YouTube
What Google can anchor on in this video
In one sentence
The YouTube video check reads one video through public surfaces and reports what a search result can anchor to: whether chapters are authored and recognisable, whether captions are uploaded or auto-generated, whether the description carries prose and lists, whether title, description, audio and declared language agree, and whether the Knowledge Graph placed the video under a specific topic. Each row names what it reads.
Chapters, captions, description and language, checked against what a search result can actually anchor to. Measured from public surfaces; the scores themselves are not visible and are not guessed.
What a search system can anchor on in a video
A video result does not just rank a whole video: it lands a viewer on a passage — a chapter, a moment, an answer — and to do that a search system has to work out what each part of the video is about. Everything it has to work with is the video’s public metadata: the chapters you authored, whether the captions are yours or auto-generated, what the description actually says, whether the title, the audio and the declared language agree, and which topics the Knowledge Graph has placed the video under. This check reads those surfaces and reports them. What it never does is claim to know the weighting: the scores are internal, and nothing here guesses them.
The 14 checks, and what each one reads
Each row on a result names what it reads, what the video currently has, and what would pass. Rows are judged only where the input is measurable: a Short with no caption track is not marked down for its captions, it is marked not measurable, and the tally says so.
| Check | What it reads | Read from |
|---|---|---|
| Chapters | Chapter anchors in the description | Description timestamps and player markers |
| Chapter spacing | Spacing between chapters | Description |
| Chapter labels distinct | How distinct the chapter labels are | Description |
| Chapter labels vs title | Chapter labels against the title | Data API |
| Chapter labels spoken in the video | Chapter labels against the spoken audio | Browser: transcript |
| Chapter labels in the description body | Chapter labels against the description body | Description |
| Generic chapter labels | How descriptive the chapter labels are | Description |
| Captions | The caption track and its confidence | Data API flag; browser tells manual from auto |
| Language agreement | Title, audio and declared language | Data API; browser adds the spoken language |
| Topics spoken but absent from the description | Spoken topics present in the description | Browser: transcript |
| Description carries a list or steps | Lists and steps in the description | Description |
| Description length | Description depth | Description |
| Topic categories assigned | Knowledge Graph topic placement | Data API topicDetails |
| Family safe | Advertiser-safe classification | Data API |
Two vantages: what a server can read, and what only a browser can
From this Worker
Title, description, tags, duration, declared language, topic categories, the caption flag, view and like counts, thumbnail sizes, and chapter timestamps parsed from the description. Read through the YouTube Data API and the public embed endpoint. The watch page itself refuses datacentre clients, so nothing on it is scraped from here.
From your browser
Whether captions are manual or auto-generated, the transcript text, storyboard frames and the player’s own chapter markers exist only inside a real, visible browser session. The extension, or a check run from your own tab, posts that observation back; the result then merges it and the transcript-based rows switch on.
The two halves are kept apart on purpose. A result never claims a transcript it did not read, and a channel tally counts caption and transcript rows only for the videos a browser has actually observed.
Stated as not measurable, never guessed
- Click and watch data (queries, unique views, watch time and retention) exist only inside Google.
- The internal ranking scores themselves: no public surface exposes them.
- Frames and on-screen text are measured only when a browser has captured the storyboard; a video nobody has observed carries no frame rows at all.
A check that filled those rows with a guess would be selling the number, not measuring it.
A channel with 86 uploads, rewritten and measured
Where it sits in the plans
| Plan | YouTube | Also |
|---|---|---|
| Free | Any single video, as often as you like | The full site scan |
| Watch | 1 channel: every upload tallied, plus the site tie-in | 3 domains under record |
| Studio | 3 channels, and the transcript layer through the browser extension | 10 domains |
| Agency | 10 channels, with a per-video change record | 40 domains |
| Network | Every channel you operate | 150+ domains |
Prices and what each tier holds back are on the pricing page. Nothing a plan unlocks changes a measurement.
Read next
Check a whole channel · Does your site carry your videos? · The case study · The browser extension · The API