YouTube Keyword Research

Keyword research that uses real competition data, not vibes.

Type a seed. The studio fans out across YouTube Suggest, Google autocomplete, and Google related searches, filters with AI to keep only on-intent phrases, then scores each one against the real top 5 ranking channels. Their median subscriber count, view ceiling, and how stale the landscape is. The keywords your niche is missing, surfaced in 30 seconds.

Find a keyword →See how it works

Free creators get one full search per cycle · ~30 seconds end-to-end · all signals from real YouTube data

The keyword table

15–25 keywords scored on what really moves rankings.

Every kept phrase carries an opportunity score, momentum tag, difficulty pill, and the cluster it belongs to. Sortable, scannable, and grounded in real YouTube competition data. Not a global difficulty estimate that doesn’t know what your channel is targeting.

Score weighted on real competitor channel size
Momentum tag. See unclaimed niches at a glance
Difficulty pill from real top-5 channel medians
3–5 thematic clusters surfaced from the same data

Top opportunities · seed "apartment cleaning"

22 of 24 kept

Keyword

Momentum

Difficulty

Score

home office cleaning vlog

cluster · Vlog

unclaimedEASY

91

minimalist desk setup tour

cluster · Setup

activeEASY

84

small bedroom organization

cluster · Vlog

steadyFAIR

78

cleaning routine motivation

cluster · Routine

activeFAIR

72

studio apartment cleaning

cluster · Vlog

unclaimedEASY

68

how to deep clean apartment

cluster · Tutorial

steadyHARD

54

Top pick

"home office cleaning vlog"

Newest top-5 video is 220 days old · top creators < 50K subs · clear topic gap.

What the score is built from

Six real signals. One opportunity score.

The opportunity score isn’t a vibe. It’s a weighted formula. Three competition signals (feasibility, traffic, freshness), one intent multiplier, one autocomplete-rank bonus, and a momentum label. Same formula every keyword, so you can trust the comparison across searches.

01

Feasibility (45% weight)

Median subscriber count of the top 5 ranking channels. Below 10K subs scores 100 (easy to outrank). 10–100K = 75. 100K–1M = 45. Above 1M = 15 (brutal. Incumbent dominates).

02

Traffic ceiling (30% weight)

Median view count of the top 5 ranking videos. 100K+ = 100 (strong traffic to win). 10K–100K = 70 (decent). Below 10K = 45 (thin, may not be worth pursuing).

03

Freshness (25% weight)

Days since the most recent top-5 video. Below 30 days = 40 (active competition). 30–180 days = 70 (normal). 180+ days = 100 (landscape stale, opportunity wide open).

04

Intent match multiplier

Exact intent = ×1.0. Strong = ×0.9. Partial = ×0.75. Off-intent drift gets scaled down even if the surface metrics look good. Keeps the score honest about what your video would really compete for.

05

Autocomplete rank bonus

Where the keyword appears in YouTube’s autocomplete pool. Position 0–4 adds +8. Position 5–14 adds +4. Earlier autocomplete position is the strongest single signal that real viewers type the phrase.

06

Momentum label

Active (newest top-5 video < 30 days), Steady (30–180 days), Unclaimed (180+ days). Cheap badge on every keyword so you can see at a glance whether competitors are actively shipping.

Feasibility

weight 45%100

Top-5 channels median < 10K subs. Easy to outrank.

Traffic ceiling

weight 30%70

Top videos median 24K views. Solid headroom for a small channel.

Freshness

weight 25%100

Newest top-5 video posted 220 days ago. Landscape is wide open.

× intent multiplier (exact match)

×1.0

Final opportunity score

91
Anatomy of an opportunity score

You see the math. So you trust the rank.

Click any keyword to expand the breakdown. Feasibility (45%) measures how dominant the top 5 channels are. Traffic ceiling (30%) measures the view headroom. Freshness (25%) measures whether anyone’s actively shipping in the niche. Multiplied by intent match. Plus a small autocomplete-rank bonus. No black box, no proprietary "DA-equivalent". Just the formula.

Feasibility

Top-5 channels. Smaller = higher score.

Traffic

Top-5 view ceiling. Bigger = higher score.

Freshness

Days since newest top-5 video. Staler = higher.

Intent × bonus

Exact / strong / partial. Plus autocomplete-rank lift.

How it works

From seed keyword to ranked opportunity in 30 seconds.

Five stages, all of them parallelized. The competition enrichment runs only on the top 10 by initial score so you don’t pay for low-signal data.

