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.
Free creators get one full search per cycle · ~30 seconds end-to-end · all signals from real YouTube data
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.
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).
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).
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).
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 actually compete for.
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.
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.
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.
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.
Multi-source fan-out
In parallel: YouTube Suggest scrape, SerpAPI Google autocomplete, Serper "related searches". 30+ raw suggestions in ~3 seconds.
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.
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.
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.
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 actually 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
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 →