YouTube SEO is the set of choices that make a video findable: the query you decide to cover before you even start filming, and the way YouTube and Google read the title, thumbnail, description, chapters and spoken audio.
The guides on the first page of Google UK for this search (HubSpot, Backlinko, Sprout Social, SocialBee) repeat more or less the same list of techniques. YouTube has played down at least two of those techniques in its own public documentation, and the top-ranking result, HubSpot's, still opens its list of tips with renaming the video file (DataForSEO, Google UK, September 2026). This article starts from there: what YouTube states, what the public data shows, and what is habit handed down.
What YouTube says about search signals
The help page on how YouTube Search works lists three groups of signals, and doesn't hide any of them behind vague wording.
- Relevance: how well the title, tags, description and video content match the query.
- Engagement: the watch time the video gets for that specific query, not overall.
- Quality: signals of expertise, authoritativeness and trustworthiness on the topic.
On the same page YouTube adds two clarifications that are rarely repeated. The first: the user's search and watch history feeds into the result, so the position you see is not the position your customer sees. The second: placement in organic results can't be bought.
The first point matters more than it seems. There is no single ranking to climb: a video's ranking is a function of video, query and user together. Anyone who shows a client a screenshot of their video at number one on YouTube is often looking at their own history, not at the market. The same goes for Google, where ranking factors are documented far more extensively.
What counts and what doesn't, with the source alongside
I have put the individual elements you can work on into a table, with what the most authoritative available source says about each. Where the source is a correlation study on a US sample I have said so: correlation is not causation, and a 2017 figure should be read for what it is.
| Element | What the source says | Source | Weight |
|---|---|---|---|
| Title | Among the more important pieces of metadata for discovery | YouTube guide to tags | High |
| Thumbnail | Among the more important pieces of metadata | YouTube guide to tags | High |
| Description | Among the more important pieces; but zero correlation between a description optimised for a keyword and ranking for that keyword | YouTube guide / Backlinko 2017, 1.3m videos, US | High as context, none as keyword stuffing |
| Watch time on the query | Engagement signal stated explicitly | YouTube Search guide | High |
| Chapters and timestamps | Official technical requirements; 31% of videos cited by AI engines carry timestamp signals | YouTube guide / Otterly.ai 2026 | Medium on YouTube, high in AI answers |
| Tags | "Tags play a minimal role in your video's discovery", useful mainly when the content is commonly misspelled | YouTube guide to tags | Low |
| Comments, shares, subscriptions generated | Strong correlation with first-page rankings | Backlinko 2017, US sample | Correlation, not a direct lever |
| Name of the uploaded file | No official YouTube source lists it among the signals | None | None documented |
The last two rows are the divisive ones. Renaming the file youtube-seo-guide.mp4 before uploading appears in almost every list of techniques, including HubSpot's, which ranks first in the UK for this query. I haven't found it on any YouTube help page, and it isn't in Backlinko's study of YouTube ranking factors either. It takes ten seconds, so go ahead: but know that you are following a custom, not a source.
Search and suggested feeds are two different systems
Two mechanisms coexist on YouTube and behave in opposite ways. Internal search orders results for an explicit query, and responds to metadata. Home, suggested videos and the Shorts feed are recommendation: there is no query there, and what counts is whether people keep watching.
The practical consequence: optimising the title and description moves search, not suggestions. And internal search, for a business channel, is almost always a fraction of total views. If the video doesn't hold people, the metadata takes it to the top of a query nobody searches for.
A reference for length: in the Backlinko study of 1.3 million results, first-page videos lasted 14 minutes 50 seconds on average. It is a US sample from February 2017, so it is not a rule to apply today: it is a hint that the platform doesn't penalise long form when long form is justified.
Finding the queries before you film
The difference between a video that ranks and one that doesn't is decided before the camera. Three sources, in the order I use them.
- The Research tab in YouTube Studio. It shows what people who already watch the channel search for on YouTube, and flags queries where few results are available. It is the only source that talks about your audience rather than the audience in general.
- YouTube search bar autocomplete, which returns real phrasings, almost always in question form.
- Google Ads Keyword Planner, to see whether the same question also exists on Google and what it is worth.
An example with the numbers for this very page. In the UK the query "youtube seo" gets around 480 searches a month according to Keyword Planner, with a peak of 880 in September 2025 and lows of 210 in June and August 2026. "How to rank youtube videos" sits at 20 (DataForSEO, Google UK, September 2026). These are tiny volumes: the traffic on these topics lives inside YouTube, not on Google. Anyone who sets a video strategy starting from Keyword Planner volumes reaches the wrong conclusion, and it is the same mistake people make with long-tail keywords when they only look at the volume column.