01

Seed keyword

Type the broad topic you want to publish about. The studio uses your channel’s niche keywords as additional context to keep the suggestions on-brand.

02

Multi-source fan-out

In parallel: YouTube Suggest scrape, SerpAPI Google autocomplete, Serper "related searches". 30+ raw suggestions in ~3 seconds.

03

Claude intent filter

Sonnet 4.6 reads all 30+ suggestions, drops off-intent / branded / duplicates, keeps 15–25 phrases that match your seed intent. Tags content angle + intent strength on each.

04

Real competition fetch

For the top 10 by initial score: real YouTube top-5 results. Channel size, view ceiling, days since newest. Parallel, anonymous API key, ~10 seconds.

05

Score + cluster

Final score = feasibility×0.45 + traffic×0.30 + freshness×0.25, intent multiplier, autocomplete bonus. 3–5 thematic clusters surfaced. Top pick named.

Output structure

Seven distinct output blocks. Every keyword is publishable.

You don’t leave the studio with a list of phrases. You leave with a ranked, scored, clustered, intent-filtered keyword set. Each one carrying a content angle so the next step is "open SEO Studio", not "now what".

Opportunity score 0–100

Weighted on feasibility (45%) + traffic ceiling (30%) + freshness (25%), with an intent multiplier and autocomplete-rank bonus. Same formula every keyword. Explainable, repeatable.

15–25 filtered keywords

Claude reads 30+ raw suggestions, drops off-topic / branded / duplicate, keeps only those that match your seed intent. Each one tagged with content angle and intent strength.

3–5 named clusters

Keywords grouped into thematic clusters (e.g. "Tutorial · cleaning routines", "Vlog · room makeovers"). Tells you what 3-month content arc you could ship from one search.

Momentum tag per keyword

Active (creators ship weekly) · Steady (normal) · Unclaimed (top videos > 6 months old). Lets you spot landscapes nobody is currently fighting over.

Top pick + reason

The single highest-opportunity keyword surfaced, with a one-sentence "why this one". So you stop scrolling and start scripting.

Content angle per keyword

Each kept keyword carries a one-sentence angle suggestion. Not a template. Anchored to the intent and the niche, so the angle reads as a video idea you’d publish.

Real competition snapshot

For the top 10 by initial score: result count, top-5 median subs, top-5 median views, days since newest. Live YouTube data, not estimated. Your score is built on these numbers.

What powers it

Three live data sources. One AI judgement layer.

Real autocomplete from YouTube and Google, real "related searches" from Google’s SERP, and real top-5 ranking data from the YouTube Data API. Claude Sonnet 4.6 sits in the middle as the intent filter and clustering layer. But every score component is grounded in numbers we genuinely fetched, not estimated.

Autocomplete (YouTube)

Direct Suggest API · ranked positions surface real demand

Autocomplete (Google)

SerpAPI Google Suggest · localized to your region

Related searches

Serper Google SERP · "people also searched" panel

Competition data

YouTube Data API · search.list + videos.list + channels.list

Intent filter + clusters

Claude Sonnet 4.6 · drops noise, tags angle + intent strength

API quota

Anonymous YouTube key · doesn’t touch your OAuth quota

By plan

How many keyword searches you get each month.

Free creators get one full search per cycle so you can prove the engine on a real keyword. Paid plans charge one credit per search. Same engine, no feature differences. Re-running an old seed creates a fresh entry so you can track how a niche shifts over time.

Free

1

search

per cycle

One full keyword search per cycle. Same engine as paid plans

Solo

20

searches

included per month

Search every video idea · 3 channels

Most popular

Growth

50

searches

included per month

Same engine, higher monthly allowance · 5 channels

Agency

150

searches

included per month

Pooled across 10 channels

Same engine across every plan, including free.

See full pricing →
FAQ

The keyword engine, answered honestly.

Real answers from how the product behaves. The data sources, the score formula, the YouTube quota boundary, and what won’t work.

