Output Format Cookbook
About this resource
The most common daily friction with these tools is not answer quality. It is shape. You wanted three bullets and got four paragraphs, or you wanted a table and got prose describing a table. This is entirely fixable and almost nobody is taught how, because format control gets one line in most prompting guidance.
This cookbook gives you the instruction for each output shape you actually need: tables with specified columns, structured summaries at a stated length, comparison layouts, ranked lists with criteria, drafts with a defined structure, and output clean enough to paste into a document without reformatting. Each entry is a copy-and-paste line you append to whatever you were already asking, so nothing here requires rewriting your prompt.
Every recipe is a line you add to the end of a prompt you have already written. None of them require you to start over.
How to use it
Pick the two shapes you need most often and commit those to memory or to your custom instructions. Format control compounds: once your output arrives in the right shape by default, the reformatting time disappears entirely, and that recovered time is usually larger than whatever the drafting saved in the first place.
Inside the download
With the columns you named, in a form you can paste straight into a document.
Why word counts work better than "short" and how to phrase them.
A stated shape rather than whatever the model defaults to.
Side by side against criteria you specify.
Common questions
Name the columns explicitly and say you want a table rather than a description of one. Specify how many rows, what each column contains, and what to do when a value is unknown. Vague requests for "a table" produce prose with headings roughly half the time.
Give a number rather than an adjective. "Under 120 words" works reliably; "short" does not, because short means something different in every context the model has seen. For lists, specify the count. For sections, specify how many sentences each should run.
Instruct it explicitly: no preamble, no restatement of the question, no closing offer to help further, and no summary of what it just said. These habits come from training that rewarded thoroughness, so they persist until you say otherwise.
Free. No email list, no follow-up sequence.
More from the resource library, chosen because they pair well with this one.
A decision tree for fixing a disappointing answer: diagnose the symptom, then change one thing.
Eight reusable prompt structures that transfer to work you have not encountered yet.
Tokens, context windows, training cutoffs, and why these systems produce confident wrong answers. One page, kept beside a monitor.
Tell us a little about yourself and this document is yours. We don't have an email list. We collect this so we know who's finding our resources useful, and that's it.
You're downloading
this document
Your information is kept confidential and never shared with anyone.
