How Scoring Works

Last updated: July 2026

TruthScore gives every YouTube video a score from 0 to 100. A higher score means fewer warning signs were found. The score is built from up to three separate signals, and the report shows which signal each finding came from, so you can judge the evidence yourself.

Important: TruthScore estimates risk. It does not prove that a video or creator is legitimate or fraudulent. Treat it as a prompt to look closer, not as a final verdict.

The three signals

Signal	What it looks at	Weight
📊 YouTube Signals	Title, description, engagement, dislikes, comments, and channel history	55%
🌐 Web Research	Scam reports, complaints, and fact-checks found elsewhere on the web	30%
🎙️ Video Transcript	What is actually said in the video, when captions are available	15%

If a signal isn't available for a video, for example because the creator turned captions off, it is left out and the remaining signals share its weight proportionally. A missing signal never counts against a video.

📊 YouTube Signals

This starts from a neutral baseline and moves up or down based on what it finds:

Title and description: unrealistic income or fast-result claims, known scam phrases, urgency or FOMO language, and heavy use of hype emojis.
Engagement: likes and comments relative to views. Very low engagement on a large view count can point to purchased views or bot traffic.
Dislike ratio: YouTube hides public dislike counts, so this uses estimates from the third-party Return YouTube Dislike project. It is an estimate, not an exact figure.
Comments: whether comments are disabled on a popular video, and whether the comments that exist are mostly positive, negative, or explicitly call the video a scam.
Affiliate and sales pressure: the strongest warning needs both a tracked link and urgency or pressure language together. A link shortener on its own is common and only counts as a minor note.
Channel trust: channel age, subscriber count, and upload history. A brand-new, small channel with a sudden viral video is treated as a notable warning sign.

🌐 Web Research

For every video, TruthScore uses Google's Gemini AI with Google Search to look for scam reports, complaints, refund disputes, and fact-checks tied to the creator or to specific claims in the title. Each finding is listed with where it came from. If the search turns up nothing notable, the report says so and the signal stays neutral. No results does not mean a creator is trustworthy. It only means nothing has been found yet, which is common for newer channels.

To do this, the video's title, channel name, and description are sent to Google's Gemini API.

🎙️ Video Transcript

When a video has captions available, TruthScore also checks what is said for unverifiable income claims, fake urgency, pressure tactics, and hidden sales pitches. Many creators disable transcripts, so this signal is often missing. That is why it carries the smallest weight.

Reading the result

75 and above: Likely Legit. Few or no warning signs were found.
Below 75: the lower the score, the more warning signs were found and the more carefully you should check the claims yourself.

Findings are color-coded: red for strong warning signs, yellow for things worth checking, blue for context, and green for positive signs.

Where TruthScore can be wrong

Keyword checks are blunt. A legitimate creator can trigger a warning, and a careful scammer can avoid every keyword.
Good numbers can be faked. Engagement, comments, and subscriber counts can be bought or manipulated.
Web research is limited to what is findable. New or small scams may have no public reports yet.
AI can make mistakes. The web research and transcript checks use an AI model, which can misread context.
Results are cached. A video's result may be reused for up to 6 hours, so very recent changes may not show up yet.

Never make a financial decision based on a score alone. If a video asks for money, check the seller independently first.

Analyze a video · Privacy Policy