YouTube's recommendation system in 2026 is not a mystery, but it is multi-factor: whether a video starts sessions, whether it earns Browse placement, how it performs in Suggested, click-through relative to its impressions, average view duration as a percentage, how fast engagement arrives, whether people share it, and whether it converts subscribers. The Algorithm Score Predictor evaluates a described or published video against those eight signals, plots them on a radar, and ranks what to fix first.
This guide explains the eight signals, why the radar matters more than the total, and how to use the prediction before publishing.
The eight signals
Session start: is this a video people open YouTube for, or one they stumble into? Browse eligibility: does the packaging work on the home feed? Suggested score: does it pair well with adjacent videos? CTR threshold: will it clear the impressions-to-clicks bar for its niche? AVD assessment: will it hold viewers above the duration-band median? Engagement velocity: will likes and comments arrive early? Share rate: does the content carry social currency? Subscriber conversion: does it promise more? Each is scored 0–100 from the metadata, structure and topic you provide.
Why the radar matters
Two videos can total 65 with opposite shapes: one strong on CTR and weak on AVD (clickbait risk), one strong on AVD and weak on CTR (under-packaged). The fixes are opposite too. The radar shows the shape; the priority action list picks the lowest signal with the largest weight in your traffic mix.
Before publishing
Run the predictor on your draft title, thumbnail concept and structure. A low session-start score suggests a topic tweak; a low CTR score suggests packaging changes; a low AVD score suggests structural edits. Fixing these before upload is cheap; after upload it is expensive.
Tips
- Describe the thumbnail concept and the first 30 seconds; those drive four of the eight signals.
- Run a published video URL to see the actual metadata scored; compare with the pre-publish prediction to calibrate.
- Aim for no signal under 50 rather than a high total.
Common mistakes
- Optimising CTR at the expense of AVD — High click, low hold is the pattern YouTube punishes most.
- Ignoring session start — It is the highest-weighted 2026 signal and the least understood.
- Treating the score as a forecast — It is a structural assessment. Actual performance depends on execution.
Worked example
Input: "Budget grinder comparison; thumbnail: three grinders + "$40 WON"; opens with the winner then explains."
What the report returned: Total 71. Radar: CTR 82, AVD 74, Session start 48, Browse 70, Suggested 77, Velocity 66, Share 55, Sub conversion 62. Priority: session start — reframe title toward a problem people open YouTube to solve ("Why your cheap grinder ruins espresso").
Frequently asked questions
What is session start?
Whether a video is the reason a viewer opened YouTube (from search, notification or external link) rather than something they found mid-session. Videos that start sessions are weighted heavily in 2026.
Can it score a published video?
Yes, by URL. It reads the public metadata and structure.
How is this different from the 2026 Algorithm Checklist?
The Checklist is a 25-item pass/fail. The Predictor scores eight weighted signals and shows the shape. Use the Checklist for a quick pre-publish pass and the Predictor for diagnosis.
What score is "good"?
Above 70 with no signal under 50 is a strong position for most niches.