Write for the Skimmer, the Scanner and the Machine

Step 5 of 7 in The Definitive Guide to Marketing in the Age of AI (in Seven Steps). Originally published November 19, 2024. Rewritten for the AI era October 26, 2026.

Writing in the age of AI means writing for three readers at once: the skimmer who reads only headings, the scanner who reads tables and callouts, and the machine that lifts single passages into an answer. Lead every section with the answer, make every passage stand alone, and keep the persuasion principles that always worked. Then produce it with an AI-native team that has a brief, a fact check and a human who approves.

A note on proof: Proof below predates generative AI; an AI-era result replaces it by mid-December.

Who reads your content now?

Three readers, in this order. The machine often reads first, the skimmer decides whether to stay, and the scanner decides whether to believe you.

Here’s the number that should change how you write. In Pew’s study of real Google searches, the median AI summary was 67 words long (Pew). Whatever you want buyers to take away has to survive being squeezed into about one short paragraph. So write that paragraph first.

How do you write a passage AI can lift?

Treat every section like a standalone answer. Five habits do most of the work:

1. Lead with the answer. The first two sentences under a heading should answer the heading.
2. One idea per passage. If a paragraph makes two points, split it.
3. Name things. Say “Gartner’s 2026 CMO Spend Survey,” not “a recent study.”
4. Put numbers next to their scope. “45% of 645 B2B buyers surveyed in 2025,” not “nearly half of buyers.”
5. Use real HTML tables and lists for comparisons and steps, so both scanners and machines can read them.

None of this makes writing dull. A clear answer up front earns the right to tell the story underneath it.

Do classic persuasion principles still work?

Yes. People haven’t changed. The principles that made the original version of this chapter useful still hold, but each needs an honest AI-era check.

The underlying idea is the same one the old chapter started with: we decide on feeling and justify the decision with reasons afterward. Earn the feeling, then supply the proof. More on that in Why We Like Stories.

How should an AI-native team produce content?

Like a content team, not a typewriter. The mistake is asking AI for “a draft” and editing whatever comes back. The fix is to give AI the roles a good editorial team has, with gates between them. I described the method in Stop Asking AI for a Draft. Staff It Like a Content Team.

1. Brief. One page: the reader, the single takeaway, the proof you’ll use and the action you want.
2. Research. Sources opened and logged, with dates. Nothing invented.
3. Draft. Answer-first, against the brief.
4. Edit. Cut 15–25%. Check headings, rhythm and the one call to action.
5. Independent fact check. A fresh pass that didn’t write the piece opens every source and checks every claim against the proof bank.
6. Design. Diagrams and tables that teach, with alt text.
7. Human approval. A named person signs off before anything goes live.

Most teams aren’t set up for this yet. In Gartner’s 2026 CMO Spend Survey, 70% of CMOs said their internal processes aren’t mature enough to implement and scale AI, and only 30% reported mature AI readiness (Gartner). The gap is process, not tools.

What does AI expose about your team?

Everything your team never wrote down. AI is only as good as the context you give it: the positioning decisions, the approved claims, the reasons a past campaign was killed. In most teams, that lives in people’s heads and old decks. When those people leave, the context goes with them, and AI fills the gap with the average of the internet.

So the first AI investment isn’t a tool. It’s a written record of what you know and why you decided it. I make the full case in AI Won’t Replace Your Marketing Team. I have a stake in it: I’m Chief Marketing Officer of Nvitis, an institutional memory and context layer. But you don’t need any product to start. A shared proof bank and a decision log will do (see chapter 2).

What still takes a human?

Judgment, taste and experience. Design is a good example. Texas Heritage for Living, which D Custom produced, won three rounds of Content Marketing Awards: Best Use of Photography and Best Overall Design (Print) in 2017, Best Overall Design (Print) in 2020 and Best Print Publication in 2023. Those wins came from people making hundreds of small choices about what to leave out. AI can speed up the making. It can’t decide what’s worth making.

Checklist: write for three readers

• Every section answers its heading in the first two sentences
• Headings are real questions or clear claims
• One idea per passage; numbers next to their scope and source
• At least one real table and one standalone callout per piece
• A brief, an independent fact check and a named approver for every piece
• A written AI-use policy, published where readers can find it

Where to go next

Well-written content still needs to reach the places AI reads. Chapter 6 covers how to earn citations instead of chasing clicks. If you want to set up an AI content team with real gates, talk to me about building your AI content team: work with me.

AI can write the sentence. Only your team knows which sentence is true.