Quick answer: a good outline prompt names the topic, the reader, the reader’s goal, the format and length you want, and the sources or facts the post must include, then asks the assistant to flag anything it is unsure about. OpenAI’s own guidance says to be clear and specific, give context and refine iteratively. Treat the first answer as a draft structure that you edit, not a finished plan.
Outlines are the safest place to start using AI in a blog workflow. Google’s guidance on generative AI says it can be particularly useful for researching a topic and for adding structure to original content, which is exactly what an outline does. An outline contains no facts the reader will rely on yet, so mistakes are cheap to catch. This guide shows how to write prompts that produce usable outlines, with templates you can adapt. It is part of the guide to AI tools for bloggers and creators.
The templates here are examples written for this post. They are not tied to a single tool, and they work in any chat-based assistant. Results will vary by model and by day, so test them on a topic you already know well.
What an Outline Prompt Has to Include
OpenAI’s help center on prompt engineering gives three plain pieces of advice. Be clear and specific, supplying enough context for the model to understand what you want. Refine iteratively: start with an initial prompt, review the response and adjust the wording, add context or simplify. And request a tone using descriptive adjectives. Those three habits cover most of what makes an outline prompt work.
Turn that advice into a checklist of ingredients for every outline request:
- Topic: the exact subject, in a phrase a reader would search for.
- Reader: who is reading and what they already know.
- Goal: what the reader should be able to do or decide after finishing.
- Format: how-to, comparison, list, explainer or checklist, plus a rough length.
- Constraints: sections you require, points to exclude and words to avoid.
- Sources: the primary sources the post must cite, which you will check yourself.
- Uncertainty: an instruction to mark any point it cannot support so you can verify it.
Leaving out the reader and the goal is the most common failure. Without them, the assistant falls back on a generic structure: introduction, five tips, conclusion. A prompt that says who the reader is and what they need produces headings that answer real questions.
A Basic Prompt Template You Can Reuse
Here is a starting template. Replace the bracketed parts with your own details.
“I am writing a blog post titled [working title] for [reader description]. They already know [baseline] and want to [goal]. Create an outline with an H1 and 6 to 8 H2 headings for a [how-to / comparison / explainer] of about [length] words. For each H2, give two or three bullet points of what to cover. Include a short FAQ with questions a reader would actually ask. Do not invent statistics, quotes or sources. If a point needs a fact I should verify, mark it with [VERIFY]. Use a [plain, friendly] tone.”
Several choices in that prompt matter. Asking for H2 headings with bullet points gives you structure without prose you might be tempted to paste in. The instruction not to invent statistics or sources reduces, though it does not remove, the risk of fabricated facts. The [VERIFY] marker turns the assistant’s uncertainty into a to-do list. And the tone instruction follows OpenAI’s advice to describe the tone you want.
Run the prompt, then read the outline as an editor. Which headings overlap? Which promise something the post cannot deliver? Which are filler? Cut or merge those before you do anything else.
Refine in Rounds Instead of Rewriting the Prompt
OpenAI’s guidance describes prompt engineering as iterative, and outlines show why. Your second message can be short and specific, because the assistant already has the context. Useful follow-ups include:
- “Combine headings 3 and 4, since they cover the same question.”
- “Reorder the sections so the reader hits the decision point earlier.”
- “Add a section on common mistakes, and remove the history section.”
- “Rewrite the headings so each one is a question a reader might type into a search box.”
- “Mark which sections need first-hand examples from me, since I will write those myself.”
That last follow-up is a quiet way to protect originality. Google’s helpful content page asks whether content demonstrates first-hand expertise and depth of knowledge, for example expertise that comes from having actually used a product or service. If the outline tells you where your own experience belongs, you will not fill those gaps with generic text.
Keep each round small. Changing five things at once makes it hard to tell what helped. If the assistant drifts, restate the key constraints in one line and continue from there. For a better understanding of why prompts behave the way they do, the earlier post on how ChatGPT processes prompts is a useful companion.
Check the Outline Against Real Search Intent
An assistant can propose headings, but it cannot see live search results the way you can. Before you accept an outline, search the main phrase yourself and look at what already ranks. Note the type of page, the questions answered and what is missing. Then compare it with the outline.
Google’s page on helpful content gives a useful test. It asks whether content provides substantial value compared to other pages in search results, and whether the main heading or page title gives a descriptive summary without exaggerating or shocking. Apply both to your outline. If every heading is already covered by the top results with no new angle from you, add something only you can supply, such as a specific example, a source you reviewed, or a clearer step-by-step.
Be careful with numbers. An assistant may suggest “ideal” word counts or keyword densities. Google’s helpful content page says directly that it has no preferred word count, and lists writing to a particular word count as a warning sign. Choose length based on what the reader needs, not on a figure from a chat window.
