Video guides

How to use
KweliAI

Two short videos showing exactly how to paste a social media link and get a real-or-AI verdict in seconds. No downloads. No technical knowledge required.

How to use KweliAI — detecting AI images from social media

This video shows the complete workflow: finding a public post on X or Reddit, copying the post URL, pasting it into KweliAI, and reading the verdict. The scan takes under 10 seconds and requires no account for the first try.

Video transcript

KweliAI is a free tool that checks whether images and videos on social media are real or AI-generated. You don't need to download anything — just paste the post link.

Go to the post on X or Reddit that contains the image or video you want to check. Copy the URL from your browser address bar. Then go to kweliai.com, paste the link into the scan box, and press Scan.

KweliAI extracts the media directly from the post and runs it through three detection models simultaneously — one for AI-generated images, one for deepfakes, and one for AI-cloned audio in video. The result arrives in under 10 seconds.

You receive a clarity score from 0 to 100 — where 0 is definitively real and 100 is definitively AI-generated — along with a verdict of Real, AI-generated, or Deepfake, and a confidence rating of High, Medium, or Low.

Free users get 5 scans per day on X and Reddit. No account is required for the first scan.

How to use KweliAI — understanding the verdict and confidence score

This video explains what the clarity score means, how to interpret high versus medium confidence results, and what to do if the result is inconclusive. It also covers checking multiple posts from the same account and using the Pro plan for TikTok, YouTube, Instagram, and Facebook.

Video transcript

Once KweliAI finishes scanning a post, you receive three pieces of information: a clarity score, a verdict label, and a confidence rating.

The clarity score runs from 0 to 100. A score close to 0 means the content shows strong signals of being camera-captured. A score close to 100 means the content shows strong signals of AI generation. Scores in the middle — roughly 30 to 70 — are ambiguous and come with a medium or low confidence rating.

The confidence rating tells you how certain the detection models are about the result. High confidence means the signals were clear and consistent across all three models. Medium or low confidence means the image may have been heavily compressed, edited, or processed in ways that reduce detection reliability.

If you receive a medium or low confidence result, scan additional posts from the same account rather than treating a single result as definitive. A pattern of AI-generated verdicts across multiple posts is much stronger evidence than a single result.

Pro users can scan TikTok, YouTube, Instagram, and Facebook posts in addition to X and Reddit, with unlimited scans per day. Both free and Pro accounts use the same detection models.

How to detect AI-generated content — step by step

1
Find the post you want to verify
Go to the specific post on X, Reddit, TikTok, YouTube, Instagram, or Facebook. The post must be publicly visible. Open the individual post page — not the user's profile page.
2
Copy the post URL
On desktop, copy from the browser address bar. On mobile, tap the Share button on the post and select Copy link. Make sure you have the URL of the specific post — not a search result or profile page.
3
Paste it into KweliAI and press Scan
Go to kweliai.com. Paste the URL into the scan box and press Scan. X and Reddit work on the free plan. TikTok, YouTube, Instagram, and Facebook require a Pro account at $12/month.
4
Read the verdict
You receive a clarity score (0–100), a verdict (Real, AI-generated, or Deepfake), and a confidence rating (High, Medium, or Low). High-confidence results are the most actionable. For medium or low confidence, scan additional posts from the same account.

Try it right now

Free for X and Reddit · No account needed for first scan

Paste a link →
Written by Mohammad Nyundo, Founder of KweliAI  ·  Last updated July 2026