YouTube Keywords: Intent and Metadata Instead of Tags

I wrote the first version of this article in 2018, and almost all of it was devoted to tags: where to find them, how to arrange them, how many characters to squeeze out of the 500-character limit.

Giving that kind of advice in 2026 would mean lying to the reader.

YouTube has long been analyzing videos with neural networks: it “hears” speech, reads captions, and understands what is happening on screen. The tag field has gone from the main optimization tool to a vestige.

But keywords themselves haven’t gone anywhere. What changed is where they work: the title, the description, chapters, captions, and — increasingly — the spoken words in the video itself.

Our SEOquick channel is living proof: for years, the bulk of its traffic has come from YouTube search, with Google search in second place. That is the result of working with semantics, not tags. We have a separate in-depth guide covering the channel as a whole — YouTube promotion — while this article focuses narrowly on keywords and metadata.

In short: in 2026, YouTube keywords are gathered from YouTube autocomplete, vidIQ/TubeBuddy, and Google Trends with the “YouTube Search” filter, and placed in the title, the first lines of the description, chapters, and captions. According to YouTube’s official position, tags play a minimal role in video discovery — spending hours on them is pointless.

Why Tags Are No Longer the Main Thing (The Honest Answer)

Let me start with something that would have sounded like heresy in 2018.

YouTube tags are a utility field in video settings that historically helped the algorithm understand what a video is about. Today, YouTube’s Help Center states outright: “Tags play a minimal role in your video’s discovery.” The same wording comes up regularly on the official Creator Insider channel.

Why keep tags at all, then? For two narrow tasks:

1. Typos and spelling variants. If your brand or a key term is often misspelled (alternative spellings, transliterations), tags help connect the query to the video.
2. Disambiguation. When a word has several meanings, a tag clarifies the context.

That’s it. No more “first tag = video title,” no more “fill 450 of the 500 characters.” Research backs this up: in analyses of YouTube search results, only about a third of top-ranking videos contain the query in their tags, while ~90% of top videos have it in the title.

My practical verdict: spend 60 seconds on tags — the main keyword, 2-3 synonyms, common misspellings, your brand name. Then move on to what actually affects rankings.

Search vs. Recommendations: Two Different Algorithms

Before collecting keywords, you need to understand which kind of traffic you are actually competing for. YouTube has two fundamentally different distribution systems.

Recommendations (the home page, “Up Next,” the Shorts feed). Viewer behavior rules here: retention, thumbnail CTR, the audience’s watch history. According to Google itself, around 70% of watch time on the platform is driven by recommendation algorithms. Keywords work indirectly here — they help the system understand who to show the video to.

YouTube search. Here the viewer formulates the query themselves, and the match between the query and the video’s metadata and content is a direct ranking factor. It is a smaller share of the platform’s traffic, but the most evergreen one: a search-driven video collects views for years, not just 48 hours after publication.

This leads to the main rule of keyword research:

• Trending and entertainment topics — optimize for recommendations (thumbnail, first 30 seconds, retention). Semantics is secondary.
• Educational, review, and how-to topics — optimize for search. Here semantics is decisive, and the rest of this article is about exactly that.

You can check where your traffic comes from in YouTube Analytics → “Traffic sources.” If “YouTube search” is in your top 3, this article will pay for itself tenfold.

▶ Как продвигать видео на YouTube без бюджета

Where YouTube Reads Your Keywords

Let’s go through every keyword placement — in descending order of importance.

Video Title

The strongest signal. A classic Ahrefs study showed the correlation between an exact-match keyword in the title and ranking at the top of YouTube search — and that relationship has only grown stronger since.

The rules for 2026:

1. Main keyword in the first third of the title. On mobile, long titles get truncated after roughly 35-40 characters, and the viewer sees an ellipsis after that.
2. Optimal length is 50-60 characters. The technical limit is 100, but shorter titles both read better and correlate with higher positions in SERP studies.
3. One keyword — one video. Don’t try to cram three queries into a title separated by slashes. Shoot a second video for the second intent.
4. Keyword + click trigger. The formula for a working title: exact query + benefit or intrigue. “YouTube keywords” is boring. “YouTube keywords: the method that took our channel to the top” — now that’s a title.

