Mission Pointers

How to spot AI images and deepfake videos: a UK guide

How to tell if a photo or video is AI-generated: a simple stop-and-check routine, the labels to look for, and what to do if you find a fake.

Mission Pointers · 4 October 2026 · 10 min read

In short

  • AI fakes are getting harder to spot by eye, so checking where something came from matters more than looking for glitches.
  • Before you share, check who posted it, whether trusted news outlets are reporting it, and run a reverse image search.
  • Labels and Content Credentials can help, but a missing label doesn't prove something is real.
  • If you find a fake, don't share it, even to mock it. Report it on the platform instead.
On this page
  1. Why it matters
  2. Why looking closely isn't enough
  3. A stop-and-check routine
  4. Content Credentials and AI labels
  5. Clues you can still look and listen for
  6. If you think you've found a fake
  7. Talking to friends and family about fakes
  8. Printable checklist
  9. Sources

A photo of a dramatic event is everywhere on your feed. A video shows a well-known presenter recommending an investment. A clip of a politician saying something shocking is being shared in your family group chat. Is it real?

Artificial intelligence (AI) tools can now make convincing fake photos, videos and voices in minutes. These are often called deepfakes. Spotting them by eye is getting harder, so this guide focuses on simple checking habits that work whatever the technology does next.

It's written for everyday internet users, community groups and educators across the UK. You'll get a five-step "stop and check" routine, an explanation of AI labels and Content Credentials, the visual clues that can still help, and a checklist you can print.

Why it matters

Deepfakes are already common. In 2024, Ofcom found that 43% of people aged 16 and over said they'd seen at least one deepfake online in the previous six months. Only 9% felt confident they could spot one. In February 2026, the Home Office estimated that 8 million deepfakes were shared in 2025, up from 500,000 in 2023.

They're used in three main ways, which Ofcom sums up as content that can demean, defraud or disinform:

  • Scams and fake endorsements. Celebrity scam adverts were the vast majority of the Advertising Standards Authority's (ASA) scam ad alerts in 2024. In February 2024, the ASA was alerted to a deepfake video ad of King Charles recommending a cryptocurrency investment. Surrey Police, citing Action Fraud figures, reports that in 2024 victims lost more than £10 million after being convinced by deepfakes of influential people advertising fake investments.
  • Misinformation. Fake images and videos can mislead people about news events, public figures or political issues.
  • Harming individuals. Deepfakes can be used to humiliate, harass or blackmail people, including with fake sexual images. In Ofcom's 2024 research, one in seven adults who'd seen a deepfake said they'd seen a sexual one.

There's also a quieter risk. Ofcom points out that as fakes become more common, people may dismiss real photos and videos as fake. Good checking habits protect you from both.

Why looking closely isn't enough

Advice about spotting fakes used to focus on mistakes, such as extra fingers. You can't rely on those any more. Full Fact, the UK fact-checking charity, warns that because AI tools keep improving, visual tips "might not work in future". Research cited by Ofcom found that people's guesses about whether a video was fake were, for the most part, "as good as flipping a coin".

Online "AI detector" tools aren't the answer either. Full Fact doesn't rely on them because they don't work consistently.

So the most reliable approach is to check where something came from before deciding whether to believe it or share it.

A stop-and-check routine

1. Pause before you react

Fakes are often designed to make you feel shocked, angry or excited, so you share or act quickly. If a post gives you a strong reaction, take that as a cue to slow down.

2. Check who posted it

  1. Look at the account that posted it. Is it the real, official account of the person or organisation, or a lookalike?
  2. Check how long the account has existed and what else it posts.
  3. Ask whether this person would really say or do this. The ASA suggests considering whether what someone says in a video seems out of character.

3. See if reliable news outlets are reporting it

If something big really happened, established news organisations will usually be covering it. Full Fact notes that if an image only appears on social media and hasn't been published by a reliable news website, that may be a clue it isn't real.

A reverse image search shows where else a picture has appeared online. It can reveal an older original, a different context or a fact-check.

Google Lens on a computer:

  1. Go to google.com and select Search by image in the search box.
  2. Select Upload a file and choose the image, or paste the image's web address.
  3. In Chrome, you can also right-click an image on a web page and choose Search with Google Lens.
  4. Look through the results for websites that use the image or similar ones, especially older ones.

Google Lens on a phone:

  1. In the Google app or Chrome, touch and hold the image.
  2. Tap Search image with Google Lens (on an iPhone in Chrome, it says Search image with Google).
  3. To search a saved photo or screenshot, tap the Google Lens icon in the search bar and choose the image.

Bing Visual Search:

  1. Go to Bing Images and select the camera icon in the search box.
  2. Drag in an image, upload one from your device, or paste an image or its web address.
  3. In the Microsoft Edge browser, you can right-click any image and choose Visual Search.

5. Check the fact-checkers

Professional fact-checkers often investigate viral images and videos. Search their websites for the topic:

  • Full Fact, the UK's independent fact-checking charity.
  • BBC Verify, the BBC's team for fact-checking and checking footage.

