Specificity: defining the task and removing ambiguity

Say what you actually want

Compare these two requests:

• "Can you look at my essay?"
• "Check my essay's introduction (pasted below) for unclear sentences. List up to five sentences that a first-year student might misunderstand, quote each one, and suggest a clearer version. Do not rewrite the rest of the essay."

The second prompt is longer, but it removes guesswork. It states the action (check for unclear sentences), the scope (the introduction only), the deliverable (a list of up to five quoted sentences with suggestions), the reader whose perspective matters (a first-year student) and a boundary (do not rewrite the rest).

Text description

A comparison of Vague: “Can you look at my essay?” and Specific rewrite. Action. Vague: “Can you look at my essay?”: “Look at”: the model must guess. Specific rewrite: Check for unclear sentences. Scope. Vague: “Can you look at my essay?”: The whole essay, by default. Specific rewrite: The introduction only. Deliverable. Vague: “Can you look at my essay?”: Anything at all. Specific rewrite: Up to five quoted sentences, each with a clearer version. Reader. Vague: “Can you look at my essay?”: Not mentioned. Specific rewrite: A first-year student. Boundary. Vague: “Can you look at my essay?”: None. Specific rewrite: Do not rewrite the rest of the essay.

Four questions that make a task specific

Before sending a prompt, answer:

1. What is the action? Use a precise verb: summarise, compare, classify, rewrite, extract, list, critique, translate, draft, plan, explain. "Help with" and "look at" are not actions.
2. What is the deliverable? A paragraph? Three options? A table with named columns? A 5-step plan?
3. What is in and out of scope? Which part of the material, which time period, which audience, which options?
4. How will you judge success? "Each suggestion must be under 20 words." "Every claim must cite the paragraph it came from." Success criteria guide the model and give you something to check.

Ambiguity: words that mean different things

Many everyday words are ambiguous to a model:

Pronouns cause trouble too. In "Compare the plan with the budget and fix it", it could be either. Name the thing.

Positive instructions beat vague prohibitions

"Don't be boring" gives the model little to work with. Say what to do instead: "Open with a concrete example from a small business." Prohibitions are still useful for real limits ("Do not invent statistics; if a figure is not in the text, write 'not stated'"), but pair them with what you do want.

Specific does not mean long

Specificity is about precision, not word count. A 40-word prompt with a clear action, deliverable and criterion beats a 300-word prompt that rambles. Remove anything that does not change the output — you will practise this in Module 4.

Worked example

Goal: prepare for a job interview.

1. Vague: "Help me prepare for an interview."
2. Action and deliverable: "Write 8 likely interview questions for me to practise."
3. Scope: "…for a junior data-entry role at a small logistics company."
4. Success criteria: "Include at least 3 behavioural questions ('Tell me about a time…'). For each question, add one sentence on what the interviewer is trying to learn."

Final prompt: "Write 8 likely interview questions for a junior data-entry role at a small logistics company, for me to practise. Include at least 3 behavioural questions ('Tell me about a time…'). After each question, add one sentence explaining what the interviewer is trying to learn. Present them as a numbered list."

Notice how each addition answers one of the four questions.