YouTube keyword research is the process of finding the phrases people actually type into YouTube search, judging how hard each one is to win, and shaping your video around the one you can realistically be the best answer for. You do not need a paid SEO suite for any of it. YouTube's own autocomplete, the search results page, competitor channels, Google Trends filtered to YouTube, and your own analytics cover the whole job for free.
A paid tool buys speed and tidier numbers. It does not buy judgement about intent — whether a searcher wants a two-minute fix or a forty-minute walkthrough — and that judgement decides whether a video ranks and holds viewers. This guide is the research layer under our broader YouTube SEO guide: finding terms, reading intent, turning both into a shortlist you can film against.
What Keywords Do and Do Not Control
YouTube surfaces videos through two very different mechanisms, and keyword research only governs one of them cleanly.
Search is query-driven. Someone types a phrase, YouTube matches it against titles, descriptions, captions, and its own understanding of the topic, then ranks by relevance and engagement history. Keywords are decisive here.
Suggested and browse are behaviour-driven. YouTube shows a video because people who watched something similar went on to watch it. Your keywords barely register; watch time, retention, and topical adjacency do the work.
So keyword research wins the durable, compounding half of YouTube — videos that keep pulling views months after upload because they answer a query people keep typing. It does not manufacture a spike. For the definitional version of how these surfaces interact, our YouTube SEO glossary entry is the short form.
The Free Toolkit
| Tool | What it gives you | Where it falls short |
|---|---|---|
| YouTube autocomplete | Real query strings, ordered roughly by popularity | No volume numbers; partly personalised |
| YouTube search results page | Intent, format, competition, content gaps | Requires manual reading |
| Competitor video tabs (sorted by popular) | Proven title patterns and topic demand | Skewed toward older, accumulated views |
| Google Trends, YouTube Search filter | Relative demand over time, seasonality, regions | Relative index only, not absolute searches |
| YouTube Studio research tools | Query data with rough demand bands | Coverage varies by niche and channel size |
| Your own YouTube Analytics search traffic | Queries already sending you views | Only reflects what you already rank for |
That is the entire stack. Everything below is a workflow for using it in order.
Step 1: Mine Autocomplete Systematically
Autocomplete is the highest-signal free source on the platform because it reflects what people type, in their own words, including the awkward phrasings no keyword tool would generate.
Work it methodically rather than poking at it:
- Seed term alone. Type "sourdough starter" and record every suggestion.
- Alphabet pass. Add a space and each letter in turn — "sourdough starter a", "sourdough starter b" — and record the new suggestions. Most seeds are exhausted after a dozen letters.
- Question prefixes. Put the question word first: "how to sourdough starter", "why sourdough starter", "when sourdough starter". These surface the problem-shaped queries that make the easiest videos to rank.
- Comparison and qualifier prefixes. "best", "vs", "for beginners", "without", "cheap", "at home". Negations are gold — "sourdough starter without discard" is a whole video that most channels have not made.
Two habits make the output cleaner. Search from a logged-out or incognito window so your own watch history stops colouring the suggestions. And paste everything into a sheet unfiltered — judging terms while you are still collecting them cuts the list down too early. A realistic first pass produces 60-100 raw phrases per seed topic in about half an hour.
Step 2: Extract Competitor Titles
Autocomplete tells you what is searched. Competitor titles tell you what has already been rewarded.
Open three to five channels in your niche, go to their Videos tab, and sort by most popular. Then read the top 20 titles of each as a set rather than individually. You are looking for:
- Recurring phrase patterns. If four channels independently published something containing "for beginners", that modifier has proven pull.
- The gap between views and recency. A three-year-old video with heavy views on a topic nobody has revisited is an opening. A topic where every strong entry is from the last two months is a crowded lane.
- Weak leaders. Search the term yourself. If the top result is off-topic, poorly made, or answers a slightly different question, that is a rankable term even if the channels above you are large.
A mini example. A small espresso channel found "how to dial in an espresso grinder" dominated by big equipment channels, but "espresso grinder retention fix" had only two videos, both over two years old and both filmed on a grinder that had since been discontinued. That is a clean opening: proven query, stale answers.
Record the title patterns separately from the keywords. When you write the actual video title later, you will want both — the term and the phrasing that has historically earned clicks. Our YouTube titles and descriptions guide covers turning those patterns into titles that hold up on the results page.
Step 3: Google Trends, Filtered to YouTube
Google Trends has a search-type dropdown that most people never touch. Switch it from Web Search to YouTube Search and the entire dataset changes to YouTube query behaviour.
Three jobs it does well:
Settling synonym fights. You have "air fryer chicken thighs" and "air fryer chicken thigh recipe" in your sheet and no idea which to lead with. Put both in Trends, compare, and take the higher line. Neither term gives you a search count, but the ratio between them is what you need.
Mapping seasonality. Trends shows the shape of demand across the year. Because ranked videos accumulate authority over weeks, publishing at the peak means arriving late. Filming and uploading two to four weeks ahead of the climb gives the video time to gather early watch signals before the traffic arrives.