Description

The second most powerful zone. YouTube uses the description to understand the video’s topic, and the viewer sees the first 2-3 lines above the “more” button — these same lines often get pulled into the search snippet.

How to write it:

• The first 150 characters — a self-contained phrase with the main keyword. This is your “meta description” for YouTube and Google search.
• The first 2-3 paragraphs — expand on the topic with synonyms and long-tail variations of the query, in natural language.
• Don’t pad it. A 4,000-character wall of comma-separated keywords is a 2015 practice. It doesn’t help and looks like spam.
• 3-7 links maximum — to related videos, a playlist, your website.

Chapters (Timestamps)

An underrated tool. Chapters are timestamped markers in the description that YouTube turns into in-video navigation.

What they do for semantics:

• Each chapter is a separate line of text indexed by both YouTube and Google. Put keywords and long-tail query variations in chapter titles.
• Google displays chapters as “key moments” right in the search results — the video gets several entry points from search instead of one.
• Viewers find the fragment they need faster and leave less often — retention grows, and that is already a signal for recommendations.

The format is simple: the first marker must be 0:00, at least three chapters, each at least 10 seconds long. Chapter title = the sub-query that the fragment answers.

Captions and Spoken Words

This is where the real revolution happened. YouTube transcribes the audio track and uses the transcript to understand the content — captions give the algorithm dozens of times more indexable text than the title, description, and tags combined.

Practical takeaways:

1. Say your key queries out loud — especially in the first 15-30 seconds of the video. The phrase “today we’ll break down how to research YouTube keywords” is a ranking signal, not filler.
2. Check auto-generated captions. Automatic speech recognition makes mistakes, especially with terminology and brand names — which means the wrong words get indexed. Upload a proofread caption file, or at least fix the auto-captions in the editor.
3. Captions in other languages expand your search reach to new markets — we have a separate video on how to do this.

▶ Как добавить субтитры на youtube на другом языке

Hashtags Today

Hashtags get confused with tags, but they are different things: hashtags are written in the description with a # sign, are visible to viewers, and are clickable.

The current rules:

• The first three hashtags from the description are displayed above the video title.
• The hard limit is 15 hashtags: exceed it, and YouTube ignores all of them.
• The working scheme: 3-5 total — one broad topical hashtag, 1-2 narrow ones specific to the video, one branded (#SEOquick).
• Specific beats generic: #legalenglish is more useful than #english.

The influence of hashtags on rankings is modest: they are navigation and extra context, not magic. But they are worth the 30 seconds.

Tags: What to Do With Them

For completeness, here is the minimal routine for the tag field: the main query, 2-3 synonyms, common misspellings, your channel name. One minute of your time, no more. The details are above, in the first section.

How to Research YouTube Keywords: From Intent to Semantics

Now for the process itself. The logic is the same as when building a semantic core for a website, but the data sources are different: Google’s classic Keyword Planner is built for ads and does not reflect YouTube search demand.

Step 1. Break Your Audience Down Into Intents

Don’t start with tools — start with people. Write out your audience segments and the questions each segment brings to YouTube.

An example from our channel: the target audience is website owners, but within it the intents are completely different:

• already doing SEO, looking for new strategies;
• unhappy with their contractor, choosing a new one;
• have a website that was never promoted — searching for “where to start”;
• website still in development — searching for “how to do it right.”

Each intent is its own series of queries and its own series of videos. The easiest way to map this out is a table or a mind map: segment → pain points → query phrasings.

An important nuance of YouTube intent: people come here to learn or to be entertained, not to buy. The query “buy a pressure canner” lives in Google, while on YouTube it becomes “how to choose a pressure canner” and “pressure canner review.” When collecting semantics, translate commercial queries into educational and review phrasings.

Step 2. Collect YouTube Autocomplete Suggestions

The most honest data source is YouTube’s own search bar. Suggestions are generated from real user queries.

The mechanics:

1. Type in a seed query and write down the suggestions.
2. Add a space after the query and cycle through the letters of the alphabet — you’ll get long-tail variations.
3. Put an _ (underscore) before the query — YouTube will show variants with words added in front.
4. Repeat in incognito mode to remove the influence of your own history.

