Free Sora Video Checker

What this check can tell you about a suspected Sora clip

It can tell you whether the file says it was made with AI. It cannot tell you that Sora, or any other specific tool, made it.

That is a deliberate line, not a gap we are hoping you will not notice. Many video models produce similar looking footage, clips get edited and re-encoded on their way to your feed, and a confident "this came from tool X" is a claim no free checker should be making. What you get instead:

• A verdict: AI generated, Likely AI, Likely real, or No AI signs found.
• The evidence behind it: whether the file carries Content Credentials or an AI label that says AI, an AI tool name in its metadata, or nothing the check could read.
• Still frames from across the clip, so you can look for the tells yourself.

If you suspect Sora specifically, the useful extra evidence is a visible watermark, Content Credentials in the file, and where the clip was first posted. Those are covered below. Keep in mind that a free check is one signal among several, and "No AI signs found" never proves a clip is real.

Worth knowing for context: OpenAI's own Creating with Sora safely page now carries a note that, as of April 26, 2026, the Sora product is no longer available. Clips made before then are still out there and still get reposted, so the question has not gone away.

The visible watermark and why cropping matters

On the same page, OpenAI says every video generated with Sora included both visible and invisible provenance signals, and that many outputs also carried visible, dynamically moving watermarks that include the name of the creator.

A visible mark is helpful when you can see it. Its weakness is that it lives in the picture itself. That has two practical consequences:

1. It can be removed. Cropping the frame, zooming in, blurring a corner, or laying a caption or sticker over it can take the mark out of view. A clip with no visible watermark tells you very little about where it came from.
2. It can be faked. A watermark is just pixels. Someone can paste a convincing looking mark onto real footage to make it seem AI made, or onto a clip from a different tool. The mark alone is not proof either way.

The Sora watermark guide goes deeper on what the visible mark looks like and how to read it.

Content credentials on Sora videos

OpenAI also says all Sora videos embed C2PA metadata, an industry standard signature. In plain terms, Content Credentials are information attached to the file that records where it came from and how it was made. The C2PA specification defines how that record is stored and signed, and an AI made file usually marks its source with the IPTC digital source type for generative AI. Those records are exactly what the check above looks for.

The catch is how easily they disappear. OpenAI's help article on provenance signals in OpenAI generated content says Content Credentials can be lost through everyday actions such as taking a screenshot or converting a file, and that platforms, editing tools and file conversions may remove metadata. A clip that has been downloaded, trimmed and reposted a few times may carry none at all.

The same article lists video as a content type covered by Content Credentials, and describes openai.com/verify as a way to check supported image or audio files. Video is not on that list, so for a suspected Sora clip the practical options are looking for the visible mark, asking whether the file you have is close to the original, and reading the file with a checker like this one.

For a broader method that does not depend on any one tool, see how to tell if a video is AI. If the clip you are wondering about looks more like Google's style, the Veo video checker covers that side, including SynthID.

Suspect a Google tool instead? The Veo video checker covers the visible mark and SynthID side of the question.

Saw the clip on TikTok first? The TikTok AI video checker covers saving it and reading the platform label.