Why AI Hands You a Wall of Text (and the Format Fix)
- Tomasz Dylik
- Jul 19
- 3 min read
Updated: Jul 31
You ask a clear question and get back a dense grey paragraph you now have to mine for the actual answer. The information is in there somewhere. You just have to do the work of digging it out, which is the work you were trying to hand off in the first place.
That's a Format problem, the fourth input in the CROFTC framework. Format is the shape of the answer, and when you don't specify it, you accept whatever shape the model defaults to. Its default is prose, and prose is rarely what you actually need.
What Format actually is
Content is what the model says. Format is how it arrives: a five-bullet list, a table, a subject line plus body, three options labeled A, B, and C, a short intro under 150 words. Same information, but the shape decides whether you can use it in ten seconds or have to reshape it yourself.
The model doesn't withhold structure on purpose. It just has no idea what you're going to do with the answer, so it hands you the safest container it knows: a paragraph. If a paragraph isn't how you'll use it, that's a container you have to unpack by hand.
Why skipping it quietly costs you
Format is the input people forget because the answer looks finished. It reads fine. But "reads fine" and "usable" aren't the same. A wall of text hiding three options is worse than a short list that shows them, even if the wall is better written, because you can act on the list immediately.
Every time you copy an AI answer and then spend two minutes cutting it into a list, pulling out the one line you needed, or turning a paragraph into a table, that's a format you should have asked for up front. It's small, it's constant, and it adds up.
How to specify it well
Ask yourself where the answer is going, then describe that shape. Going into a slide? Ask for three short bullets. Comparing options? Ask for a table with columns. Sending as a message? Ask for a subject line and a body under 80 words. Be concrete about structure and length.
Here's a prompt with no format:
What should I consider when choosing an email platform?
And the same request with a format:
Format: A table with three columns: factor, why it matters, and one question I should ask a vendor. Keep it to the five most important factors. What should I consider when choosing an email platform?
The first gives you a lecture. The second gives you a decision tool you can take straight into a sales call. Same knowledge, but one of them is ready to use and the other is homework.
Want prompts with the output shape already set? My free bonus pack includes 500+ ready-to-use prompts plus the one-page CROFTC cheat-sheet, each built to return something you can use, not reshape. Get the free prompt pack.
Format delivers the rest
Format is how the other five inputs reach you. Context, Role, Objective, and Tone decide how good the answer is; Format decides whether that good answer arrives in a form you can actually use, and Conditions keep it honest along the way. A brilliant answer in the wrong shape still costs you time. The right shape is what makes it effortless.
If you change one habit, make it this: before you hit enter, picture where the answer is going and tell the model that shape. You'll stop reformatting AI output and start pasting it straight in.
CROFTC is the structure behind every one of the 1,350+ prompts in my series, and Format is what makes each one land ready to use. If you want a running start, grab my free bonus pack: 500+ ready-to-use prompts plus the one-page CROFTC cheat-sheet, so you can stop digging through paragraphs and start getting answers in the shape you need.
Format is the quickest of the six to fix, which is exactly why it is easy to stop there. The other five work the same way, and I put all six in one place, with the prompts already written, in AI Prompt Engineering Bible (7 Books in 1). Seven books in one volume, each input taken apart and rebuilt, with prompts you can copy and adapt to your own work.
The full CROFTC series
This is one input in a six-part framework. Here's the whole series:
Comments