For bulk collection of suggestions, web scrapers like Keyword Tool or Kparser will do — both work specifically with the YouTube database and export results to CSV. Their metrics are limited in the free versions, but as a long-tail generator they are a workable option.

Step 3. Run Your Keywords Through vidIQ or TubeBuddy

Once the draft list is collected, you need metrics: how many people search for each query and how high the competition is.

vidIQ is our primary tool. A Chrome extension plus the Keyword Research web tool: it shows search volume, competition, and an overall score for each query, and suggests related keywords and questions. The competitor analysis is separately useful: you can see which queries other channels’ videos rank for and cover those same intents with your own content.

TubeBuddy is its direct competitor, with the Keyword Explorer tool: the same “volume + competition + overall score” principle, plus content-planning features like best time to publish. The free tier is more limited than vidIQ’s.

Which one to choose is a matter of taste; over the years we settled on vidIQ. What matters more: the metrics of both services are estimates. Use them to compare queries against each other (“this one is searched noticeably more often than that one”), not as absolute truth.

Step 4. Check the Trend in Google Trends

Google Trends has an underrated filter: in the dropdown menu, switch from “Web Search” to “YouTube Search” — and you’ll see demand dynamics specifically on YouTube.

Why you need this:

• Seasonality. There’s no point shooting “how to set a New Year’s table” in February.
• Growing vs. dying topics. A declining trend is a reason not to invest in producing an expensive video.
• Comparing phrasings. “SEO for beginners” vs. “SEO from scratch” — Trends will show which phrasing is more popular, on a relative scale.

Trends doesn’t give absolute numbers, but it gives direction and comparison — for free.

Step 5. Add Data From Classic SEO Tools

Ahrefs has kept its Keywords Explorer for YouTube — the largest public database of YouTube queries with volume estimates. For some languages and regions the database is thin, so I use a workaround: in the regular Keywords Explorer for Google, I apply the SERP features filter “contains video” — and get the queries for which Google itself shows videos in the results. Such keywords hit two targets at once: YouTube search and Google’s video carousels.

Step 6. Cluster the List and Turn It Into a Content Plan

The collected list (usually hundreds of queries) needs to be grouped. Our free keyword clustering tool helps here: upload the list, set up negative keywords — and get ready-made groups.

Then a simple rule:

• A large cluster (dozens of queries, different subtopics) → a playlist or a series of videos.
• A small cluster (one subtopic, 3-10 phrasings) → one video: the main query in the title, synonyms in the description and chapters, all of it spoken out loud in the video.

Treat your semantics as a living document: mark which clusters already have a video, and review the plan quarterly — demand on YouTube shifts faster than on Google.

SEOquick experience: the “intent → cluster → content” approach works beyond video. In our medical website promotion case study, working out the semantics around real patient questions took the clinic to TOP-2 for the query “dentistry Kyiv,” and the site’s pages appear in Google’s AI answers for 26,714 queries. The principle is the same: first understand what people ask and how, then produce the content.

Keywords for Shorts

Shorts live by their own rules, and carrying over the logic of regular videos is a mistake.

The main thing: the Shorts feed is almost entirely recommendation-driven. What decides is not query matching but viewer behavior in the first seconds: did they watch through or swipe away. Studies of the algorithm describe an “explore and exploit” model: a Short is tested on a small sample, and with good retention signals it gets rolled out to an ever-larger audience.

Does this mean Shorts don’t need keywords? No:

1. Shorts are indexed in search — both YouTube and Google. A clear title with a keyword (up to ~70 characters so it doesn’t get cut off on mobile) gives a Short a long search tail after the feed has “worked it through.”
2. Shorts have descriptions too — 1-2 sentences with a keyword and context are better than nothing.
3. Hashtags — 3-5 specific ones, by the same rules as for long-form videos. A dedicated #Shorts hasn’t been required for a long time, but it doesn’t hurt.
4. Spoken words get indexed — name the topic out loud in the very first second of the video.

And a sober assessment: for getting into AI answers and long-term search, Shorts are nearly useless — according to OtterlyAI research (2026), 94% of YouTube video citations in AI answers go to long-form videos, with less than 6% for Shorts. Shorts are a tool for reach and subscribers, not search traffic.

