How videos are found: search, suggested, browse, external
YouTube Analytics groups where views come from into traffic sources, and four of them account for most of what video SEO can influence. YouTube search is a query typed into YouTube; the results are a ranked list of videos, channels and playlists, personalised by the viewer's history. Suggested videos are the recommendations beside and after a video, driven by what a viewer has watched and by which videos tend to be watched together. Browse features cover the home feed, the subscriptions feed and other surfaces where YouTube offers videos without a query. External is traffic from outside YouTube: Google Search results, embeds on websites, links in email and social posts.
The traffic sources guide explains how to read each of these in the reports. For SEO the point is that each surface is won differently. Search is won by relevance to a query and by satisfying the people who searched it. Suggested is won by being the natural next video after something the viewer just watched, which is partly a topic question and partly a retention question. Browse is won by having an audience that returns. External is won by being embedded, linked and shown in Google's own results. A video that is built only for search can rank and still never be suggested; a video built for suggestion can travel and still be invisible to a query. The YouTube marketing hub describes the surfaces at channel level; this section is about winning them video by video.
What YouTube documents about ranking signals
YouTube publishes a description of how its search works, and it names three groups of signals. Relevance is how well the title, description, tags and the content of the video match the query. Engagement is aggregated signals from viewers, such as watch time for a particular query, that indicate which videos people found valuable for it. Quality is signals that help identify expertise, authoritativeness and trustworthiness for a topic, applied with more weight on subjects where accuracy matters. Search results are also personalised: two people typing the same query see different orders depending on their history.
For recommendations, YouTube explains that the system draws on the viewer's own watch and search history, their subscriptions, what people with similar histories went on to watch, and direct feedback through likes, dislikes, "not interested" and surveys. It describes the goal as predicting satisfaction rather than clicks, and it says explicitly that clicking a video and leaving quickly is a signal against it.
What YouTube does not document is the weight of any signal, the effect of any particular metadata field, or a threshold at which a video is "ranked". Claims that a given keyword density, tag count or upload time moves rankings are not supported by the help centre and should be treated as practitioner folklore. The safe reading is that metadata earns the chance to be shown, and viewers decide what happens next. That is also the reading that keeps the core guides in this section honest: the titles, descriptions and chapters guide covers the text that establishes relevance, and the thumbnails and click-through guide covers the image that earns the click, with click-through rate (CTRShare of impressions that result in a click.) read together with retention rather than alone.
Keyword research for video
Video keyword research answers the same question as web keyword research: what do the people this channel wants to reach actually type? The differences are the sources and the intent mix, because people search YouTube for things they want to watch rather than read. The intent classification guide sorts queries by what the searcher wants; on YouTube the useful split is between queries that want a demonstration, queries that want an explanation, and queries that want a comparison, because each wants a different video.
| Source | What it gives | How to use it |
|---|---|---|
| YouTube search suggestions | The completions YouTube offers as a query is typed, drawn from real searches | Type the seed term and each letter after it; collect the phrases and group them by what the viewer wants |
| Research tab in YouTube Analytics | Searches made across YouTube and by the channel's own viewers, with gaps where few good results exist | Look for the gaps in the channel's own subject; they are the queries most worth a video |
| Google Trends with the YouTube Search option | Relative interest in a term on YouTube over time and by region | Check seasonality before scheduling; compare phrasings of the same idea |
| Competing channels | The titles that earn views on the same subject | Read the pattern of words, not just the topic; note which queries nobody has answered well |
| Google Search results (SERPThe page of results a search engine shows.) | Which queries show a video carousel or video results | A query that already shows videos is one Google has decided wants video; it can send external traffic too |
| Comments and community posts | The questions viewers ask in their own words | Each recurring question is a search someone makes before asking |
| The business's own search and support logs | What customers ask once they have arrived | Care-stage videos that cut support load and answer searches existing customers make |
Group the phrases by intent and by the video that would satisfy each, and give each video one primary query and a small set of secondary ones, in the way the search intent guide describes for web pages: read the results for the query, see what format already satisfies it, and make that format better. A query whose results are all long demonstrations is not won by a short talking-head video. Record the primary query with the video brief so that the title, description and chapters are written against it, and so that the search terms report can later show whether the video is being found for what it was built for.
A kitchen-appliance brand maps its video queries
A brand that sells stand mixers lists 140 search phrases from YouTube suggestions, the Research tab and competitor titles. Grouping them by intent produces 32 "how to" queries about using attachments and cleaning, 18 comparison queries naming competitor models, 11 "is it worth it" queries, and a long tail of recipe searches that are not about the mixer at all. The brand plans 12 videos for the quarter: 8 tutorials, 3 comparisons and 1 explainer about motor types, and assigns each a primary query recorded in the brief. The search terms report after 60 days shows the attachments tutorial being found for 46 distinct queries, 70% of them variants of its primary query, and the explainer being found for almost none; the explainer's title had used the brand's own phrase "planetary action" rather than the "how do stand mixers work" phrasing viewers type. Retitling it against the recorded primary query lifts its search impressions from 900 to 4,300 a week.