The rest of the method is the same as for text: start from search intent and build the video around the answer, not the keyword. I have written a separate guide on how to set up the groundwork: keyword research.
Title, thumbnail, description
These are the three elements YouTube itself calls "more important pieces of metadata for your video's discovery" in its official guide to tags. That sentence is worth rereading: the page explaining how to use tags spends a paragraph saying that tags matter little and attention should go elsewhere.
The title should contain the phrasing people actually use, at the start, and survive being cut off in the mobile preview. The thumbnail is not decoration: it is the factor that decides the click-through rate, which YouTube Studio reports as impressions click-through rate. If impressions CTR drops below the channel average, the problem is almost always the title-thumbnail pair, not the video.
The description serves two different readers: the person who reads the first two lines before "more", and the system that has to understand what the video is about. Write the first two lines for the person and the rest for the system, with the chapters at the bottom.
Chapters and transcripts, the two most underrated fixes
Chapters have precise technical requirements, published by YouTube in its guide to video chapters: at least three timestamps in ascending order, the first starting at 00:00, and a minimum length of ten seconds per chapter. If any of the three is missing, chapters don't appear. That is why on many business channels chapters are written in the description but don't show under the progress bar.
Outside YouTube, the same timestamps feed key moments in Google results. Google Search Central documentation describes three routes: automatic detection, Clip markup and SeekToAction. The last is supported in twelve languages, English included. If you host videos on your own site, the rest is in the guide to structured data.
The transcript is the other half. YouTube generates it automatically, with a quality that drops noticeably when speakers talk over each other or use product names. Uploading a hand-corrected subtitle file costs little and turns speech into machine-readable text, which is exactly what the next section needs.
The video outside YouTube: Google and generative answers
A video can bring visibility in three different places: inside YouTube, in Google results, and inside answers generated by AI engines. The third is the one few people write about, and it is also where the gap between those moving now and those who aren't is widest.
The largest study available is by Otterly.ai, which in 2026 analysed over 100 million citations collected over thirty days across six engines: ChatGPT, Google AI Overviews, Google AI Mode, Perplexity, Microsoft Copilot and Gemini. The study doesn't state the market or the language of the queries, so it should be read as a global snapshot, not as UK data.
Three numbers change the way videos should be produced.
- YouTube citations are concentrated on a few engines: Perplexity at 38.7%, AI Overviews at 36.6%, AI Mode at 19.6%. ChatGPT stops at 4.4%, Copilot at 0.5%, Gemini at 0.2%.
- 94% of citations go to long-form videos. Shorts get 5.7%, playlists, channels and live streams the remaining 0.3%.
- 31% of cited videos carry timestamp signals, and of these 78% receive multiple citations across two to five different chapters of the same video.
The last figure is the practical one: chapters don't only help the viewer navigate, they split the video into units a generative engine can cite separately. A twenty-minute video without chapters is one resource; with eight chapters it is potentially eight. The same reasoning applies to written content when you work on generative engine optimisation or aim to appear in ChatGPT answers.
On the Google side, if the video sits on a page of your site rather than on YouTube, the technical requirements of Google Search Central apply: the watch page must be indexed, the video embedded with standard tags, the thumbnail at a stable URL and the streaming file reachable by Googlebot. I cover this side in more depth in the guide to video SEO. On how videos appear inside generated summaries, see also the article on AI Overviews and the one on how to do SEO for Perplexity.
How much YouTube really weighs in the UK
Global YouTube statistics get shared a lot and say little to a business selling in the UK. These are the UK figures I use as a baseline.
| Figure | Value | Period | Source |
|---|---|---|---|
| YouTube reach among UK online adults | 94% | May 2025, UK | Ofcom, Online Nation 2025 (Ipsos iris) |
| Same reach a year earlier | 94% | 2024, UK | Ofcom |
| Same reach two years earlier | 91% | 2023, UK | Ofcom |
| Average time per person on YouTube | 51 minutes a day (47 in 2024) | May 2025, UK | Ofcom |
| People aged 16+ with home internet access | 95% | 2025, UK | Ofcom |
| People aged 65+ without home internet | 13% (20% among 75+) | 2025, UK | Ofcom |
Two caveats on the first row. Ofcom's Online Nation 2025, published on 10 December 2025, measures reach among online adults, not the whole population, so 94% is a share of internet users. And the figure is flat: unchanged from 2024. In the UK YouTube has stopped growing in reach because there is almost nobody left to reach; what is still growing is time spent.
