Ask for the Output You Need | AI Prompting Course

By the end of this lesson, you can:

Name at least five output formats and know when to use each one.
Write a prompt that specifies length, structure, and delivery format simultaneously.
Avoid the most common output specification mistakes that cause extra editing rounds.

Why format is part of the prompt, not an afterthought

When you do not specify a format, an AI assistant defaults to what it was most frequently rewarded for in training: conversational prose. Prose is fine for simple questions, but it is rarely what you need when producing deliverables for work. Consider these common use cases:

Reports & Briefings

Prose summaries often lack the scannable structure decision-makers need. A table or structured sections with headings communicates faster.

Social Media Copy

Each platform has strict character or word norms. Specifying platform + word count prevents copy that's too long to publish.

Developer Workflows

Code, config files, or data extractions benefit from JSON, YAML, or code block output — not a prose explanation of what the JSON would look like.

Training & Guides

Step-by-step instructions work best as a numbered list so readers can follow along without re-reading.

The six most useful output formats

Master these six formats and you will cover the vast majority of professional output needs:

Format 1

Markdown table

"Return a Markdown table with columns: Feature | Benefit | Example."

Best for comparisons, specifications, and any multi-attribute dataset.

Format 2

Numbered step-by-step list

"Give me a numbered list of exactly 7 steps, each under 25 words."

Best for tutorials, SOPs, onboarding guides, and recipes.

Format 3

Bullet-point summary

"Summarise the key findings in 5 bullet points. Each bullet: max 2 lines."

Best for executive briefings, meeting notes, and slide decks.

Format 4

Structured prose with headings

"Write 3 sections with H2 headings: Introduction, Key Findings, and Recommendations."

Best for reports, blog posts, and long-form documentation.

Format 5

JSON or YAML

"Return the extracted data as valid JSON with keys: name, date, amount, category."

Best for developers, data processing, and integration workflows.

Format 6

Q&A or FAQ pairs

"Generate 5 FAQ pairs (Question: / Answer:) about our return policy below."

Best for help centres, chatbot training data, and knowledge bases.

Controlling length: the three levers

Length control prevents responses that are far too long (burying your key insight) or far too short (lacking necessary detail). Use whichever lever matches your need:

Word or character counts: "Under 120 words" or "no more than 280 characters." Use for social copy, emails, and brief summaries.
Item counts: "Exactly 5 bullet points" or "a numbered list of 7 steps." Use when you need a fixed number of actionable items for a slide or checklist.
Section depth: "One sentence per section" or "2–3 paragraphs per heading." Use for long-form documents where each section needs a consistent level of detail.

Three common output specification mistakes

Even experienced users fall into these traps. Avoid them by adding a short output block to every complex prompt:

Mistake 1 — Assuming a default: Writing "give me a summary" without specifying length or format. The AI will choose, and it often chooses a long paragraph when you wanted three bullet points.
Mistake 2 — Mixing two formats in one instruction: Asking for "a table and a short description" without saying which comes first, or how long the description should be. Separate them into numbered instructions.
Mistake 3 — Omitting structural anchors: If you need the output to start with a specific phrase (e.g., "Our recommendation is…") or end with a call to action, state that explicitly. The AI will not infer it.

Illustrative example

Prompt Makeover

A startup founder needs a competitive analysis of three AI writing tools to share with investors. Watch how specifying output format turns a wall of prose into an investor-ready document.

Original prompt

Compare Jasper, Copy.ai, and Writesonic for me.

What is missing

Format: table, bullet list, prose paragraph, or numbered ranking?
Criteria: what dimensions matter (price, features, ease of use)?
Length: how detailed should each entry be?
Audience: is this for a technical team, investors, or personal research?

Improved prompt

Act as a SaaS product analyst. Compare Jasper, Copy.ai, and Writesonic for an investor presentation.

Return a Markdown table with these exact columns: Tool Name | Monthly Price (Starter Plan) | Best For | Key Limitation | Verdict (one sentence).

Below the table, add a 3-sentence executive summary starting with "For a bootstrapped startup focusing on..."

Do NOT include feature counts or technical API details.

Why it is better

It requests a named format: Markdown table with exact column labels.
It defines the audience (investors) so the AI calibrates its language.
It appends a structured prose section with an explicit opening phrase.
It uses negative constraints to remove irrelevant technical detail.

What still needs checking

Verify current pricing on each tool's website — AI training data has a knowledge cut-off.
Confirm the verdict sentences reflect your actual product strategy, not a generic one.

Practice

Spot the prompt that specifies format and length

A content manager wants to repurpose a blog post into short social media copy. Read both requests and select the one that fully defines the output format, platform, and length constraints.

Illustrative example

Repurposing a 1,200-word blog post titled 'Five habits of highly productive remote workers' for two social media channels.

Which prompt will produce usable social content on the first attempt? Option A: Turn my blog post into social media posts for LinkedIn and Twitter. Option B: Rewrite the blog post below into: (1) One LinkedIn post — professional tone, 150–200 words, ending with a question to encourage comments. (2) Three Twitter/X threads — each tweet under 280 characters, no hashtags, conversational tone. Separate each platform's output with a clear heading.

Check the model answer

Option B specifies platform, format, length, and tone.

Option B defines output for two separate platforms with individual word/character limits, tone requirements, and structural separators. Option A leaves every one of those dimensions undefined — the AI will guess, and you will spend extra time editing the result.

Knowledge check

Why is specifying output format in your prompt more effective than reformatting the AI's response after you receive it? Option A: It forces the AI to produce shorter answers. Option B: It reduces editing time and avoids a second round-trip prompt to reformat. Option C: AI assistants ignore format instructions and always respond in prose.

Check the answer

Option B is the correct answer.

Specifying format upfront (table, numbered list, JSON, etc.) means the output arrives ready to copy-paste or share. Reformatting after the fact costs extra editing time or requires a follow-up prompt — both of which are avoidable.

Keep this

The Output Specification Formula

Add these three output controls to any high-stakes prompt before you send it:

FORMAT: [Table / Numbered list / Bullet points / JSON / Markdown / Plain prose]
LENGTH: [Under X words / X bullet points / X sentences / X paragraphs]
STRUCTURE: [Section headings / Column labels / Opening phrase / Sign-off]

Example:
Return a Markdown table with columns: [Column A | Column B | Column C].
Below the table, write a 2-sentence summary starting with "[Required opening phrase]".

Combine FORMAT + LENGTH + STRUCTURE for the most predictable results. You only need all three for complex deliverables — a simple list just needs LENGTH.

Optional browser progress

Your progress

0 of 5 available lessons completed

0 of 5

Progress stores only the course identifier and completed lesson identifiers in this browser. It does not save your answers, prompt text or identity, and it does not sync to another device. Clearing browser data removes it.