A limited result usually means the detector did not have enough accessible, usable evidence for a fuller screening assessment. It is not a finding that a video is real, fake, safe, or deceptive. Check what was available, what was missing, and whether you can verify the original source or file before acting on it.
Quick answer
Use this sequence:
- Confirm that you used one direct, public video URL with the link-check guide.
- Read the result state, confidence, checked signals, unchecked signals, and limitations in that order.
- Find the earliest available full version and compare its caption, date, account, and context.
- Upload an original file only when you lawfully have the right to analyze it.
- For a high-stakes claim, use independent sources and qualified human review—not a detector score alone.
What “limited signals” means
| What you see | Likely explanation | Best next step |
|---|---|---|
| Limited public-link result | The page or video stream was unavailable or restricted. | Check the direct URL and find the original source. |
| Low confidence | The clip is short, blurry, compressed, cropped, or incomplete. | Treat it as inconclusive; seek a higher-quality source. |
| Audio unavailable | The post has no usable audio, or access is limited. | Do not infer speech or voice conclusions. |
| Visual clues unavailable | There were too few readable frames or severe processing. | Use an original file only with a right to submit it. |
| Source context missing | The link does not expose enough post context. | Compare independent uploads and account history. |
AI Video Detector is a screening aid. A public post may play in an app while still not exposing enough accessible media for a fuller link review. A limited result is a transparent limitation, not evidence against AI generation.
Why a public video may not be usable
Start with a direct URL to one public video post—not a profile, search page, channel, hashtag, photo carousel, Story, or share page that cannot resolve. Private accounts, removed posts, login walls, age gates, regional limits, and platform controls can prevent a scan from accessing the same material that you can view in an app.
Platform-specific paths have different limits. Review the supported public-link guidance for Instagram Reels, TikTok videos, and YouTube links before treating an unavailable scan as meaningful.
Social-video processing can reduce evidence
Reposting, screen recording, heavy compression, speed changes, filters, subtitles, cropping, and replacement audio can alter frame detail and timing. They can make genuine footage look odd and can also hide genuine signs of manipulation. Treat a strange-looking frame as a lead to investigate, not proof.
Find the earliest available full version before drawing a conclusion. A useful review looks for patterns across frames and makes clear what it could not inspect.
Provenance can be absent without proving anything
When present, Content Credentials can offer provenance information about an asset’s origin or edit history. They are helpful source evidence—not a universal truth test. Their absence does not establish AI generation, because metadata can be absent from the start or removed by editing and reuploading.
Use provenance alongside the account, original recording, full context, and independent reporting. See C2PA’s overview of Content Credentials and Meta’s explanation of AI-content-labeling limits.
Confidence is not likelihood
Likelihood describes how strongly checked signals point toward possible AI generation or manipulation. Confidence describes how much usable evidence the particular review had. A high likelihood with low confidence deserves restraint; a low likelihood with low confidence does not authenticate a video.
Read the evidence report guide before using a result in a decision. The report does not determine copyright, policy compliance, the factual truth of a caption, or a person’s intent.
When an original file is a better next step
Use clearer evidence when available
Upload the original only when you are entitled to analyze it.
An original file can provide a clearer frame sequence, duration, and audio track than a restricted social link. It still is not proof: authentic footage can be paired with a false caption, and manipulated video can include authentic source footage.
Read deepfake screening guidance- Copy the direct URL for one public video post.
- Record its caption, account, date, and claimed event before it changes.
- Look for the earliest available full-length version and compare independent reporting.
- Read every limitation before focusing on a score.
- Escalate identity, safety, money, reputation, news, and legal claims to qualified human review.
Frequently asked questions
Does a limited result mean the video is real?
No. It means the available input did not support a fuller review. It does not authenticate the video or establish that it was generated.
Why can a public link have less evidence than an upload?
A public page can restrict direct media access or show a processed version. An original file can preserve a clearer frame sequence and audio track when you have the right to analyze it.
Can I share a video after a low-likelihood result?
Not on that result alone. Verify the source and the factual claim first, especially where a video could affect safety, identity, reputation, or money.
Bottom line
Limited signals are a transparent limitation, not a verdict. Use the report to decide what you need next: a direct public link, a better source, an original file you are entitled to analyze, or human verification. Pause before sharing when evidence is incomplete.
Sources and update notes
Reviewed September 24, 2026. Product behavior and social-platform access can change; verify official guidance before relying on platform-specific steps.