Evidence-verification operational skill for small channels (0–10K subscribers). Separate the claims in advice videos from official documentation, and run a 90-day experiment loop using only free YouTube Studio features.
When to use this skill
- When asked to address YouTube channel growth strategy or the initial stagnation period (valley of death)
- When an execution plan is needed for titles, thumbnails, niche, and upload frequency
- When claims from growth advice videos/articles need fact-checking
When not to use this skill
- Artificial boosting such as buying ads or buying subscribers → refuse (violates fake engagement policy)
- Shorts-only strategy, revenue settlement and tax → requires separate research
- Reproducing the motion/design of the videos themselves →
video-to-superprompt
Instructions
1. Diagnosis (Day 0)
- Collect 28-day impressions, CTR, average view duration, and traffic sources in Studio > Analytics.
- Define the channel in one sentence: "A channel where [audience] comes to solve [need]".
- Judge in a table whether the last 8 videos match that sentence. Exclude mismatched videos from the series.
2. Lock the niche (Weeks 1–2)
- Plan 8–10 videos with one topic, one format, and one tone (video claim; official basis is only at the level of "consistent topic and format help form regular/returning viewers").
- Pick 3 topics where demand > supply using Content gaps in Studio > Analytics > Trends (official feature).
3. Packaging (every video)
- Design thumbnail = emotion/scene and title = specific promise so each delivers different information. Do not repeat the same wording.
- Confirm within the first 30 seconds what the title and thumbnail promised (official: intro metric).
- No clickbait — even if CTR rises, recommendations decrease when watch duration is low (official).
4. Experiment loop (Weeks 3–13)
- Run Studio A/B tests (up to 3 variants, up to 2 weeks, winner = watch time share) on every eligible long-form video. Requires desktop + advanced features.
- Do not judge by CTR immediately after upload. Look after impressions accumulate, by traffic source (official: half of channels have CTR 2–10%).
- Decision rules: impressions↑ CTR↓ → fix packaging, CTR↑ duration↓ → fix promise mismatch, both low → re-examine topic demand.
- Check metrics only once a week and spend the rest of the time on planning and production. Do not check views daily.
5. Verdict (Day 90)
- Use a 28-day comparison to see new/regular/returning viewer flow. The returning viewer ratio itself does not affect reach or monetization (official).
- If all metrics show no response after filling 8–10 videos, change the niche, but do not count one viral video as proof of success.
- Not stopping but varying: change one variable at a time in the order of keep topic + change format → keep format + move to adjacent topic.
Safety boundaries
- Refuse requests for buying views/subscribers, sub4sub, or engagement manipulation (violates official policy).
- Warn that mass template/AI slide mass-production strategies risk violating the 2025-07-15 "inauthentic content" policy.
- For realistic AI synthesis (altering real people or events), instruct on the Studio 'AI use' disclosure obligation.
- Unverified figures (5% search, average of 77, etc.) shall be marked as [uncertain] and not stated as fact.
Verification
- Does the plan output attach an evidence grade (official/expert commentary/unverified) to each claim?
- Does the 90-day calendar specify the decision date, stop rule, and one primary metric?
- Do the proposed features (A/B testing, Trends, retention report) match the official URLs in REFERENCE.md?
References
REFERENCE.md— Video claim verification table, ledger of 20 official sources, list of algorithm myths, 90-day measurement design