What Gistil shows you on a YouTube page, and why each part is there

Gistil puts one panel next to a YouTube video. Everything in it answers a question you have in the ten seconds before you decide to press play, and nothing in it is there to be impressive.

This page is a walk through the panel, block by block, with what each block claims and where each one stops being trustworthy.

The verdict, and one line of why

At the top: one of three words. Worth it, worth skimming, or skip — and a single sentence in our own words explaining which of the numbers below produced it.

The sentence matters more than the word. A verdict with no reason is a thing you either trust or ignore, and neither is useful. A verdict that says why is a thing you can disagree with, which is the only kind worth publishing. About 20% of measured videos end up as a skip.

Three numbers

Value density. What share of the runtime carries information rather than orientation and restatement. It runs 0–100 and the library average sits at 59.

Padding share. The proportion of the runtime that restates, recaps or sets up rather than adds. The median measured video sits at 40%. This is the number people find most immediately useful and the one most easily misread — it does not mean that share of the video is worthless, it means the same material could have been delivered in substantially less time.

When it starts delivering. The second at which the video stops setting up. The median across the library is 1:00, and one video in ten does not arrive until 3:46. This is the most practical of the three: if you already know why you clicked, that is where to start.

All three are defined at length, along with the cases where each one is the wrong thing to look at, in what value density means.

The receipts

A number with no evidence under it is an assertion. So each of the three carries something you can check against the video itself.

Examples of the restating we counted, so a padding figure of 45% is a claim you can verify against the video rather than a number to take on faith.

A title check. How closely the title describes the video that follows, with one line saying what the gap was. Across the library 71% of titles land in the range we would call accurate.

A labelled map of the video’s sections, each with what it is doing and how it scored — the opening, the body, the digression, the sponsor read, the close. It is worth knowing what this is not: a section list is not a map of value unless the sections carry measurements, which is why ours do.

The things that are only sometimes there

Four blocks appear when there is something to say and are simply absent when there is not. No placeholders, no dashes, no “not applicable”.

What you need to already know. Some videos assume a background, and finding that out four minutes in is the second most common way a video wastes your time.

A note when somebody is being sold to you, and whether that was disclosed — shown only when the commercial intent is loud enough to be worth flagging. A flag on every video would train you to stop seeing it.

Rare, loud flags for the small number of videos that show synthetic-sounding assembly or a clickbait pattern strongly enough to name.

A warning about our own confidence, on the small share of videos where our map of the video is a hint rather than a measurement. The normal case is not announced, because announcing it on every video is noise; the exception is.

What the panel deliberately does not do

It does not rank the channel. We publish measurements of individual videos and aggregates over a shelf. There is no list of worst channels and no verdict on anybody’s body of work — a channel with one thin video is a channel that made one thin video.

It does not replace the video. The panel answers whether to play and where to start. What the video actually says is the video’s job, and a summary is a poor substitute for it whoever writes the summary.

It does not check whether the video is right. We measure how much a video says per minute, not whether what it says is true. A densely argued video about a claim that does not hold up scores well on every number in the panel.

It does not touch YouTube. The player, the recommendations and the page are unchanged; the panel sits alongside them.

Where the numbers come from

Every figure on this page except the panel description is read live from the library on each visit, so nothing here can drift out of step with the product. The library currently holds 3,486 measured videos across 15 subjects, last recalculated 2 September 2026.

Scores are recalculated as the measurement behind them improves, which means a number you saw last month may have moved. That is why every page on this site carries the date its figures were last recalculated, and why the dated reports on the index freeze their figures instead: a measurement people cite should not change under them.