# Spotting AI fakes · AI in Bed

Beginner level · 5 min · https://www.aiinbed.com/learn/beginner/spotting-ai-fakes

*AI can make convincing photos, voices and articles. How to slow down, check the source and not get fooled.*

Generative AI can now produce a realistic photo of an event that never happened, clone a voice from a few seconds of audio, or write a fake news article in any style. A fake picture or video of a real person is called a **deepfake**.

### Looking for glitches isn't enough

Extra fingers and melted text used to be giveaways. Newer models rarely make those mistakes, so "it looks real" proves very little.

The better question is not "does it look fake?" but "where did it come from?"

### The three-question check

**Who posted it first?** A named news organisation, or an anonymous account created last month?

**Is anyone else reporting it?** Real big events get covered by several independent sources within hours.

**Does it want you angry or scared?** Content designed to make you share fast deserves the slowest look.

**Quick check:** A shocking video of a politician is going viral. What's the most reliable first check?

A. Zoom in and look for blurry edges
B. See whether reputable news outlets are reporting it and who posted it first
C. Count how many likes it has
D. Ask a chatbot whether it's real

Answer: B. Provenance and independent confirmation beat pixel-peeping. Likes measure popularity, not truth, and detectors are unreliable.

### Voice clones and scams

Scammers use cloned voices of family members asking for money urgently. Agree a family code word, and if a call feels off, hang up and call back on a number you know.

### Labels and watermarks

Some tools now add invisible watermarks or signed "content credentials" (the C2PA standard) that record how an image was made. They help when present, but their absence proves nothing.

> If something online makes you instantly furious, that's the moment to wait ten minutes before sharing.

**Key takeaways**

- Realism is no longer evidence. Ask where content came from.
- Check who posted first and whether independent sources confirm it.
- Use a family code word against voice-clone scams.

**Go deeper (free sources)**

- [News Literacy Project](https://www.newslit.org/) by News Literacy Project
- [C2PA: Content Credentials](https://c2pa.org/) by Coalition for Content Provenance and Authenticity
- [Ethics of AI (free course)](https://ethics-of-ai.mooc.fi/) by University of Helsinki