Metadata, thumbnails and the core guides
Metadata is the text a video carries: the title, the description, chapters and their timestamps, tags, hashtags, the category, captions and the file name. It is what the search system reads to establish relevance and what the viewer reads to decide whether to click. The thumbnail is the image the viewer sees first, and together with the title it decides whether an impression becomes a view.
The core guides divide this work. The titles, descriptions and chapters guide covers the text: a title that carries the primary query and a reason to click within the characters that show on each surface; a description whose first lines do the work because they are what shows before the fold; chapters that meet YouTube's documented requirements and can surface as key moments in Google Search; and what tags and hashtags do and, more importantly, do not do. The thumbnails and click-through guide covers the image: design principles, the thumbnail and title as one unit, consistency across a channel, testing, and how to read impressions click-through rate without being misled by it.
One rule runs through both. The packaging must describe the video that exists. A title and thumbnail that earn a click the video cannot honour produce a viewer who leaves early, and YouTube says plainly that early abandonment counts against the video. A good video with honest packaging may get fewer clicks than a misleading one, but it keeps the viewers it gets, and that is the signal the systems reward. The indicators (KPIA metric tied directly to an objective.) to watch are impressions, click-through rate and retention read together, never one alone.
Further guides
Further tutorials apply the core guides to the videos a channel has already published and to the searches it shares with the business's website:
- Auditing a YouTube video library — a tutorial for auditing a channel's back catalogue: sorting videos by job and traffic source, fixing packaging, chapters and next steps, then logging every change.
- Planning YouTube videos and web pages for the same searches — a tutorial for putting the website's keyword map and the channel's query register in one sheet, reading each results page, deciding for every query what each side owns and pairing a page with a video where both are wanted.
Video in Google Search results
A YouTube watch page is a web page, and Google indexes it like any other. Videos appear in Google Search as video results in the main results, in video carousels, in the video tab, and sometimes with key moments: a list of jump points under the result that lets a searcher land at the part of the video that answers them. Google documents that for YouTube videos the key moments can be drawn from the chapters a creator adds to the description, which makes chapters a Google Search feature as well as a YouTube one.
This is the external traffic source, and it behaves like organic search traffic to a website: it depends on the query wanting video, on the title and description matching it, and on the video being a good answer. Queries that show a video carousel on the results page are the ones to target, because Google has already decided they want video. The same video can be embedded on the business's own page for that query, which gives the page a reason to rank and gives the video another route to views; the broad search engine optimisation discipline (SEOImproving visibility in unpaid search results.) and the YouTube-specific version are the same work viewed from two sides, and an SEO strategy that plans both avoids the common split where the website team and the video team target the same queries without knowing it.
Common video SEO problems
Most video SEO problems show up in the traffic sources and reach reports as one of a few patterns. The table maps each to the guide that diagnoses it.
| Symptom | Usual cause | Where to go |
|---|---|---|
| Few impressions from search for a video built for a query | The title and description do not use the words viewers type; the query wants a different format | Titles, descriptions and chapters; keyword research section above |
| Many impressions, low click-through | Thumbnail and title do not promise what the audience is looking for, or they look like every competitor | Thumbnails and click-through |
| Good click-through, early abandonment | Packaging promises something the video delays or does not deliver | Both guides; the traffic sources guide for which surface the viewers came from |
| Search views but no suggested views | The video answers a query but is not a natural next video for anything; retention weak after the answer | Traffic sources guide and the retention guide it links |
| Found for the wrong queries | Title or description phrasing matches a different intent | Titles, descriptions and chapters; intent classification |
| No key moments or chapters in Google results | Chapters missing or not meeting the documented requirements | Titles, descriptions and chapters, chapters section |
| Several videos competing for one query | No primary query assigned per video; the channel answered the same search more than once | Keyword research section above |
| The website and the channel answer the same search without knowing it | The keyword map and the channel's query register are kept apart, so nobody decides which side owns each query | Planning YouTube videos and web pages for the same searches |
| Older videos keep earning views on packaging written when the channel knew least | The back catalogue has never been reviewed against the channel's own baseline | Auditing a YouTube video library |
Two of these symptoms have their own step-by-step diagnoses in YouTube channel troubleshooting: the low impressions click-through rate diagnosis and the diagnosis for a video not showing in search.
Video SEO is reviewed on the same rhythm as the rest of the channel: the search terms per video after publishing, impressions and click-through weekly, and a quarterly pass over the library to retitle, add chapters to and re-thumbnail the videos that earn search traffic but could earn more, which the library audit tutorial works through step by step.
