Skip to content
EN

Lesson 4 of 8 · 6 min · +100 XP

How can you tell if an image is AI-generated?

To tell if an image is AI-generated, look for odd details like garbled text, strange hands or shadows, then check for an AI label or Content Credentials, run a reverse image search and find the source. For text, detectors are unreliable. No single test is proof, so combine clues and trust where something comes from more than how it looks.

▶ Try it

Why it matters

AI images and text are getting very convincing, and visual clues are fading. You'll learn to combine clues with labels, reverse searches and the original source.

In 6 minutes you'll be able to

  • Look for clues like garbled text and odd details, but don't rely on them alone.
  • Check for AI labels or Content Credentials, and run a reverse image search.
  • Find the original source and see if trusted outlets confirm it.

What you'll practise

  1. Real or generated?

    Tap every sign that this post needs checking before you share it.

    @daily_wow_news
    BREAKING: giant wave hits the city harbour!! 📷 Photo: the shop sign reads "HARBUOR CAFFE" in wobbly letters 📷 The people in front have blurry, merged hands Share before they delete it! 📷 The photo is sharp and high resolution

  2. Quick checks

    Which check helps here?

    1. You want to know where the wave photo first appeared.

      • Run a reverse image search (Answer)
      • Read more comments under the post

    2. A photo comes with Content Credentials saying it was made with an AI tool.

      • That's useful evidence about how it was made (Answer)
      • Ignore it, labels mean nothing

    3. A teacher runs an essay through an AI text detector, and it says "92% AI".

      • That proves the student used AI
      • It's a hint, not proof. A conversation is fairer. (Answer)

  3. Myth or fact?

    Tap each card to see the truth.

    ✋ "AI always gets hands wrong."
    Myth. Older tools often did. Newer ones often don't, so a perfect hand proves nothing.
    🤖 "A detector can tell for sure."
    Myth. Detectors for images and text make mistakes in both directions. Use them as one clue among several.
    🏷️ "Some AI content is now labelled."
    Fact. Some tools and platforms add labels or Content Credentials, and EU rules require deepfakes to be disclosed. But labels can be missing or removed.
    📍 "The source matters more than the pixels."
    Fact. Who posted it first, and do trusted outlets confirm it? That's often the strongest test.

Remember

No single test is proof. Combine clues, check labels, and find the original source.

No single test is proof. Check where it came from.

Key facts

  • The US standards agency NIST found that provenance tracking, watermarking and detection each have technical limits, so no single method can reliably flag all synthetic content. · Source: NIST, AI 100-4 Reducing Risks Posed by Synthetic Content, 2024
  • Under the EU AI Act, people who publish deepfake images, audio or video must disclose that the content was artificially generated or manipulated. These transparency rules apply from 2 August 2026. · Source: EU AI Act, Regulation (EU) 2024/1689, Article 50
  • Content Credentials are an open standard from the Coalition for Content Provenance and Authenticity (C2PA) that attaches tamper-evident information about how a file was created and edited.
  • Visual tells such as extra fingers or garbled lettering are becoming rarer as image generators improve, which makes checking the source more important than inspecting pixels.

Updated:

Questions people ask

Do AI image detectors work?

Sometimes, but not reliably. They can miss AI images and wrongly flag real ones, so use them as one clue alongside the source, labels and a reverse image search.

How can you tell if text was written by AI?

Often you can't tell for sure. AI text detectors make mistakes, so look at context instead: drafts, sources, the writer's usual style, and whether the facts check out.

What are Content Credentials?

They are tamper-evident information attached to a file that records how it was created and edited, including whether an AI tool was used. A missing credential doesn't prove an image is real.

What is a reverse image search?

You upload an image or its link to a search engine, which shows where the same or similar images appear online. It helps you find the original and any fact-checks.

Related terms