Everyone Who Runs Prompt One Gets The Same Five Niches
“Claude could be your YouTube growth strategist.” Seven prompts, posted in full in the comments: find your niche, build your brand, generate thirty ideas, write scripts, optimise for clicks, build community, and a ninety-day roadmap toward income.
Two of them carry lines that most cards of this kind leave out, and they deserve naming first. Prompt 5 ends “maximise click appeal without misleading viewers”. Prompt 7 ends “focus on sustainable growth, not guaranteed results”. Prompt 3 asks for ideas “a new channel can realistically compete with”. Somebody was thinking about the reader.
Which makes prompt 1 worth looking at properly, because it is the one everything else depends on.
The source post — “Claude could be your YouTube growth strategist”, BANI Academy. Seven prompts, each posted as its own comment, building toward a ninety-day roadmap.
Three of the seven ask for measurements
Prompt 1 reads: “Identify 5 promising niches with strong demand, manageable competition, monetization potential, and faceless-channel opportunities. Rank them and recommend the best fit for me.”
Demand is a measurement. Competition is a measurement. Monetisation potential is a measurement. None of the three is a matter of judgement that a well-read assistant can reason its way to — they are numbers that exist in search tools, in competitor channel pages, and in advertiser bidding, and a language model has none of them in front of it.
Prompt 3 asks for the same things again: “audience demand, competition insights” across thirty video ideas. And prompt 7 asks for “current monetization requirements” — a published, specific, changeable fact, which a model will state fluently and may state from a year ago.
The other four — names and branding, scripts, titles and thumbnails, a thirty-day community plan — ask it to build something. That is the job it is genuinely good at, and nobody needs a measurement to do it.
So the ranking is a frequency readout, and it is the same for everybody
Ask a model to rank niches by profitability and it will produce a ranking. It will be ordered, confident and plausible. What it reflects is how often those niches appear in the text it was trained on — which correlates with how much has been written about them, which correlates with how many people are already doing them.
That is close to the opposite of what you want. But the sharper problem is not accuracy, it is sameness. Two people in different countries, typing the same prompt into the same model about broadly similar skills, get broadly the same five niches. The prompt that promises you an edge is structurally the one least able to give you one, because its output is a property of the model rather than a property of you.
And prompt 1’s answer becomes the [NICHE] that prompts 2, 3, 6 and 7 all take as input. Four of the remaining six inherit it, and none of them can detect that it was wrong.
Diagram: ranking a market versus intersecting your specifics
| Ranking a market | Intersecting your specifics | |
|---|---|---|
| The question | “Which niche is most profitable?” | “What sits in the overlap of these three things only I have?” |
| What it needs | Search volume, competitor data, advertiser rates | Your skills, your access, your constraints — which you typed in |
| Does the model have it | No | Yes, because you supplied it |
| Is the answer yours | No. Everyone gets roughly the same list | Yes. Nobody else has your three things |
| What good output looks like | A confident ranking you cannot verify | A list of angles, and search terms to go and check |
| Prompts in this set | #1, #3, and the monetisation half of #7 | #2, #4, #5, #6 — and what #1 should have been |
All seven, and which kind of job each one is
| # | Prompt | Asks for | Can the model supply it? |
|---|---|---|---|
| 1 | Find your most profitable niche | Demand, competition, monetisation potential, ranked | No. Those are measurements. What comes back is a frequency readout |
| 2 | Build a powerful channel brand | 10 names, an About section, a value proposition, a visual identity | Yes. Pure construction, from the niche and audience you typed |
| 3 | Generate 30 high-potential video ideas | Titles and hooks — plus “audience demand, competition insights” | Half. The ideas and hooks yes; the demand and competition no |
| 4 | Write scripts that keep viewers watching | A hook, storytelling, examples, transitions, an ending | Yes. The strongest prompt in the set |
| 5 | Optimise every video for clicks | Titles, thumbnail concepts, description, keywords, pinned comment, CTA | Yes — and it carries the set’s best clause: “without misleading viewers” |
| 6 | Turn viewers into loyal fans | A 30-day engagement plan, ready to publish | Yes. Construction again |
| 7 | Build your 90-day growth roadmap | Publishing targets, experiments, milestones, “current monetization requirements” | Mostly — except the requirements, which are a published fact that changes |
The ninety-day promise, with the real numbers
Since prompt 7 asks the model for the monetisation thresholds, here they are from YouTube’s own help page, read this week. Two routes, both requiring a thousand subscribers:
“Get 1,000 subscribers with 4,000 qualified watch hours in the last 12 months.”
