The most expensive mistake on YouTube is not a bad thumbnail; it is three days of production spent on a topic nobody searches. Topic research is twenty minutes, happens before the camera turns on, and is the cheapest insurance in the whole craft.
Here is the routine: mine real searches, stack the demand evidence, read the competition, and pick from a matrix instead of a mood.
Step 1: mine what viewers actually type
YouTube autocomplete is the demand ledger: every suggestion is a completion of real searches. The generator on this site pulls those suggestions in bulk for any topic, which turns one seed into a map of what your niche asks YouTube for.
Run your channel's three core topics and collect everything: the phrasings, the questions, the qualifiers. This pile is the raw material; the next steps decide which entries deserve production.
Step 2: read the intent shape
Video searches carry format instructions if you read them.
| The search says | Viewer wants | Format to film |
|---|---|---|
| how to, tutorial, step by step | To follow along | Process video, chapters, real time where it counts |
| fix, not working, why does | Rescue, now | Direct answer first, diagnosis second |
| best, vs, review, worth it | A decision | Comparison with a verdict |
| what happens if, can you | Curiosity settled | The experiment, shown |
| for beginners, explained | A gentle start | Assumption-free walkthrough |
Matching the format to the intent matters as much as picking the topic: the search "sourdough proofing fix" clicked into a 20-minute vlog is a back-button; the same search into a 6-minute direct fix is a subscriber.
Step 3: stack the demand evidence
The small-channel test deserves the emphasis: a phrase where a 3,000-subscriber channel holds 200,000 views is a topic that pulls its own audience. A phrase only won by giants proves their reach, not the topic.
Step 4: read the room before entering
Search the phrase and study the top ten as competition: how big, how recent, how good. Three readable situations: all-giant and fresh means walk on by for now; strong but aging winners means demand with stale supply, the classic opening; scattered weak coverage means the topic is yours to define.
Note the angle gaps too: ten results all filming the same approach leave every other approach unclaimed, and "same topic, missing angle" is where mid-size channels live.
Step 5: decide from the matrix
Each candidate topic gets three quick scores: demand (the evidence stack), openness (the competition read), and fit (can you make this genuinely well). Shoot the topics scoring high on all three; park the high-demand-closed-room ones for when the channel is bigger; skip the no-evidence hunches or make them knowingly as suggested-feed bets.
The surviving list becomes the shoot list, and each topic arrives with its research attached: the exact phrasing for the title, the variants for the description's context zone, and the two-minute tag set already in hand.
One phrase is usually a series
Before leaving the research, look at the variant cluster around any winning topic: the questions, the levels, the situations. A topic with eight strong variants is not one video; it is a series with the episodes pre-validated, which compounds in ways single uploads never do.
The one-line takeaway: twenty minutes before filming: mine autocomplete, match the format to the intent, demand-check with views and the small-channel test, read the room, and shoot only what scores on demand, openness and fit. Film demand, not hunches.