How to spot a deepfake: AI-made images, videos and audio
Warning signs, official tools such as SynthID (Google) and Content Credentials (C2PA), and what to do if you receive a deepfake or a cloned-voice scam. With cited sources.
A deepfake is content created or altered with AI that imitates a real person. Checking whether it “looks odd” is no longer enough: today’s generators are very realistic.
Warning signs
- Urgency and emotion: “Mum, I lost my phone, send me money now”.
- Unknown origin: a forwarded video with no source, or a brand-new account.
- Inconsistent details: hands, text, reflections, shadows or lips out of sync with the voice. Careful: their absence proves nothing.
Tools that actually help
| Tool | What it does |
|---|---|
| SynthID (Google) | Invisible watermark in images, video, audio and text generated with Google AI. You can upload the file to Gemini and ask whether Google AI created or edited it. |
| Content Credentials (C2PA) | Open standard that works like a “nutrition label”: it shows a piece of content’s origin and edits when the creator includes it. |
None detects everything. Use them as clues, not verdicts.
What the law says
According to the European Commission, under the AI Act deepfakes must be clearly and visibly labelled, and generative AI providers must make generated content identifiable. The transparency rules apply from August 2026.
If you receive one
- Don’t share it.
- If you’re asked for money or data, hang up and call the person or organisation on a number you know. Agree a family code word.
- Get help (in Spain, INCIBE’s 017) and, if there’s a crime, report it.
Frequently asked questions
If SynthID finds no mark, is the content real?
No. SynthID only detects content made or edited with Google tools that carry that mark. Other generators don’t add it.
Where can I get help if I’ve been scammed?
In Spain, INCIBE’s Cybersecurity Helpline, 017, free and confidential. If a crime was committed, report it to the police.
Glossary · Safety and good use
Keep reading
Sources
- Google SynthID — Google DeepMind
- Content Credentials (C2PA) — Coalition for Content Provenance and Authenticity
- Ley de IA: riesgo de transparencia — Comisión Europea
- Línea de Ayuda en Ciberseguridad 017 — INCIBE
Change log
- : First published with the SynthID and C2PA documentation checked that day.
Educational content, prepared with the help of AI tools, with sources cited. It is not a substitute for professional advice. Tools change fast: if you spot something out of date, write to us at contacto@cognenza.com. How we work.