Videos in Google Search and AI Overviews

Well-chosen keywords give a video a second life beyond YouTube — in Google search results and in AI answers.

Numbers worth knowing. According to a Surfer SEO analysis (46 million citations, 2026), YouTube is the most-cited domain in Google AI Overviews: around 23% of all links, more than Wikipedia. And OtterlyAI’s research revealed something unexpected: almost 41% of videos cited by AI had fewer than 1,000 views. AI systems pick videos by relevance to the query, not by popularity — that is, by the very metadata and transcript we are optimizing.

What this means for keyword work:

• Question-based queries (“how,” “why,” “what is”) are the priority: these are exactly what AI systems rephrase.
• Chapters and an accurate transcript are critical: AI cites a specific fragment of the video, and marked timestamps help it find that fragment.
• A direct spoken answer at the start of the video is the video equivalent of GEO-optimizing text.

We covered Google video snippets and channel strategy in detail in the YouTube promotion guide, and AI search optimization in general in our articles on GEO optimization for GPT and content for AI answers. I won’t duplicate that here — just keep in mind: a video’s semantics now works on three fronts at once (YouTube search, Google search, AI answers).

▶ SEO Оптимизация видео на YouTube + в Поиске Google

SEOquick experience: presence in AI answers is already a measurable channel. In our service website promotion case study, the project’s content appears in Google’s AI answers for 1,478 queries — and video content with the right semantics gets there by the same mechanics as text.

Updating Keywords on Old Videos: What to Expect

A video can’t be re-edited after publication — but metadata can be changed as much as you like. Is it worth it?

We ran an experiment on our own channel: we bulk-updated tags and descriptions on old videos and measured the rankings. The honest result — there was an uplift, but a minor one. Metadata “lifts” a video only if the content itself answers the query; new tags won’t help a dead video.

What actually works with a channel’s archive:

1. Rewrite the title and the first lines of the description around the actual query the video is already getting impressions for (see the “Traffic sources → YouTube search” report in your analytics).
2. Add chapters — old videos often never got them, and they are new entry points from Google search.
3. Proofread the captions — fix speech-recognition errors in terminology.
4. For hopelessly outdated videos — shoot a new version targeting the same keywords, link to it from the old video and its description, replace the embeds on your website, and set the old video to “unlisted.”

FAQ

Do YouTube tags still work in 2026?

Practically not. YouTube’s official Help Center states outright that tags play a minimal role in video discovery and are mainly needed for misspellings and ambiguous words. Spend one minute on them — the main keyword, synonyms, misspellings, your brand — and focus on the title, description, chapters, and captions.

Where is the best place to put keywords in a video?

In order of importance: the title (keyword in the first third, 50-60 characters long), the first 150 characters of the description, chapter titles, captions, and the spoken words in the video. Say the main query out loud in the first 30 seconds — YouTube indexes the transcript.

How can I research YouTube keywords for free?

A combination of three free sources: YouTube search bar autocomplete (cycle through letters after the query), Google Trends with the “YouTube Search” filter for assessing trends and comparing phrasings, and the free tiers of vidIQ or TubeBuddy for assessing competition. Our free keyword clustering tool makes grouping the list easy.

How are YouTube keywords different from website keywords?

By intent. On Google people search to find and buy; on YouTube — to learn and be entertained. Commercial queries (“buy X”) barely work on YouTube — their place is taken by educational and review phrasings (“how to choose X,” “X review”). That’s why channel semantics is collected separately, not copied from the website.

How many hashtags should I put under a video?

3-5 specific hashtags: one topical, 1-2 narrow ones specific to the video, one branded. The first three are displayed above the video title. The hard limit is 15: add more, and YouTube ignores all of the video’s hashtags. Hashtags have a weak influence on rankings — they are navigation and context, not a top-ranking factor.

Do keywords help get into AI Overviews?

Yes. YouTube is the most-cited domain in Google AI Overviews (around 23% of links according to Surfer SEO, 2026), and AI picks videos by relevance, not views: 41% of cited videos had fewer than 1,000 views. What matters: question-based phrasing in the title, chapters, and an accurate transcript with a direct answer at the start of the video.