The wider picture comes from the same report: 95% of people aged 16 and over have internet access at home, and 87% of those aged 65 and over do. The audience is there above sixty too, which contradicts the idea that YouTube is a channel only for the young. If you are thinking about where to shift your presence, the piece on search everywhere optimisation may help.
What to look at to tell whether it is working
Views say almost nothing. The four numbers I read first, in this order:
- Impressions and impressions CTR in YouTube Studio. They separate a distribution problem (few impressions) from an appeal problem (plenty of impressions, few clicks).
- Average view duration, not completion percentage. On a four-minute video, two and a half minutes is worth more than a theoretical 60%.
- Traffic source: YouTube search, which returns the actual queries people arrived through. It is the closest thing to a Search Console inside the platform.
- Referral sessions from youtube.com in Google Analytics, cross-referenced with conversions. It is the only point where video stops being a platform metric and becomes a business number.
If the videos are also on pages of your site, the video indexing report in Google Search Console shows which ones Google has actually indexed and which it has discarded, with the reason.
Why video is rarely a traffic channel for a B2B business
This is the part where I usually lose a client or two and win over someone else. For most industrial SMEs, YouTube doesn't bring traffic in meaningful quantities, and chasing channel rankings is time taken from something else.
A reference from our own data, already public: at Macropix, an Italian LED wall manufacturer, the work that moved revenue was on the website. In July 2026 the domain holds 25.55% share of voice on the ledwall cluster, ahead of Amazon at 18.08%, with over a hundred keywords in the top three on Google Italy and around 7,000 enquiries generated in seven years. The YouTube channel had nothing to do with it. The full story is in the case study on SEO for manufacturing.
For these businesses video has a different job: it is proof. An installation filmed on site answers the question "what does it really look like" better than any paragraph, and embedded on the product page it keeps people on the page and gives the sales team something to attach to an email. Anyone working on long sales cycles already knows this, and it is the same principle behind B2B SEO: content isn't there to be found by everyone, it is there to help a few people decide.
If you want to understand where you stand before filming anything, this is the kind of analysis we do within our SEO services and, for generative engines, in SEO for AI. More measured results are in our projects.
The number that will matter over the next two years
In the reports I send clients I have stopped putting views at the top. In their place is a more uncomfortable question: can a machine read this video? A clean transcript, chapters that meet the three requirements, speech that names products in full instead of saying "this one here".
Because the gap between a 400-view video cited by Perplexity on a commercial query and a 40,000-view video nobody was searching for is widening, and in the first case the return arrives months after publication without the counter moving. Nobody has yet published a measurement of what that difference is worth in a given market. Whoever measures it first, in their own sector, will have an advantage that budget can't buy.
YouTube SEO FAQs
YouTube names title, thumbnail and description as the more important pieces of metadata for discovery. Beyond these, internal search weighs the watch time the video gets on the specific query and quality signals on the topic. Working on the title-thumbnail pair is the fastest way to raise impressions CTR; raising average view duration is the way into suggested feeds.
Very little. YouTube's official guide says tags can be useful if the content of your video is commonly misspelled and that otherwise they play a minimal role in discovery. The Backlinko study of 1.3 million results (US sample, 2017) had already found a weak correlation between keyword-rich tags and rankings. Fill them in honestly, without spending time on them.
There is no official length. As a reference, in the 2017 Backlinko study on a US sample, first-page videos lasted 14 minutes 50 seconds on average. The figure says that long form is not penalised when it is justified, not that making a video longer pushes it up. The signal that counts is average view duration, not file length.
It depends on where you want to be found. Shorts have their own recommendation system and work for discovery. In AI engine answers the picture flips: according to the 2026 Otterly.ai study of over 100 million citations, 94% of YouTube citations go to long-form videos and only 5.7% to Shorts. The study does not state the market it was measured in.
Yes, and on Google UK a video carousel often appears for informational queries: it shows up on page one for youtube seo itself. Google can also show key moments, meaning individual chapters as separate entry points. For videos hosted on your own site rather than YouTube, the Google Search Central requirements apply: indexed watch page, video embedded with standard tags, thumbnail at a stable URL and a streaming file Googlebot can reach.
They are two separate jobs. On the platform you optimise metadata and retention for internal search and recommendation feeds. On your own site you work on the page hosting the video: indexing of the watch page, VideoObject structured data, a thumbnail at a stable URL and chapters through Clip or SeekToAction markup, which Google supports in English and eleven other languages.