Still have questions? Email us →
Correct. YouTube doesn’t expose true search volume publicly. We use a multi-source proxy that’s genuinely close. We pull suggestions from YouTube’s own autocomplete (only surfaces high-volume queries by definition), Google’s autocomplete via SerpAPI, and Google’s "related searches" via Serper. The strongest signal is autocomplete rank. Earlier positions correlate strongly with real query volume because that’s how Google ranks them. We’re not estimating an exact monthly number; we’re ranking phrases by relative demand using signals YouTube’s own ranking engine uses.
Result count alone is misleading: 50K results dominated by one mega-channel is easier to break into than 5K results split across small channels. So for the top 10 keywords from your search, we hit the YouTube Data API and pull the real top 5 ranking videos, then look up median subscriber count across those channels. Below 10K = easy. 10–100K = fair. 100K–1M = hard. Above 1M = brutal. Plus median views on those top 5 (your traffic ceiling) and days since the newest one was published (the freshness signal). Three real numbers, not a vague difficulty label.
No. Pytrends and the paid Trends API are unreliable for niche queries. We derive momentum from data we already have: active (newest top-5 video < 30 days = creators are shipping right now), steady (30–180 days = normal cadence), unclaimed (180+ days = nobody is actively fighting for this keyword). Unclaimed keywords are the highest-leverage gaps because the search demand exists but no recent video is competing for it.
Score = (feasibility × 0.45 + traffic × 0.30 + freshness × 0.25) × intent multiplier + autocomplete bonus. Each component is 0–100. Feasibility rewards low top-5 channel size. Traffic rewards a strong view ceiling. Freshness rewards stale (i.e. open) landscapes. Intent multiplier (exact / strong / partial) keeps off-topic drift from inflating the score. So a 91 means the top 5 ranking channels are small AND their videos pull strong views AND nobody’s posted recently AND your seed intent matches exactly. The highest-leverage combination.
Raw autocomplete + related-search dumps are noisy. They include branded queries ("VidIQ alternative" when you searched "youtube growth"), off-intent drift ("youtube to mp3" on a creator-tools search), and obvious duplicates. Claude reads all 30+ raw suggestions in one pass, drops the noise, keeps the 15–25 phrases that genuinely match your seed intent, and tags each with a content angle. The competition enrichment then runs only on those. Keeps API quota efficient and the final ranking clean.
Three differences worth knowing. First, our competition score uses real top-5 channel size + real view ceiling + real freshness. Not a global difficulty estimate. Second, every kept keyword carries a Claude-written content angle so you don’t leave the tool wondering what to make. Third, the momentum label surfaces unclaimed niches (top videos > 6 months old) that legacy tools don’t flag. We’re not trying to replace global volume estimates; we’re replacing vague difficulty scores with a real-data view of who you’d compete with. For a full side-by-side of vidIQ, TubeBuddy, Keyword Tool IO, and the free options, see our YouTube keyword research tools comparison.
Mostly yes. The autocomplete sources (YouTube Suggest, SerpAPI, Serper) all return localized results when you seed in another language. So the keyword pool will be relevant. The Claude intent filtering works across languages. The competition enrichment is language-agnostic (it’s pulling raw API data). The one weak spot is the autocomplete-rank bonus, which calibrates best for English-language queries. For non-English the bonus contributes a smaller fraction of the final score. Still useful, just lean more on momentum + feasibility for non-English niches.
~25–40 seconds end-to-end. Behind the scenes: YouTube Suggest scrape (parallel), SerpAPI autocomplete + Serper related-search (parallel), Claude Sonnet 4.6 intent filter + cluster, then YouTube Data API competition enrichment for the top 10 keywords (parallel, 5 workers). Each search is one credit on paid plans. Free creators get one full search per cycle. Same engine, no feature differences. Re-running the same seed charges a new credit because the competition data is fetched fresh.
No. That’s deliberate. The competition enrichment uses an anonymous YouTube Data API key on our side, not your OAuth credential. So even if you run 50 keyword searches in a day, your channel’s OAuth quota is untouched and remains fully available for the SEO Studio analyses, channel audits, and competitor analyses that do need read access to your private data.
The keyword table stays inside your YTGrowth dashboard for now. Every search is saved per channel and reopens with full data + the original seed. PDF / CSV export is on the near-term roadmap; if it’s critical for an agency workflow email support and we’ll prioritize. The faster pattern most users find: copy the top 5 picks straight into the SEO Studio and run title rewrites against each one.
Yes. Every search persists per channel and shows up in your search history with the seed keyword + timestamp. Click any past search to reopen the full table. Keywords, scores, momentum tags, clusters, top pick, competition snapshots. Re-running an old seed creates a fresh entry instead of overwriting, so you can compare how a niche has shifted over time.
Yes. The keyword data is shared infrastructure. The SEO Studio reuses the same intent-options + autocomplete fan-out for its title scoring. The Outliers feature uses the same intent picker for its viral video search. The clusters from Keyword Research can be sent straight into Competitor Analysis as the niche keyword set. Every tool gets sharper the more keyword searches you run.