The Human Prompt Test
One small method for using AI without handing it the steering wheel
I. The Problem
Most people open AI too early.
They have a half-formed idea, a vague discomfort, or a blank page, and they immediately ask the machine to begin. “Write me a post.” “Give me ideas.” “Make this better.” “Tell me what to think about this.”
The tool answers, because that is what tools do now. They answer even when the question is mush. Especially then. AI will happily build a cathedral on wet cardboard if you ask with enough confidence.
The problem is not that AI helps.
The problem is that AI can begin before you do.
That changes the order of authorship. The machine supplies the first shape, the first rhythm, the first assumptions, the first categories. Then you react. You edit. You accept a little here, reject a little there, and tell yourself you are still in command.
Sometimes you are.
Often, you have become the assistant.
This is how people lose their voice without noticing. No dramatic surrender. No villain music. No glowing red eye in the laptop camera.
Only a thousand tiny acts of convenience.
You ask before you have judged. You receive before you have chosen. You polish before you have made.
A person who does this every day becomes easier to steer.
The cure is small.
Before you prompt the machine, prompt yourself.
II. The Principle
Tools train the user.
A carpenter who works with dull tools develops different habits from one who works with sharp ones. A musician who practices with a metronome hears time differently. A cook who uses a knife every day learns attention through the hand.
AI also trains.
It trains speed. It trains ease. It trains the expectation that language should arrive on demand, like hot water. This is convenient, which means it is dangerous in the way sofas are dangerous. Nobody fears the sofa. Then six months pass and your spine has filed a formal complaint.
The hidden question is simple.
Who sets the aim?
If the human sets the aim, AI can serve the work.
If the machine sets the aim, the human becomes the local branch office of autocomplete.
This matters because judgment forms through repeated action. You do not become discerning by agreeing with whatever appears first. You become discerning by choosing standards before you see options.
The feed knows this.
Advertising knows this.
Bad schools know this.
A person without a prior standard can be walked around by whatever speaks fluently.
The Human Prompt Test exists to protect that prior standard.
It forces you to name what you are doing before the machine starts supplying language. It gives the work a spine. It makes AI answer to your purpose, rather than letting your purpose melt into its answer.
That is the whole game.
A tool should be held in the hand.
It should not quietly grow fingers.
III. The Method
Before every serious AI prompt, write four lines.
My goal is...
My standard is...
My constraint is...
I refuse to...
That is the Human Prompt Test.
It takes ninety seconds when you are warm and five minutes when your brain has become soup with a Wi-Fi password.
The first line names the task.
My goal is to write a clear introduction for a newsletter about parents using AI with children.
My goal is to compare three logo concepts for a small pottery studio.
My goal is to understand this biology chapter well enough to explain it without notes.
The second line names the measure of quality.
My standard is plain speech, practical examples, and no fake urgency.
My standard is a design that feels handmade, calm, and recognizable at thumbnail size.
My standard is being able to answer questions from memory, not merely paste definitions into a study guide.
The third line names the boundary.
My constraint is that the post must stay under 900 words.
My constraint is that the logo must use black, white, and one accent color.
My constraint is that I have forty minutes and cannot watch another video lecture.
The fourth line names the thing you will not surrender.
I refuse to let AI invent my argument.
I refuse to accept a style that looks like every coffee shop brand born in a beige panic.
I refuse to mistake a summary for understanding.
Now combine the four lines with your actual request.
“Using the goal, standard, constraint, and refusal below, help me improve my draft without changing the argument.”
That sentence alone will save many people from producing digital pudding.
Use the test before writing, designing, planning, studying, teaching, or making decisions. Use it when asking AI for critique. Use it when asking for options. Use it when asking for structure.
Especially use it when you feel tired.
Fatigue is when convenience puts on a little crown and starts issuing decrees.
IV. The Application
A writer sits down to draft an essay.
The lazy version is familiar: “Write me an essay about attention in the age of AI.” The machine responds with smooth paragraphs about balance, mindfulness, and navigating the modern age. Nobody has committed a crime, though the prose may be taken in for questioning.
The Human Prompt Test changes the exchange.
My goal is to argue that people should write alone before using AI.
My standard is sharp, warm, practical prose with concrete examples.
My constraint is 1,000 words for a Substack audience.
I refuse to let AI create the central argument or opening image.
Now the machine has a job. It can suggest structure. It can identify weak transitions. It can say where the example needs more detail. It may even catch a bloated sentence wandering around in a cape.
But the human remains the author.
A designer can use the same method before generating images.
My goal is to develop references for a children’s book cover about a village of swans.
My standard is old storybook beauty, clear silhouettes, and a sense of wonder.
My constraint is a horizontal cover with black, white, gold, and magenta.
I refuse to accept generic fantasy gloss.
That refusal matters. Without it, the machine will often produce something pretty and vacant, like a hotel lobby pretending to have a childhood.
A student can use the test before studying.
My goal is to understand the causes of the French Revolution.
My standard is that I can explain the sequence to a classmate in five minutes.
My constraint is that I have one hour before dinner.
I refuse to use AI to replace reading the chapter.
Now the prompt becomes: “Quiz me on the chapter, correct my answers, and ask follow-up questions until I can explain the sequence clearly.”
That is a sane use of the tool.
A parent can use it before planning a family activity.
My goal is to create a Saturday project my child and I can make together.
My standard is hands-on, low-cost, and genuinely shared.
My constraint is that it must use things already in the house.
I refuse to turn this into another screen activity.
Now AI can assist without becoming the parent.
The same pattern works for remote workers, small creators, teachers, founders, and anyone else who wants help without being absorbed.
Name the human purpose first.
Then call the machine.
V. The Warning
The Human Prompt Test can become theater.
People love turning a useful habit into a ceremonial hat. They make templates, dashboards, color-coded trackers, and seventeen-step rituals. Soon the method meant to protect judgment becomes another little bureaucracy. The soul gets a clipboard. Nobody asked for this.
Keep it plain.
The test is not a magic spell.
It is a guardrail.
The danger is not that AI answers questions. The danger is that people stop noticing which questions are theirs.
AI assistance must not become intellectual laziness. If you ask the machine to decide your opinion, name your taste, select your values, or write the first version of your conscience, you have crossed the line from assistance into surrender.
That surrender may feel pleasant.
Most bad habits do.
The same applies to style. If every piece of writing you produce sounds like a patient airport sign, stop. If every image you make looks like a glowing helmet in cinematic fog, stop. If every plan you generate has the moral texture of a corporate retreat muffin, stop immediately and seek sunlight.
Your work should carry signs of human contact.
A scar. A preference. A regionalism. A strange affection. A sentence that took the scenic path because it had something worth seeing.
The machine can assist your craft.
It cannot give you a soul on subscription.
VI. Closing
Use AI.
Use it well.
But do not arrive empty.
Before the machine speaks, place a human standard on the table.
The better prompt begins before the prompt.