All guides
Plain guides to AI video: what deepfakes are, how platforms label AI clips, what Content Credentials and watermarks prove, how the scams work and what the law says. Each guide links its primary sources and shows the date it was last checked. New guides appear here on the day they are published, newest first in each topic.
If you have a clip in front of you right now, start with the free AI video detector.
Start here
- How to Tell If a Video Is AI
How to tell if a video is AI in nine checks, from hands, text and physics to labels, the original source and a free detector, each with an example frame. - How to Spot AI Video
Learn how to spot AI video by eye: warped hands, melting text, impossible physics, odd blinking and drifting light, with annotated frames for each tell. - The Sora Watermark
What the Sora watermark looks like, why a missing mark proves nothing, how content credentials work, and how to check a clip when the mark is cropped out. - SynthID Explained
SynthID explained: what Google's invisible watermark is, where you can check for it, what a result means, and what to do when a video has no mark at all. - Best AI Video Detectors, Compared
The best AI video detectors compared on what they show you: free limits, clip length, frame evidence and signup, in a dated table with the same fields.
How the guides are organised
The guides are grouped by the question you are trying to answer, not by the date they were written. The five guides under Start here cover the basics: the checks you can run by eye, the visual tells, what a Sora watermark proves, what SynthID is, and how the free detectors compare.
The topics below them go deeper. Spot it yourself covers the tells you can find with the pause button. Verify viral video covers finding where a clip came from. Platform labels and policy explains what the big apps show when a clip is marked as AI. Provenance and content credentials covers C2PA and what a file's own records can prove. Generators and watermarks looks at specific AI video tools. Scams and fraud, Law and policy, Families and schools, Deepfakes explained and Detection tools cover the rest. A topic appears here once its first guide is published, and inside each topic the newest guide comes first.
Using a guide with the check
The check and the guides are meant to be used together. The free check reads the clip's file in your browser and reports any Content Credentials, AI labels or AI tool names it finds, with still frames to look at. Many clips carry no label, so a result of "No AI signs found" is where the guides take over. Use the visual tells guide on the frames the check shows you, then follow the source steps in how to tell if a video is AI. If the check does find a label, the guides on Content Credentials and watermarks explain what that label does and does not prove.
How guides are dated and kept current
Every guide shows the date it was last updated, and new guides also show the day they were published. Facts about platforms, labels and tools link to the page that states them, usually the company's own help page or documentation. When a platform changes a label or a tool changes what it offers, the guide is revised and its date moves. The date does not move without a real change. Comparisons say when each tool's details were checked.
Who writes the guides
The guides are written and kept current by Tessa Ostrowski, who edits the site. They are written for someone who has just seen a suspicious clip and wants a straight answer, so each one opens with a short answer and then shows its working. No guide quotes accuracy percentages or promises certainty, for this check or any other. If you spot an error or a stale fact, the contact page explains how to report it, and the about page sets out the rules every guide follows.