“Get 1,000 subscribers with 10 million qualified Shorts views in the last 90 days.”
And the sentence almost nobody quotes: “Meeting these thresholds doesn’t guarantee acceptance.” Channels go through a review that typically takes about a month.
Now put that against “a 90-day roadmap… working towards generating income”.
| Step | The arithmetic | Result |
|---|---|---|
| The bar | 4,000 watch hours | 240,000 minutes of watch time |
| In 90 days | All of it inside the window you are counting | 240,000 minutes, banked in three months |
| At 3 minutes average view duration | 240,000 ÷ 3 | 80,000 views |
| Publishing twice a week | 90 days ≈ 26 videos | ~3,000 views on every single video, from a standing start |
| If your AVD is 2 minutes | 120,000 views ÷ 26 | ~4,600 per video |
| If your AVD is 4 minutes | 60,000 views ÷ 26 | ~2,300 per video |
| The Shorts route instead | 10,000,000 ÷ 90 | 111,111 views a day, every day |
| Then | Review after you qualify | About another month |
None of that says the roadmap is pointless. It says the honest version of prompt 7 produces a publishing and improvement plan, and the income part is a separate question with a far longer clock. The prompt’s own closing line — “sustainable growth, not guaranteed results” — is pointing at exactly this, and it is the line the headline covers up.
It is also the clearest argument for doing what this blog keeps arriving at from other directions: model the small channels that are growing now, not the large ones that already arrived.
How to rewrite prompt one
Keep the deck. Change the first question from a ranking to an intersection, and ask for homework instead of a verdict:
“Here are three things that are true of me and would be expensive for most people to copy: [SKILL / ACCESS / CONSTRAINT]. Give me eight specific video angles that sit in the overlap of all three. For each, give me the exact search terms I should look up myself to judge demand and competition, and say plainly what you cannot know about either. Do not rank them.”
That version asks the model for the thing it is good at, hands the measuring back to the tools that can measure, and produces an answer nobody else gets — because nobody else typed your three things. We have made the same point about keyword tools: an assistant is good for generating candidates and useless for telling you demand.
Four honest caveats
- “Everyone gets the same five niches” is a claim about how the model works, not a controlled experiment. We did not run prompt 1 across many accounts and compare. It is also the easiest claim in this article for you to test yourself: send prompt 1 to two people, have them fill in different skills, and see how much of the list overlaps. If the answers diverge sharply, we are wrong and we would like to know.
- The ninety-day arithmetic is an estimate with assumptions on display. Average view duration, publishing cadence and the standing start are all ours. Change them and the per-video number changes a lot. What does not change is the shape: 240,000 minutes is the bar, and the subscriber count is a separate hurdle.
- Thresholds move. The two routes above were read from YouTube’s help page this week, in October 2026. That is precisely why asking a model for “current monetization requirements” is the wrong tool for that one field — and why we linked the page rather than retyping the numbers and leaving you to trust us.
- We did not run the seven prompts. This is a reading of what each one asks for against what a model can know. The four construction prompts may well produce excellent work; nothing here says otherwise.
The one-line version
Seven prompts, four of which ask Claude to build something — names, scripts, packaging, a community plan — and it will do those well. Three of them ask it to measure a market it cannot see, and the first of those three is the one every other prompt takes as input.
A ranking of profitable niches is not a measurement of where the money is. It is a readout of what has been written about most, and it comes back roughly the same for everyone who asks. Change that one question from “rank the market” to “intersect my specifics, and tell me what to go and look up”, and the other six prompts start from something that is actually yours.
Do this before anything else. Write down three things that are true of you and would cost somebody else real money or years to copy. That list is the only input in this entire deck that the model cannot generate — and it is the one the card asks you for in a single bracket marked [INSERT DETAILS].
One Comment