How to Use AI to Spot AI-Generated Images and Fake News

A dramatic photo is going round your family group chat. A shocking quote from a politician is all over your feed. A video of a celebrity endorsing an investment looks oddly smooth. AI image and video tools are now so good that you can’t always trust your eyes, and fake stories spread fastest when they make us feel angry or scared.

The good news is that the same AI tools can help you slow down and check. They can’t give you a perfect “real or fake” verdict, but used alongside a few simple fact-checking habits, they make it much easier to spot what doesn’t add up before you share it. This guide shows you a routine you can follow in about 10 minutes.

New to AI? Start with our beginner’s guide.

What you’ll need

• The image, video or story you want to check (save the image rather than screenshotting it if you can)
• A browser with Google Lens or another reverse image search, such as TinEye or Bing
• A chatbot: Gemini is useful for images, and Perplexity is handy because it shows its sources
• About 10 minutes, and a willingness not to share until you’ve checked

Step-by-step: check an image or story with AI

1. Pause before you share. Ask yourself how the post makes you feel. Content designed to make you outraged, frightened or excited is the kind most often faked.
2. Find where it came from. Run a reverse image search with Google Lens or TinEye. Look for the earliest version, the original caption and whether reputable news sites have used it. An image “from yesterday” that was online years ago settles the question.
3. Check for a watermark. Upload the image to Gemini and ask whether it was created or edited with Google AI. Gemini can detect Google’s invisible SynthID watermark. A “yes” means Google AI was involved; a “no” doesn’t prove it’s real, because other tools don’t use SynthID.
4. Look closely at the details. Ask a chatbot to point out anything that looks unusual (prompt 1), then check for yourself: garbled text on signs, odd hands or jewellery, mismatched shadows, and skin that looks airbrushed. Treat these as clues, not proof, because the best fakes have none of them.
5. Break the story into claims. Paste the text of the post or article into ChatGPT or Claude and ask it to list each factual claim and what evidence would confirm it (prompt 2).
6. Check the claims against sources. Ask Perplexity, or a chatbot with web search turned on, to find coverage from reliable outlets and fact-checkers (prompt 3). Open the links yourself and read them; don’t just trust the summary.
7. Decide and reply kindly. If it’s fake or unproven, don’t share it. If a friend or relative posted it, a short, friendly message with a link to a fact-check works better than an argument (prompt 4).

Copy-paste prompts

1. Look for signs an image is AI-generated

I'm trying to work out whether this image is a real photograph or AI-generated. Describe anything that looks unusual, such as text, hands, faces, reflections, shadows, lighting or backgrounds, and explain why each detail might be a clue. Be clear that these are clues, not proof, and tell me what else I should check.

2. Break a story into claims to check

Here is a post I've seen online: [PASTE THE TEXT]. List each factual claim it makes, in plain English. For each claim, tell me what evidence would confirm or disprove it, what kind of source would be reliable for checking it, and any warning signs in how it's written, such as emotional language, missing dates or no named source.

3. Find reliable coverage

Search for reliable news coverage and fact-checks about this claim: [CLAIM]. Prioritise established news organisations, official sources and recognised fact-checkers such as Full Fact. Summarise what they say, give me the links and dates, and tell me clearly if you can't find any reliable source.

4. Write a friendly reply to someone who shared a fake

A [RELATIVE OR FRIEND] shared this in our group chat: [DESCRIBE THE POST]. It turns out to be [FAKE / MISLEADING / UNPROVEN] because [WHAT YOU FOUND]. Write a short, warm message (3 sentences maximum) letting them know, with no blame, and include a place for me to paste the fact-check link.

Recommended tools

• Gemini: can check images for Google’s SynthID watermark, and is good at describing what’s in a picture.
• Perplexity: answers questions with numbered sources, which makes it quicker to find and open the original reporting.
• ChatGPT: strong at breaking a long post into separate claims and spotting manipulative wording; turn on web search for current events.
• Claude: careful at analysing text and good at explaining why a claim needs more evidence.

Fake pictures often travel with fake messages, so it’s worth knowing how to spot scam emails and texts with AI too. If you look after older relatives, our guide to helping elderly parents with tech and admin covers how to set them up safely, and for everyday device problems see how to use AI as your own tech support.

Tips and mistakes to avoid

• Don’t trust a single detector. No tool can reliably say whether any image is AI-made. Combine several checks.
• Remember new fakes have no history. An image made an hour ago won’t show up in a reverse image search, so an empty result doesn’t mean it’s genuine.
• Watch out for real photos with false captions. Many misleading posts use a genuine image from a different time or place.
• Check the account, not just the content. New accounts, no other posts and a sudden flood of shares are warning signs.
• Chatbots can be wrong about the news. Always open the sources they cite, and check the dates.
• Don’t upload private photos of other people just to check them. Stick to images that are already public.

FAQ

Can AI tell me for certain if an image is fake?

No. Tools like Gemini’s SynthID check can confirm that Google AI was used, but there’s no tool that can prove any image is real. Treat AI as a helper alongside reverse image search and reliable sources.

What is SynthID?

SynthID is an invisible watermark Google adds to images, video, audio and text made with its AI tools. You can upload an image to Gemini and ask whether it carries one.

How can I check a video?

Look for the original upload and the earliest date, check whether reliable outlets have reported it, and watch for lips that don’t match the words or unnatural movement. You can also ask Gemini whether a clip was made with Google AI.

What should I do if I’ve already shared something fake?

Delete it or post a correction with a link to a fact-check. Most people appreciate an honest correction.

I'm trying to work out whether this image is a real photograph or AI-generated. Describe anything that looks unusual, such as text, hands, faces, reflections, shadows, lighting or backgrounds, and explain why each detail might be a clue. Be clear that these are clues, not proof, and tell me what else I should check.

Related guides

• How to use AI to summarise research papers
• How to use AI to edit photos quickly for social media
• How to use AI to help your child with homework

Reference: Full Fact: How to spot AI-generated images