If nobody has checked it yet, don't assume it's real. Wait until you know more before you share.

Content Credentials and AI labels

Some AI-made content now carries hidden information or visible labels. These can help, but each has limits.

Content Credentials (C2PA)

The Coalition for Content Provenance and Authenticity (C2PA) has created an open technical standard called Content Credentials. It's like a "nutrition label" for digital content. It can record how a file was made, which tools were used and how it's been edited. Its steering committee includes Adobe, the BBC, Google, Meta, Microsoft, OpenAI and TikTok.

You can upload a file to the free Content Credentials Verify website to see any credentials attached to it.

The limits:

  • Adding credentials is optional, so most content won't have them.
  • Credentials can be separated from a file, for example when it's shared on a platform that doesn't support them.
  • C2PA itself says provenance information alone can't tell you whether content is true, accurate or factual.

Google SynthID

SynthID is Google's invisible watermark for content made or edited with its AI tools.

You can check for it in the Gemini app. When signed in, upload an image, a short video (under 90 seconds) or an audio file and ask, "Was this created or edited by Google AI?" The limit is important: SynthID only detects content from Google's AI. If no watermark is found, the content could still have been made by another company's AI tool.

Platform AI labels

  • Meta (Facebook, Instagram, Threads) shows an AI info label. It's applied when Meta detects industry-standard signals of AI, or when people disclose it themselves. For content that has only been edited with AI, the label may sit in the post's menu rather than on the post.
  • YouTube requires creators to disclose realistic content made with altered or synthetic media. The label usually appears in the expanded description. For sensitive topics such as health, news, elections or finance, it also appears on the video. YouTube may add a label itself if a creator doesn't.
  • TikTok requires creators to label realistic AI-generated content. It also labels content automatically using Content Credentials. By November 2025, it said more than 1.3 billion videos had been labelled.

Clues you can still look and listen for

These clues can sometimes help, but treat them as hints, not proof. Newer tools make fewer mistakes, and real photos can have odd details too. Look for several clues together, and always combine them with the checks above.

In images:

  • Hands and fingers. Look for the wrong number of fingers or awkward positions.
  • Text. Signs, labels and logos may be garbled or nonsensical.
  • Faces and details. Look for odd teeth, mismatched earrings or faces that aren't quite symmetrical.
  • Lighting and shadows. Check that shadows fall the right way and the light is consistent.
  • An unnatural finish. Some AI images have a glossy, airbrushed or cartoonish sheen.

In videos:

  • Lip-sync. Watch the mouth on "m", "b" and "p" sounds, where the lips should close.
  • Blinking and head movement. Check whether the person blinks and moves naturally.
  • Glitches. Faces may blur or distort when someone turns sideways or puts a hand in front of their face.
  • Expression. Deepfaked faces can look flat or stiff.

In audio:

  • Robotic tone, unnatural pauses, or odd changes in pitch or speed.
  • Words or phrasing the person wouldn't normally use.
  • Background noise that changes or cuts out unexpectedly.

Fake audio is a growing problem in phone scams. Our guide to AI voice-cloning scams explains how to protect yourself and your family.

If you think you've found a fake

  1. Don't share it, even to warn people or make fun of it. Sharing spreads it further.
  2. Report it on the platform. On YouTube, for example, select More (the three dots), then Report. Most social media platforms have a similar option.
  3. Report scam adverts. If it's a paid-for advert, you can also report it to the ASA using its online scam ad form. The ASA shares reports with major platforms and the National Cyber Security Centre.
  4. If you've lost money, contact your bank straight away. Then report it to Report Fraud on 0300 123 2040 in England, Wales and Northern Ireland, or to Police Scotland on 101.

Talking to friends and family about fakes

Most people who share fakes do so in good faith. A kind, curious approach works better than telling someone they've been fooled.

  • Talk privately. A direct message is less embarrassing than a public correction.
  • Share how you checked. Saying "I did a quick reverse image search and found this" helps people learn the habit.
  • Ask questions together. The Government Office for Science notes that media literacy is about asking where something came from, whether it's plausible and what it's for.

For groups and classrooms, try practising the routine on a few real fact-checks from Full Fact or BBC Verify.

Printable checklist

Before you believe or share an image or video:

  • I've paused, especially if it made me angry, shocked or excited
  • I've checked who posted it and whether it's their real account
  • I've checked whether reliable news outlets are reporting it
  • I've done a reverse image search (Google Lens or Bing Visual Search)
  • I've searched Full Fact and BBC Verify
  • I've looked for an AI label or Content Credentials, and remembered that no label doesn't mean it's real
  • I've looked and listened for clues, but haven't relied on them alone
  • If it's asking for money or promoting an investment, I've stopped and checked independently
  • If it's fake, I've reported it and not shared it

Mission Pointers offers practical sessions on AI, misinformation and online safety for community groups. Find out about our community programmes.

Sources

Next step

Want to run a session on spotting fakes and misinformation for your group?

Explore our community programmes

More guides