Written To Be Quoted

There's a piece of writing advice that ran the internet for fifteen years and is now actively costing you money.

Hook them in the first line. Build the tension. Tell a story. Make them scroll. Pay it off at the end.

That advice was correct when a human was deciding whether to keep reading. It's wrong now, and not slightly wrong. It's wrong in a way that makes your best work structurally unquotable by the systems that increasingly decide whether anybody hears about you at all.

Because a machine reading your article isn't deciding whether to keep scrolling. It's looking for a passage it can lift out, drop into an answer, and attribute to you. If your good stuff only makes sense after four paragraphs of runway, there's nothing to lift. You wrote a great article that cannot be excerpted, which in 2026 means you wrote a great article nobody's going to see.

I'm not telling you to stop being interesting. I'm telling you the interesting parts have to be able to stand up alone.

Here's how you find out where you stand right now, and it takes four minutes.

Open your best performing piece from the last year. Pick any three paragraphs from the middle of it, at random. Copy each one into a blank document by itself, with no title, no context, nothing.

Now read them as a stranger. For each one ask two questions. Does this make sense with nothing around it? And could someone tell what business wrote it?

Most people fail both on all three paragraphs. The paragraphs start with "But here's the thing" or "That's when I realized" or "So what does that mean for you?" Every one of those openings is a promise that the previous paragraph exists. Pull them out of the article and they're orphans.

That's not a writing flaw in the old world. It was good craft. Continuity is how you keep a human reading. It's just that continuity and extractability are in direct tension, and for the last two years the market has quietly repriced which one matters.

None of these will make you sound like a robot. That's the fear and it's misplaced. Most of them just make you sound more confident, because they force you to state things instead of building up to them.

One. Answer in the first sentence under every heading.

If your heading is a question, the sentence right underneath it is the answer. Full stop. Not the setup for the answer. The answer.

Bad: "How long does a bookkeeping cleanup take? This is one of the most common questions we get, and the honest answer is that it depends on a lot of factors."

Good: "A bookkeeping cleanup takes six to fourteen weeks for a business under two million in revenue with two years of backlog. Here's what moves it inside that range."

The second version can be lifted verbatim and it's still true, still useful, and still obviously yours. The first version can be lifted and says nothing.

Two. Put a number in the sentence, with its unit.

Machines weight specificity heavily and so do humans, they just don't admit it. "Significantly faster" is not a claim. "Cuts the intake call from 40 minutes to 12" is a claim.

You know these numbers. You just don't write them down because they feel too particular to be broadly useful. That instinct is exactly backwards. The particular number is the reason you get quoted instead of the generic competitor.

Three. Name the thing.

Not "a leading CRM." Not "project management software." The actual name. Not "a client in the Midwest." The industry, the size, the situation. Named entities are how these systems build the connection between your writing and the query that should surface it.

There's an obvious limit here. Don't invent client names you don't have permission to use. "A 19 person HVAC contractor in central Ohio" is specific enough to be useful and vague enough to be safe.

Four. Write the definition you assume everybody knows.

Somewhere in your work there's a term you use fifty times a week and have never once defined in public, because defining it feels condescending to your peers. Define it anyway, in one clean paragraph, in the plainest language you can manage.

Those definition paragraphs get cited constantly. They're self contained by nature, they answer a real question, and there's less competition for them than you'd expect, because everybody with actual expertise feels too senior to write them.

Five. Break the wall.

Six sentence paragraphs are a wall. A machine pulling from a wall has to grab the whole thing or nothing, and the whole thing is too long to use.

Two to four sentences. Then a break. You already know this makes it more readable for humans. It also makes each unit small enough to be usable as a unit.

Six. Put your conclusion where you'd normally put your setup.

The classic structure is context, then buildup, then payoff. Invert it. Payoff, then the reasoning, then the nuance.

You lose a little drama. You gain the thing where somebody who reads two paragraphs still walks away with your actual point, which describes almost everyone.

Beyond sentences, the architecture matters. The pages that get cited disproportionately share a shape.