Catching regional splits. If your audience is concentrated in one country, check the term there specifically. Demand can be strong in one market and flat in another, and the wrong regional read sends you after a topic your actual viewers are not searching.
Treat the numbers as relative interest, never as volume. As of July 2026, Trends still publishes an index, not counts.
Step 4: Read the Results Page for Intent
This is the step people skip, and it is the one that most often decides whether a well-chosen keyword produces a video that ranks or one that dies quietly.
Search your candidate term and read the first screen as evidence:
| What the results show | What searchers want | What to make |
|---|---|---|
| Short, tightly cut tutorials | A task completed, fast | Direct demonstration, no preamble |
| Long, in-depth videos with chapters | Understanding, not just steps | Structured deep-dive with clear sections |
| Mostly Shorts | A single quick answer | Consider a Short; see below |
| Reviews and comparisons | A purchase decision | Honest side-by-side with a verdict |
| Loosely related, off-topic results | Nobody has answered this properly | The gap — make the obvious video |
The last row is the most valuable and the easiest to misread. Off-topic results sometimes mean no demand rather than an opening, so cross-check against autocomplete: if the phrase appears there but the results are poor, that is a real gap. If it appears nowhere in autocomplete, it is probably a phrase you invented.
Where results skew heavily toward Shorts, the honest answer may be a Short rather than a long-form video — a different discovery game with its own signals, covered in our Shorts SEO and discovery guide.
Reading Suggested Intent
One extra trick. Open the top-ranking video for your term and look at its sidebar. Those suggestions are YouTube's own read of what a viewer of this topic wants next. That is a ready-made cluster: your target video plus the two or three follow-ups that keep the same viewer in your channel rather than someone else's.
Step 5: Shortlist and Score
You now have a large messy sheet. Turn it into a shortlist with five columns:
| Keyword | Source | Demand signal | Competition read | Mapped video |
|---|
Score by a single question: can I plausibly be the best answer on this results page? Not the best video in the abstract — the best answer to that specific query, for the person typing it. A channel with 300 subscribers and genuine hands-on experience beats a large generalist channel on a narrow technical term all the time.
Small channels should weight the shortlist toward long-tail and question-shaped terms. Ranking first for a query a few hundred people search each month beats placing fortieth on a term with a huge audience. The narrow win also teaches YouTube what your channel is about, which makes the next tier easier.
Step 6: Put the Keyword Where It Counts
A shortlist is worthless until it lands in the video. Four places, in order of weight:
- Title. Lead with the term, add the hook after. Check length with the YouTube character counter before publishing — the front of the title carries the search weight and survives truncation everywhere.
- First lines of the description. Write the phrase naturally in the opening sentence or two. Nothing stuffed; a keyword repeated eight times reads badly to humans and does not help.
- Spoken audio. Say the phrase out loud in the first thirty seconds. Captions are indexable text, and it costs you nothing.
- Tags and hashtags. Genuinely minor signals as of July 2026. Fill them sensibly and move on — see our YouTube hashtags and tags guide for what they still do.
This is where research meets the publishing workflow. In SocialKit you compose once and customise the title, description, and hashtags per platform, so the keyword you picked goes into the YouTube upload while the cross-posted Short or LinkedIn clip gets copy written for its own audience, all scheduled from one visual calendar. To be clear about what it is not: SocialKit has no keyword volume database and no rank tracking. The research happens in the free tools above — SocialKit is where the winning phrase lands and ships on schedule.
Step 7: Close the Loop With Your Own Data
Two to four weeks after publishing, open YouTube Analytics and filter traffic source to YouTube search. You will get the actual queries sending views to that video.
Three things to do with that list:
- Compare against your target. If you are ranking for a variant you did not target, the variant is often the better title.
- Harvest the surprises. Queries you never considered are validated video ideas — feed them straight back into your content ideas pipeline.
- Check impressions against clicks. High impressions with weak click-through rate means you ranked and lost the click, which is a title and thumbnail problem, not a keyword problem.
Our YouTube analytics guide covers the wider set of reports worth checking on a monthly rhythm.
This discipline transfers. Every search-first platform rewards the same behaviour, which is why the method here mirrors our Pinterest keyword research workflow — different surface, same loop of mining suggestions, reading intent, and checking what actually landed.
Start Here: A Two-Hour First Pass
If you have never done this properly, block two hours and work in this order:
- 30 minutes — autocomplete. Three seed topics, full alphabet and prefix passes. Everything into a sheet, unfiltered.
- 30 minutes — competitor titles. Four channels, sorted by popular, top 20 titles each. Note patterns and stale leaders.
- 20 minutes — Google Trends. Take your ten strongest candidates, compare them in pairs with the YouTube Search filter, and note seasonal peaks.
- 20 minutes — results-page reads. Search your top five terms and classify the intent for each.
- 20 minutes — shortlist. Five keywords, each mapped to one video and one publish date on the calendar.
Then set a monthly repeat: fifteen minutes of fresh autocomplete on your core topics, and one look at the search queries your last four videos actually pulled. Keyword research that happens once goes stale within a quarter. The channels that compound are the ones running this small loop every month.