How to Spot a Deepfake Before You Share It
AI-generated images, video and audio can now look convincing enough to fool people at a glance. Verification therefore has to become a habit, not a last resort.
smartphone misinformation / fact checking
What happened?
AI-generated images, video and audio can now look convincing enough to fool people at a glance. Verification therefore has to become a habit, not a last resort. NIST says modern deepfakes can be created cheaply and quickly and that detection systems can perform worse in real-world conditions than in controlled tests.
Key points
- NIST says modern deepfakes can be created cheaply and quickly and that detection systems can perform worse in real-world conditions than in controlled tests.
- UNESCO says the rise of AI-generated content makes critical thinking and media literacy increasingly important.
- INTERPOL has warned that generative AI, including deepfake video, audio and images, is being used to support impersonation and fraud.
- CISA describes deepfakes as synthetic media created or manipulated to mislead viewers.
- A suspicious image or video should be checked against independent sources before it is treated as evidence.
What we know
- NIST says modern deepfakes can be created cheaply and quickly and that detection systems can perform worse in real-world conditions than in controlled tests.
- UNESCO says the rise of AI-generated content makes critical thinking and media literacy increasingly important.
- INTERPOL has warned that generative AI, including deepfake video, audio and images, is being used to support impersonation and fraud.
- CISA describes deepfakes as synthetic media created or manipulated to mislead viewers.
- A suspicious image or video should be checked against independent sources before it is treated as evidence.
What officials/people involved say
The common thread across the guidance is simple: do not rely on your eyes alone. Context, source provenance and independent confirmation are often more reliable than trying to detect a tiny visual flaw.
Background
Deepfakes are only one form of manipulated media. A real video can also be cropped, slowed, sped up, re-captioned or placed in a false context. That is why 'it looks real' is not enough.
Why it matters
False information can influence elections, markets, reputations and personal safety. A single viral post can travel faster than a correction, so the best intervention is often to pause before sharing.
What happens next?
Use a five-step check: identify the original source; search for independent reporting; check the date and location; look for the full context; and ask whether a primary source confirms the claim. If the story is extraordinary, demand stronger evidence.
Sources & further reading
Every claim above can be traced to the documents below.
Author
Insight Media Editorial Desk — original reporting, explainers, analysis and practical guides, researched against primary documents and credible independent reporting. Developing stories are updated when significant new verified information becomes available.