They open with a two or three sentence direct answer to whatever the title promises, before any story. Then a short section that says who this applies to and who it doesn't, which sounds like a disclaimer and functions as a targeting signal. Then the substance in clearly headed sections, each of which could survive alone. Then a small block of specific facts, numbers, ranges, timelines. Then, at the very bottom, the date it was last reviewed.

That last one is small and does real work. Freshness signals matter and most business sites have no visible date anywhere, so every page looks equally old.

The section I'd push you hardest on is the second one. Who this is not for. Everybody's terrified of it. "Won't that lose me business?" It loses you the business you were going to lose anyway, later, more expensively, after two calls. And it's one of the most distinctive things you can put on a page, because your competitors will not write it. Distinctive is what gets quoted.

The objection I get here is that this requires writing constantly and nobody has time. Fair. So don't invent material. Mine it.

Pull the last ninety days of your sent email. Find every message over roughly 150 words where you explained something to a client. That's your source material and there's more of it than you think. You've already written this year's content, you just filed it in individual inboxes where it does nothing.

Take one of those explanations. Run it through whatever assistant you use with a prompt that does exactly one job:

"Below is an email I wrote explaining something to a client. Restructure it as a standalone section for my website. Rules: the first sentence must directly answer the implied question. Every paragraph must be two to four sentences. Keep every specific number, name, and timeline I used. Add nothing I did not say. If a claim in my email has no number attached, flag it at the end and ask me for the number rather than inventing one. Do not change my voice. Do not add transitions like 'moreover' or 'in conclusion.'"

That last clause matters more than it looks. Most models default to connective tissue, and connective tissue is precisely what makes a paragraph unable to stand alone. You're actively fighting the model's instinct to make things flow.

Then you edit. Every time. The model restructures, you keep it honest. It will confidently smooth "roughly six to eight weeks" into "quickly" if you let it, and that trade destroys the entire point.

If you're pulling material out of client calls rather than email, Fathom transcripts work the same way, and the spots where you explained the same thing for the fifth time this quarter are the highest value paragraphs you'll write all year.

Beyond individual paragraphs, certain small structures get pulled far more often than prose does. All three are cheap to add to pages you already have.

The range block. Three or four lines stating what something costs, how long it takes, or what it requires, in ranges with the variable named. "Two to five days for a single location. Three to six weeks if you have more than one payroll system." This gets quoted more than anything else on a typical service page, because it answers the question people actually asked.

The comparison line. One sentence stating the real difference between two options in your category, without pretending both are equally good. "A fractional controller makes sense above roughly two million in revenue. Below that you want a bookkeeper and a quarterly CPA review, and the fractional hire will cost you more than it returns." People are searching for exactly this and almost nobody writes it plainly because it feels like giving away the consultation.

The mistake list. Four or five things people get wrong in your area, each one sentence long, each stated as the mistake rather than the fix. These are highly extractable by nature and they demonstrate expertise faster than any credentials block ever has.

Writing this way makes you quotable. It does not make you findable, and confusing those two will waste your quarter.

If you have no independent footprint, no third party mentions, no citations anywhere but your own domain, then perfectly extractable writing sitting on your own site is a beautifully formatted tree falling in an empty forest. The structure work and the presence work are two different jobs and you need both.

What structure buys you is this. When something does point at you, whether that's a podcast mention or a forum thread or a trade directory, the page they land on is built to be used rather than built to be scrolled. You convert the attention you get instead of leaking it.

Take one page. Your highest intent page, the one closest to the money. Usually pricing or your main service page.

Rewrite the first three paragraphs so the first sentence answers the title directly. Add three real numbers with units. Add a short "who this is not for" block. Break every paragraph over four sentences. Put a review date at the bottom.

Ninety minutes, maybe two hours.

Then run the standalone test on it. Pull three paragraphs out at random, read them cold, and see whether a stranger could tell what you do and whether they'd believe you.

That's the bar now. Not whether it reads well top to bottom. Whether it survives being taken apart.

See you tomorrow.

Jordan Hale
The AI Newsroom