What's the shape?

Answer in the first two lines. Then why it's true. Then how to do it. Then one line summarising the whole page.

That's it. It takes about twenty minutes per page and it's the highest-yield edit available to a small business right now, because it serves two readers with one change.

Most business pages do the opposite. Three paragraphs of context, a paragraph establishing credentials, then — halfway down, in the middle of a sentence — the actual answer. That structure made a kind of sense when a human was scrolling and you were trying to hold them. It makes none at all when the first reader might be a machine deciding whether there's a liftable chunk near the top.

Most business pages bury the answer three paragraphs down. Invert the page, write subheads as questions, and one edit serves both skimmers and machines.

Why does answering first work on both readers?

Because skimming humans and extracting models want the same thing, which is unusual and worth exploiting.

A person landing on your page has one question and about eight seconds. If the answer is in the opening lines, they either get what they came for and think well of you, or they realise it isn't for them and leave — both good outcomes. Burying it produces neither; it produces a bounce with no impression left behind.

A model assembling an answer is looking for a self-contained passage it can quote without distorting. A paragraph that begins mid-argument, or depends on three paragraphs above it, is hard to lift cleanly. A paragraph that states a complete claim in two sentences is easy.

So the same edit improves the human experience and the machine one. That is not usually how optimisation works — normally you trade one against the other — and it's the main reason this is worth doing before anything else in the theme.

The Princeton-led GEO study found visibility lifts of up to around 40% from content changes, with statistics, citations and quotations the strongest levers. Structure isn't itself one of those levers — but it's what makes them reachable. A sourced statistic buried in paragraph nine is a statistic nobody lifts.

How should the subheads be written?

As questions a buyer would actually type. Not as labels.

"Our Process" is a label. "How long does this take?" is a question. The second one wins twice: a human scanning the page finds their exact question, and a model matching a query to a passage has something to match against.

Four that earn their place on almost any service page:

• How long does this take? With a real range, not "it depends". If it genuinely depends, say what it depends on — that's the more useful answer anyway.
• What does it cost? A band, a starting point, or an honest explanation of why there isn't one. Pages that dodge this entirely read as expensive and evasive.
• What do I need to have ready? Nobody writes this and everybody wants it.
• Who is this not for? The line competitors never write, because it feels like turning away work. It does the opposite: it's the single most credible sentence on most websites, and it's the one that gets quoted, because it's the only sentence that couldn't have been generated from the category.

That last one is the same instinct as saying what AI does in your business before anyone asks. Volunteering the limit is what makes the rest believable.

What about the closing line?

End with one sentence that summarises the whole page, written so it makes sense read alone.

This is the cheapest thing in the entire theme and the most often skipped. A standalone summary line is the passage most likely to be quoted verbatim, because it is the only part of the page that doesn't need the rest of the page. Write it last, write it plainly, and resist making it a call to action — a sales line won't get lifted, a clear claim will.

If you've been reading these decodes for a while you'll recognise the shape: it's the same discipline as the one-pager that doubles every AI answer, turned around. That post was about writing a brief clear enough for a model to work from. This is about writing a page clear enough for a model to quote from. Same skill, opposite direction.

What this won't do

It won't rescue a thin page.

Structure makes a good page findable. It does not make a weak one worth finding, and there is a version of this advice circulating that quietly implies otherwise — reformat your way to visibility, no new substance required. That isn't what the research says and it isn't what happens.

If the honest answer to "what does this page know that a general article doesn't" is nothing, the fix isn't the shape. It's having something to say: a number from your own work, a limit you've learned, a definition you'd defend. That's Friday's post, and it's the harder half.

Start with your single best page — the one that already earns you work. Twenty minutes. Answer first, questions as subheads, one standalone line at the end.

For the prompts that make this rewrite fast — free, 5-min read: 7 AI Prompts That Run a One-Person Business.

Related Decodes

• Be the Answer: How a One-Person Business Gets Found When Customers Ask AI Instead of Google — the week's hub; this is move two of five.
• Ask the Machine What It Says About You — do this first, so you know which page to rewrite.
• The One-Pager That Doubles Every AI Answer — the same clarity discipline, pointed at your inputs.
• The 10-Minute Website Copy Audit — three prompts for finding which page to fix.

AI the Brief (@aithebrief) — AI advice you can act on and trust, for time-starved founders.

TL;DR: Invert your best page — answer in the first two lines, then why, then how, then one standalone summary sentence at the end. It works because skimming humans and extracting models want the same thing: a self-contained passage near the top, which is unusual enough to be worth exploiting since normally you trade one reader against the other. Write subheads as questions a buyer would type rather than labels: how long does this take, what does it cost, what do I need ready, and who is this not for — that last one is the most credible sentence on most websites and the one most likely to be quoted, because it couldn't have been generated from the category. Close with a plain summary line that reads alone; don't make it a call to action, because sales lines don't get lifted and clear claims do. Structure makes a good page findable, but it will not rescue a thin one — if the page knows nothing a general article doesn't, the fix is substance, not shape.