. Several checks together give you a much better answer.

Why it is worth doing

AI tools can now make convincing photos, videos and voices of real people. Last week US prosecutors charged three people over an alleged scheme that used AI to mimic the voices and faces of company bosses and lawyers, and took more than $129m from businesses (AIooze brief). Most fakes you meet will be smaller: a shocking picture in a family group chat, a “celebrity” advert, or a video of a politician saying something odd. A few minutes of checking stops you passing on a fake, or acting on one.

The steps

  1. Pause before you share or act. If a picture makes you angry, scared or excited, or asks you to send money, treat that as a reason to check first.
  2. Look at where it came from. Who posted it first? Is it a known news organisation or an account you have never heard of? If something big really happened, several trusted outlets will usually be reporting it.
  3. Do a reverse image search. This shows where else the picture appears online, and often finds the original.
    • Google: search for the image, open it, select the info icon, then About, then Learn more about an image. According to Google’s help page, as of 9 Nov 2026, this can show when Google first saw similar images, other pages that use it, and sometimes whether it was taken with a camera or made with AI. Google says the feature is only available in some regions.
    • TinEye: drag the image onto tineye.com, paste it, upload it or enter its web address. TinEye’s tutorial explains you can sort results by Oldest, which helps you find the first time it appeared.
    • For a video, take a screenshot of a clear frame and search with that.
  4. Check for Content Credentials. Content Credentials are a kind of digital label (an industry standard called C2PA) that some cameras and AI tools attach to a file, saying how it was made and edited. Go to contentcredentials.org/verify and upload the file or paste its address. According to the Verify guide, it shows who signed the file, which app or device made it, and whether AI-generated content is listed. If it says “No Content Credential”, that does not mean the picture is fake: the project’s own FAQ notes that “a screenshot of an image wouldn’t include C2PA metadata”, and re-sharing a copy can strip it too.
  5. Ask Gemini whether Google’s AI made it. Google adds an invisible watermark called SynthID to images, video and audio made with its own AI tools. At gemini.google.com, add the file and ask “Is this AI-generated?”. According to Google’s help page, as of 9 Nov 2026, it checks one file at a time, up to 100 MB, with videos under 90 seconds, and also reads Content Credentials. The big limit: “Gemini can currently only recognise content created by Google AI tools.” A “no watermark” answer only rules out Google’s AI, not anyone else’s.
  6. Look closely. The FBI’s advice on AI fraud lists signs to look for: distorted hands or feet, odd teeth or eyes, unrealistic shadows, voices that do not match lip movements, and jerky movement in video. Newer tools make fewer of these mistakes, so treat flaws as a warning sign, not proof.
  7. If money or a person is involved, check with the person directly. Call them back on a number you already have, not one in the message. The FBI also suggests agreeing a secret word or phrase with your family.

A worked example

Your aunt shares a dramatic photo of a flooded high street in your town, “taken this morning”. You open it in Google, select About, then Learn more about an image: similar images were first seen online two years ago, on a page about a flood in another country. On TinEye, sorting by Oldest finds the same picture from 2024. You do not need to go further. You reply kindly in the group with the link to the original, and nobody spends the morning worrying.

Watch out for

  • No tool is a lie detector. Gemini’s check only knows Google’s watermark. Content Credentials can be missing, removed by a screenshot, or come from an app the checker does not recognise. Google warns that the “how it was made” details can be changed by editing tools, so they may not be accurate.
  • Uploading sends the file to a company. Do not upload private family photos, children’s pictures, intimate images or documents with personal details to any checker. Where you can, search using the image’s web address instead.
  • Real photos can be misleading too. An old, genuine picture with a new caption is one of the most common fakes. Reverse image search is the best check for this.
  • If you are being pressured, stop. Urgency, secrecy and requests for money or gift cards are the real warning signs, whatever the picture looks like.
  • Free tools and limits can change; check each maker’s help page if a step looks different.

Go deeper

Sources

Related