Vague In, Vague Out: The Five Slots That Turn a Prompt Into a Usable Answer
If the answers coming back from an AI assistant are bland, the model is rarely the problem. The prompt is.
“Help me with marketing” returns a listicle anyone could have written. “Write 10 marketing hooks for [PRODUCT] aimed at [AUDIENCE] on Facebook. Mix story, data and FOMO angles” returns something you can post this afternoon. Same model, same minute, completely different output. Nothing changed except how much of the job was specified.
That gap is the entire value of a good prompt pack. Not secret words – most of these prompts are plain English – but the habit of filling in the parts most people leave blank. This article breaks down what those parts are, so you can write your own rather than hunting for someone else’s.
A usable prompt has five slots
Nearly every prompt that produces work you can ship fills these five slots. Nearly every prompt that disappoints leaves three of them empty and hopes the model guesses.
| 1. Role → | 2. Context → | 3. Task → | 4. Constraints → | 5. Format |
|---|---|---|---|---|
| Who is answering
“You are a YouTube packaging strategist.” Changes vocabulary and what gets treated as obvious. |
What it is working with
The channel, the audience, the video, the numbers. The slot people skip most. |
One verb, one object
Write ten. Rank these. Rewrite that. Not “help me with”. |
The boundaries
Length, tone, what to avoid, which angles to mix. This is where taste lives. |
The shape you want back
A table, ten numbered lines, a script with timestamps. Saves the reformatting pass. |
Look again at the marketing example. The weak version filled slot three only, and filled it vaguely. The strong version filled task, context, constraints and format in one sentence. That is the whole trick, and it takes about fifteen extra seconds.
Six rewrites, for jobs you actually do
Abstract advice about prompting is easy to nod at and hard to use. Here are six jobs a channel runs on, written the way most people type them, and written the way that gets a usable answer.
| What most people type | What comes back | The rewrite |
|---|---|---|
| “Give me video ideas about fitness” | Twenty topics you have seen on every channel in the niche | “List 10 video ideas for beginners over 40 restarting exercise after an injury. For each, give the question the viewer is actually typing into search.” |
| “Write a YouTube title” | A title with a colon in it and no reason to click | “Write 12 titles under 60 characters for a video about [TOPIC]. Half should lead with the keyword, half with the curiosity. Mark which is which.” |
| “Write a script about X” | An essay nobody would say out loud | “Draft a 7-minute script on [TOPIC] for [AUDIENCE]. Open on the viewer’s problem in the first two sentences. Short spoken sentences. Mark where a diagram should appear.” |
| “Summarise this transcript” | A summary that flattens the interesting part | “Summarise this transcript into the five claims the speaker makes, and flag which ones are asserted without evidence.” |
| “Make my description better” | Keyword soup with emoji | “Rewrite this description. The first two lines must work as the search snippet and contain [KEYWORD]. Then a 4-line summary, then chapters. No hype adjectives.” |
| “Why is my channel not growing?” | Generic advice about consistency | “Here are my last 15 videos with views, CTR and average view duration. Group them into what worked and what did not, and name the single pattern that separates the groups.” |
Notice how many of the rewrites hand over information rather than instructions. The last one is the clearest case: it only works because real numbers went into it. A model asked to diagnose a channel it cannot see will invent plausible reasons, and plausible reasons are the most expensive thing on this list – they feel like answers. If you want a real audit, the method matters more than the prompt, and we set that out in thirty videos in and nothing moved.
The post this came from
Watch the original reel
Fifty prompts, swipe through and take what you need – the set this breakdown was written around, posted on the BANI Academy page.
Nine levels, and where most people stop
Prompting has more depth than the five slots suggest, and it is worth knowing what sits above the level you are working at. Think of it as a ladder rather than a set of tricks.
| Level | What it is | Worth reaching for when |
|---|---|---|
| 1. Zero-shot | Just ask. | Quick factual tasks where any competent answer will do. |
| 2. Role | “You are a senior editor.” Tone and expertise shift immediately. | You want a specialist’s vocabulary rather than a generalist’s. |
| 3. Structured | Role, context, task, format – the five slots above. | Almost always. This is where precision starts. |
| 4. Few-shot | Show two to five examples of what good looks like. | You need your style copied, not a generic house style. |
| 5. Chain-of-thought | Ask it to reason step by step before answering. | The task has a middle – comparisons, diagnoses, trade-offs. |
| 6. System prompting | Set the rules once, for the whole conversation. | You are doing the same job repeatedly and retyping the setup. |
| 7. Tool use and agents | Let it act: search, summarise, save a file. | The work spans steps rather than fitting in one answer. |
| 8. Context engineering | Feed it the right documents and data. | The answer depends on your material, not on general knowledge. |
| 9. Multi-agent | Several models, one job each: research, challenge, combine. | Rarely. Worth knowing it exists before you pay for it. |
There is no prize for climbing quickly. Moving one rung – from typing a request to filling five slots – changes more about your output than jumping to multi-agent setups you will not maintain. The pattern is the same as picking software: one change, three videos, then judge it. That is the logic behind sixty AI tools, ten jobs, and it applies just as well here.
What a better prompt cannot fix
Three things stay yours no matter how well you write the instruction.
Judgement about the viewer. No prompt knows who your video is for or which promise you should refuse to make for a click. A title generator that produces twelve options still needs someone to reject the ten that oversell. We went into that specific failure in why AI titles sound like AI.
Verification. A fluent answer is a draft, not a source. The better your prompt, the more confident the output sounds, which makes the checking more important rather than less.
Disclosure. If generated material ends up on screen, platform rules apply regardless of how it was made, and those rules get rewritten. Disclosing generated visuals without losing trust covers how to handle that without deflating the video.
And if you are assembling a whole production process rather than improving single answers, prompts are one layer of it – the 12-step faceless workflow shows where they sit relative to everything else.
Frequently asked questions
Do I need a prompt pack at all?
Not really. A pack is a shortcut past the learning, which is useful on day one and a crutch by month three. Keep the ones you reuse, and rewrite the rest to fit your channel.
Do the same prompts work across different assistants?
Broadly yes, because the five slots are about specifying a job, not about one model’s quirks. Tone and length of output differ; the structure travels.
Are longer prompts always better?
No. Longer prompts are better when the extra words carry context or constraints. Padding with adjectives does nothing.
Why does the first answer often feel worse than the second?
Because the first answer tells you which slot you left empty. Treat it as a diagnostic rather than a failure, and fix the prompt instead of rerolling it.
Write one prompt properly today
Take the job you asked an AI to do most recently, and write it again with all five slots filled. Compare the two answers side by side. That single comparison teaches more than reading another list of prompts, because you see exactly which missing slot was costing you.
Then do it on your next three videos. Learning a lot does not make anyone good; doing a lot does.
If you want the structure that makes that stick, that is what our free challenge is built for: no barrier to entry, a genuine commitment to action, and a refundable commitment fee you get back when you finish. You are not paying for the knowledge. You are betting on yourself finishing, and we hold the stake.
Want the full set of fifty prompts? Comment PROMPTS on the original reel and we will send it over. Leave your name in the comment and we will reply to you directly rather than to the thread.
For more on building channels for English-speaking audiences, that is what we work on at mmoyoutube.com.
3 Comments