State a concrete task, not a topic
The most common broken prompt is not badly worded. It is not a prompt at all — it is a subject line.
Database indexing and stuff
There is no verb here. Nothing tells the model what to do with database indexing. Explain it? Summarize it? Compare two approaches? Write a tutorial? Debug something? Each of those is a different answer, and the model has to pick one on your behalf.
Instant checks
States a concrete task
ConsiderNo clear task verb, or the prompt is very short.
Recommendation — Open with the action you want — summarize, extract, rewrite, classify — so the model is not guessing at the job.Avoids vague language
ConsiderContains vague words such as "good", "better", or "appropriate".
Recommendation — Replace judgement words with observable criteria. "Make it better" gives the model nothing; "cut it to 100 words and drop the passive voice" does.Avoids dangling references
PassedAvoids over-aggressive phrasing
Passed
Why this happens
It happens because you already know what you want. The job is fully formed in your head, and the topic is the label you have filed it under. Typing the label feels like communicating the job.
Chat interfaces make it worse. They feel conversational, and in conversation a fragment works fine — you can say "database indexing?" to a colleague and they will ask what you want to know. A model will not ask. It will guess, confidently, and produce a thousand words you did not want.
The fix is a verb
Explain how a B-tree index speeds up a WHERE clause, in three short paragraphs, for a developer who has never tuned a query.
Instant checks
States a concrete task
PassedAvoids vague language
PassedAvoids dangling references
PassedAvoids over-aggressive phrasing
Passed
The rewrite is barely longer, and every added word removes a decision the model would otherwise have made for you.
"Explain" names the job. Not summarize, not compare, not implement.
"how a B-tree index speeds up a WHERE clause" names the object precisely. Not "indexing" in the abstract — one specific mechanism, which means the answer can be specific too.
"in three short paragraphs" sets the shape, so you get an explanation rather than a treatise.
"for a developer who has never tuned a query" sets the level. This single clause changes the entire answer: it rules out both the introduction to what a database is, and the discussion of B-tree fanout.
Verbs that carry weight
Some verbs genuinely name a job. Others are decoration:
Real: summarize, extract, classify, rewrite, translate, compare, critique, debug, implement, refactor, rank, explain.
Not real: help me with, look at, think about, handle, deal with, work on. These are hedges, not instructions. They shift the decision back onto the model.
If your prompt starts with "can you help me with", delete that clause and see what is left. Usually there is a real verb hiding two words later. Sometimes there is nothing at all — and that is the useful discovery, because it means you have not decided what you want yet.
The test
Read your prompt and ask: could two competent people, given only this, produce answers that differ wildly in kind? Not in quality — in kind. One writes an essay, the other writes a checklist.
If yes, the task is not stated. Add the verb, name the object, and set the shape. It costs one sentence and it is the highest-leverage sentence in the prompt.
Every check in this article ran live in your browser. The same nine run on any prompt you paste in, for free.