Example: Turning a Vague Topic Into a Useful Outline
Suppose your topic is “meal prep.” A weak prompt is “Write an outline about meal prep.” The result will be a generic list: benefits, tips, recipes. Now try a stronger version: “I am writing a how-to for busy parents who have never meal prepped and cook for four. They want to prep Sunday lunches for the week in under two hours. Create an outline with 6 H2 headings and bullet points, include a short FAQ, do not invent statistics, and mark any claim needing a source with [VERIFY].”
The second prompt produces headings shaped around a reader and a goal, such as choosing containers, planning the menu, ordering the work and storing food safely. Notice that a food-safety heading flags a place where you must check an authority before writing anything specific. The assistant gave you structure, and you supply the verified facts.
The same approach works on any subject. State the reader and the job, then let the assistant propose structure. This example is invented to show the method. It is not a claim about what any tool will return.
Mistakes to Avoid With Outline Prompts
A few habits make outline prompts less useful. Asking for the whole article in the same request tempts you to publish it, and it also changes what the assistant does: it starts producing prose that sounds confident whether or not it is true. Keep outlines and drafting as separate steps.
Asking for keywords is another trap. Assistants will produce keyword lists with confidence, but they do not have your Search Console data or a keyword tool’s database, and any volume numbers they offer should be treated as unsupported. Use a proper keyword source for demand, then use the assistant for structure.
Copying the outline unchanged is the third problem. Many people ask the same assistant the same question, so unedited outlines converge on near-identical structures. Google’s helpful content questions include whether you are mainly summarizing what others have to say without adding much value. Your edits, examples and sources are what make a post yours.
Finally, do not skip the facts. An outline can look authoritative while hiding errors. Once you have a structure, build the research list: each [VERIFY] item gets checked against a primary source before it goes in the draft. The next guide in this cluster, how to fact-check AI-written drafts, covers that step in detail.
A Short Workflow You Can Repeat
- Write down the reader, the goal and the working title in a note.
- Run the base template, with your details filled in.
- Read the result as an editor and cut overlapping or padded headings.
- Refine in two or three short follow-ups, using questions and constraints.
- Search the main phrase, compare with what ranks, and add a unique angle.
- List every [VERIFY] item and the primary source you will use for each.
- Write the post yourself, section by section, using the outline as a map.
That is the whole method. It takes a few minutes, and it keeps the thinking where it belongs. For image prompts, which follow similar principles, see prompting Midjourney with ChatGPT.
Adjusting the Prompt for Different Post Types
The base template works for most posts, but each format benefits from a small change. For a how-to, ask for steps in order with a short “before you start” section and a troubleshooting section. For a comparison, ask for the criteria first, then a section per option, then a “who should choose what” section, and tell the assistant to avoid ranking options without stated criteria.
For a list post, ask for the items to be grouped by purpose and for each item to include a “best for” and a “watch out for” line, which pushes the structure toward judgment and away from filler. For an explainer, ask for the question the reader is likely to have first, then the definition, then examples, then common misunderstandings. For a review, tell the assistant that you will supply the first-hand details yourself, and ask it to leave labeled gaps where your experience goes.
Whatever the format, name the reader’s level. “Beginner who has never used the tool” gives a very different outline from “experienced user looking for edge cases.” The same topic can need two entirely different posts, and the prompt is where you decide which one you are writing.
Putting Your Own Expertise Into the Outline
Google’s helpful content page asks whether content provides insightful analysis beyond the obvious. An outline is the cheapest place to add that. After the assistant proposes headings, add three things of your own: one heading it missed, one example or source you will use, and one point where you disagree with the common advice. These edits, small as they are, change the post from a generic summary into your own view.
If you have no first-hand experience of a topic, say so honestly in your planning and lean on named sources instead. Do not ask an assistant to invent a personal story. The FTC’s influencer guidance states that you cannot talk about your experience with a product you have not tried, and the same principle applies to any claim of personal experience in a post.
Frequently Asked Questions
Can I use the outline exactly as the assistant writes it?
You can, but you should not. Edit it for overlap, add the points only you can make, and compare it with what already ranks.
Should the prompt include keywords?
Include the main phrase a reader would search, in the topic line. Do not rely on the assistant for keyword volume, since it does not have that data.
How long should an outline prompt be?
Long enough to include reader, goal, format, constraints and a request to mark uncertain points. A short paragraph is usually enough.
Does it matter which AI tool I use?
The method works in any chat assistant. Results differ by tool, so test on a topic you know and judge the structure yourself.
Does Google penalize posts planned with AI?
Google says appropriate use of AI is not against its guidelines. What matters is whether the finished content is helpful, original and accurate.