Fourteen prompt templates published by AI Central share one design principle: the template supplies the reasoning structure, and you supply the specifics through bracketed slots. Most of them do not ask ChatGPT to write anything. They ask it to critique, interrogate, compare, translate or predict. Two of the fourteen produce a finished artifact. The other twelve produce a better decision. That distinction is the whole point of the set.
Reviewed August 7, 2026. The source is a first-party AI Central document, 14 ChatGPT Essential Prompt Templates, held in the library. Everything here is drawn from it.
The bracket is the interface
Every one of the fourteen is a sentence with holes cut into it. Complex topic. Field I know well. Specific expert. Option A and Option B. Recipient, context, tone, length. The brackets look like placeholders and are easy to read straight past. They are the working part.
The fixed text around them does something a normal request cannot. It sets the model's stance, the shape of the answer, and the standard that answer has to meet, before the model knows anything about your situation. You then drop the situation in.
A prompt written from scratch tends to be all specifics and no structure. You describe the problem at length and finish with a vague ask. These templates invert that, which is why twenty words of template beats two hundred words of description.
Much of the set exists to make the model argue with you
The most striking pattern across the fourteen is how many are engineered against the assistant's default manner. A model tuned to be helpful is also tuned to be agreeable and reluctant to tell you your draft is thin. Several templates attack that head on.
The second one, the Brutally Honest Critic, assigns the stance first and then fires three questions in a row that a polite answer cannot survive.
What’s weak? What’s unclear? What’s missing? Be direct
That line is AI Central's. Notice it asks for absence, not quality. What is missing is the thing a model almost never volunteers, because producing text is far easier than auditing text that is not there.
The Strategic Idea Stress Test runs the same play on an idea rather than a draft. The Weighted Decision Matrix runs it on a choice.
I need to decide between [Option A] and [Option B]. Create a decision matrix with weighted criteria and surface hidden assumptions
AI Central's phrasing rewards a close reading. Weighted criteria forces the model to declare what it thinks matters most, which you can then dispute. Hidden assumptions forces it to name the beliefs the comparison rests on. A request for pros and cons produces neither.
AI Central's Deep-Dive Expert Interview reaches the same place through a different door.
You are a [specific expert]. Answer my questions with real-world tradeoffs, constraints, and what’s overrated or underrated about [topic]
Tradeoffs, constraints, overrated and underrated are all words that force a position. Ask an expert what is good about a topic and you get a brochure. Ask what is overrated and you get an opinion.
The reverse brief flips who is asking
The fifth template is the one most worth stealing, and the only one in the set that forbids an answer up front.
I want to achieve [goal]. Before giving advice, ask me 5 clarifying questions about constraints, resources, timeline, and priorities
AI Central calls it the Goal-Oriented Reverse Brief. Advice given without constraints is generic by necessity, so the template withholds it until the constraints exist. It even names the four that usually decide the answer.
There is a second, less obvious effect. Answering five clarifying questions forces you to state things you had never made explicit to yourself. Plenty of people finish the questions and no longer need the advice.
Negative constraints carry as much weight as positive ones
A surprising share of the instruction across these templates is about what not to do. The Intelligent Concept Explainer closes on two clauses that pull in opposite directions on purpose.
Don’t dumb it down. Make it accessible
That is AI Central's wording, and the tension is deliberate. Without the first clause, a request for accessibility gets you an explanation pitched at a child. The negative clause fences off the exact failure mode the positive clause invites. The move repeats throughout.
- The email template ends with an explicit avoid list, alongside recipient, context, tone and a word limit.
- The multi-channel repurposer sets the output counts, then insists the core message stays intact.
- The analogy generator asks for accurate and memorable in the same breath, so the model cannot buy vividness with distortion.
If you take one habit from the collection, take that one. Name the way the answer usually goes wrong in the same sentence where you ask for it.
Analogy is treated as a comprehension tool, not a flourish
Two of the fourteen are built entirely on analogy and a third carries it as a clause. The sharpest version asks for analogies from a field the reader already knows well, which makes your own expertise the raw material. A lawyer and a chef learning the same idea should not get the same explanation. The bracket tells the model where your knowledge sits, so it builds from there rather than from zero.
Only two of the fourteen produce something you could send
Sorting the templates by what each one outputs makes the bias obvious. I grouped all fourteen using nothing but the stated output in each template's own wording.
- Two produce a finished deliverable: the email speedrun and the multi-channel content repurposer.
- Four produce comprehension: the concept explainer, the jargon translator, the code explainer and the analogy generator.
- Two produce critique: the harsh critic and the idea stress test.
- Six produce analysis or a decision: the expert interview, meeting preparation, the reverse brief, research synthesis, the decision matrix and second-order thinking.
On method, that is a straight read of each template's stated output, and reasonable people could move one or two across a line. The jargon translator ends by asking for three smart questions, which is closer to analysis than comprehension. The direction of the split is not in doubt though. Twelve of the fourteen exist to change what you understand or decide, not to save you typing. The value here is not output volume, it is that each template installs a thinking procedure you would otherwise forget.
The one line that keeps paying out
The last template is a single sentence and the cheapest of the fourteen to run.
If [event] happens, what happens next? Then what happens after that? Continue
AI Central files that under strategic second-order thinking. The word doing the work is Continue. It turns a one-shot question into a chain, and chains are where the non-obvious consequences live. The first order effect of a price rise is obvious to everyone. The third order effect decides whether the rise was a good idea.
How to actually use the set
Fourteen templates is small enough to learn and large enough to sit unused. A few things make the difference.
- Pick three, not fourteen. The reverse brief, the harsh critic and second-order thinking cover most of the ground.
- Fill every bracket before sending. An empty slot collapses the template back into the generic request it was built to avoid.
- Chain them. Reverse brief to define the problem, decision matrix on the options it surfaces, stress test on whatever you picked.
- Keep the negative constraints when you adapt the wording. They are the first thing people delete and the reason the template worked.
The document's own closing instruction is blunter than any of that.
Don't wait for inspiration
A few questions come up whenever someone starts working through the set.
Do these prompts only work in ChatGPT?
The document is written for ChatGPT and that is the tool it names. Nothing in the fourteen depends on a ChatGPT-only feature though. They are plain instructions built from role, structure and constraint.
What if I do not know what to put in the brackets?
That is what the reverse brief is for. It asks five clarifying questions about constraints, resources, timeline and priorities before giving you anything, and your answers tell you what belongs in the other slots.
Which one should I try first?
The Brutally Honest Critic, on something you have already written and are slightly proud of. It is the fastest way to feel the difference between an agreeable model and a useful one.
Where this comes from
The source document is 14 ChatGPT Essential Prompt Templates, published by AI Central and held in the library alongside this article. All fourteen prompts are stored with it in full, ready to copy.