Three questions that decide it
1. Is there demand right now?
Look at channels that share your viewers. Did a video on this topic do better than usual for them in the last few weeks? A video at 2× or more of its channel's normal views is real evidence. Your own interest is not.
2. Is it yours to make?
Does it fit what your viewers come to you for? Have you already covered it? If you only mentioned it in passing, that counts as open.
3. Can you add something?
Watch what the others made. What did they skip, get wrong, or only talk about? If you can test it, show it, or answer what they left open, you have a reason to exist in that search result.
If all three answers are yes, make it. If one is no, move to the next candidate.
A weekly routine
If you publish every week, you need next week's idea now. This takes about an hour by hand:
- Open your competitors' channels and list their uploads from the last 7 days.
- Mark the ones doing better than usual for that channel.
- Check your own channel. Which of those topics have you covered?
- Pick three candidates and run the three questions.
- Choose one. Write the title, the hook and three or four sections.
Keep the list short. A long idea list feels productive but makes the choice harder. Three to five ranked picks is enough.
How YouTubePilot shortens it
YouTubePilot does steps 1 to 4 for you. It follows your competitors, works out how each new video did against its channel's normal, reads the transcripts, and compares them with your channel. You open it and see a short, ranked list. Each pick shows:
- Why now: one or two lines with the real numbers.
- The evidence: the competitor videos behind it, with their multiples.
- Quotes: what they said, with a timestamp you can click to check.
- Confidence: high, medium or low, with the reason.
Nate Herk's direct Dots-versus-Muse test reached 4.6× his usual views and 313,024 views in three days. Your separate Muse video also hit 2.8×, but you have not put the two agents through the same workflow.
10:021.4×
4.6×
For step 5, "Create brief" writes a starting point: a title, a hook, an outline, and notes on what the existing videos missed. You still make the video. The brief keeps it yours, not a copy.
Already have too many ideas?
Then the question is which one has evidence. Ask YouTubePilot: "Who covered this, and how did it do?" The answer comes from your competitors' transcripts, with links to the exact moments. If nobody covered it, it says so.
Nate Herk covered Claude Code Mods most extensively, with two recent videos. He said mods are plugins installed with /plugin in the CLI or desktop app and can be created through natural-language requests. He demonstrated five practical mods for next-step suggestions, cache monitoring, privacy redaction, goal tracking, and preventing file collisions. In a separate Cashkeeper demo, he showed how the mod tracks usage, cache timing, limits, and estimated API costs, then warns when the cache is close to expiring.1374
WorldofAI covered Mods as part of its Claude Code update. It described a moderator plugin that scans Claude’s output for important details, warnings, and useful information that could otherwise get buried. It also explained that users can customize the interface, behavior, and features, then highlighted a Token Weather mod that forecasts context-window fullness and charts usage across the previous 12 turns. WorldofAI concluded that Mods turn Claude Code into a customizable coding environment built around the user’s workflow.28
Across the recent videos as a whole, Nate has the broader hands-on coverage, while WorldofAI gives a shorter news-style explanation.
What not to do
- Don't copy a single outlier. One video can win by luck. Look for the topic working on two channels.
- Don't trust a score. Title and SEO scores don't predict views. Evidence does a better job.
- Don't chase old news. If the evidence is two months old, the moment has passed.
More on the method: finding YouTube video ideas and content gaps.