Key takeaways: We changed the way we write because AI search guidance told us to. We flattened introductions, front-loaded answers and organized content so machines could quote it cleanly. The writing became easier to extract and harder to recognize as ours. That raised a question we can’t shake: Did we improve the reading experience or did we give something up? This article digs into what changed and why we’re rethinking our approach.
Confession time.
I’ve been editing the life out of our writing, and I’m the one who decided we should. Here’s one example:
Sandy Asanovic, our president, sent me a draft that opened with a personal story. She’d suffered years of chronic physical glitches, including arms going numb on a walk, a nose that seemed to swell and shrink and splotches that came and went. Every specialist she saw gave the same verdict: The tests look fine.
After my edits, the article opened with a summary of what healthy-looking marketing data can hide. Her story was still there. I’d just moved it down to the third paragraph, after readers already knew the point of the piece.
I did that on purpose. I’ve been doing it for a couple of months. Now, I’m questioning whether I made the right call.
Our founder noticed before I admitted it
The first person to say it out loud wasn’t me.
Our founder told Sandy that our content didn’t sound like us anymore. It had lost something. It read as if we were trying to impress our audience. Sandy agreed. She told me it seemed like our content no longer had a soul.
Neither of them was wrong. I knew it in my gut because I was the one editing and “perfecting” the work. I was doing what marketers and search experts kept recommending — structure content so that every section is easy for AI to quote.
Those recommendations have become almost impossible to avoid. Every blog post, webinar and LinkedIn thread offers the same checklist:
- Write in plain, direct language
- Keep sentences short and punchy
- Don’t bury the facts in storytelling and long-form narratives
- Chunk content into sections that stand on their own
- Begin each section with a brief answer block that summarizes the key point
- Include FAQs to help AI find and quote content
This isn’t old advice I’m dredging up. Just this week, I read two new articles that said the same thing.
I’m not against change. That’s what makes this hard
Good writing is never finished evolving. Writers learn from the authors they admire, other writers and editors who catch what they miss. Almost any piece can be reworked into something better, and good writers never stop looking for how.
Readers drive some of that change. We’ve learned to write tighter for short attention spans, and we still write long-form content when the subject calls for it. The format follows the reader and the purpose.
Writing for AI systems is different from writing for people. If we don’t make our content easy for a machine to extract, AI search may skip right past it. That visibility now matters much like page-one search rankings once did. I’d rather stay visible than disappear. Still, I avoid flattening writing on command. I know the difference between tightening a sentence and gutting it.
Most marketers agree that we must write for people and AI, but I haven’t seen anyone figure out how to nail that yet.
We’ve been here before
This cycle is a familiar one. A new search technology arrives, marketers try to reverse-engineer it, experts publish optimization “rules” and we drive ourselves crazy making our content fit in a box. Then the search technology evolves.
Anyone who lived through Google’s major algorithm releases in the 2000s knows exactly what I’m talking about. We once edited copy to hit an arbitrary keyword percentage because experts told us that’s what Google rewarded. Other typical advice: never publish under 300 words and put the keyword in the first 100 words.
Google didn’t publish its algorithm, so an entire industry gave advice based on observation and guesswork. Some of it worked. I suspect a lot of it resulted in content written for a machine that had already moved on to the next big thing.
I’m not saying AI-search advice is wrong. But marketers and search experts are racing to decode it and harden checklists into law before anyone can confirm it works.
Back in the SEO days, we watched the metrics the advice was aimed at, including rankings and traffic, and whether the tactics paid off. What nobody tracked was what the advice did to the writing itself. This time, that’s the part I’m watching.
What the tracked changes show
I reviewed my edits to Sandy’s article about her health scare and what marketing data can hide. Her draft had a running joke in it. Describing years of specialist visits that never found anything, she wrote that every doctor delivered the same verdict: “The [INSERT TEST TYPE HERE] look fine.” That bracketed placeholder represented years of appointments with every test coming back normal, compressed into one dry line. It’s also where the article’s title came from.
I changed it to “The test results look fine.” It’s accurate, clear and easy to quote, but it stripped out the humor. I made it easier for AI search to quote the content, but Sandy’s unique point of view is gone.
Nothing in the checklist required that edit. Answer-first guidance covers where the summary goes and how sections stand alone. Jokes were never on the list. Killing that line was my own overreach. I cut voice the guidance didn’t ask me to touch, and once I saw it in one edit, I started spotting it in others. This isn’t the reviewer pile-on we’ve written about, where 11 cautious people each delete a little more. Nobody reviewed these cuts. I made them alone.
The bigger change was quieter and everywhere. Following the answer-first “rule,” I added a summary sentence to the top of nearly every section, stating the point before the story beneath it could earn it.
Sandy’s section on pitting edema opened with the image itself. A doctor presses a thumb into swollen skin and the dent lingers. By the time I was done, the paragraph opened with, “Pitting edema offers a useful way to think about stagnant marketing data.” The metaphor survived, but the storytelling hook didn’t.
A story with a conclusion bolted on top isn’t a story anymore. It reads measured, dispassionate, controlled and cool. In other words, it’s as flat as a lab report. The machine gets its clean answers. I’m not sure what readers get.
What I know and what I don’t
Our writing has definitely changed. That’s not an opinion. I’m our editor, so I know it has. What I can’t say is whether the change is a problem.
The honest case for my edits goes like this: If buyers now get their answers from AI before they visit a website, then content a machine can’t quote may as well not exist. Point of view does nothing for a reader who never finds you. Maybe the answer-first format is the new standard, the way tight writing is what short attention spans require.
The case against is the one this article keeps circling. Everything we know about answer-first formatting focuses on content discoverability. How readers react to it is a different story. If content wins in AI search but loses readers, what will we gain?
And this stopped being an editing problem a while ago. Sandy structured her health story to the guidance before I saw it. I write my first drafts this way now, too. When Sandy sends me an article she’s already flattened, her voice is gone before it hits my desk. She followed the style I set.
Where I’ve landed, for now
Anyone who has a confident verdict about writing for AI search right now is guessing.
What I have is a distinction worth testing. Some content answers readers’ questions. Think how-to guides, checklists and the formulas in Sandy’s marketing ROI articles. An answer-first approach costs that content nothing.
But thought-provoking content is different. It needs a distinct point of view, real experiences from the person writing it and a setup that doesn’t reveal everything immediately. Sandy’s health story is the perfect example. She’d lived through the thing she was describing, and readers could tell. Flatten that kind of content and the story goes with it.
So now I ask myself one question before every edit, and I’d hand it to any marketer following this guidance. Am I making a structural change the checklist calls for, or am I cutting voice the guidance never asked me to cut? The structural changes are a defensible bet on where search is going. The voice cuts aren’t. I was following a rule nobody wrote.
I could be wrong either way. Maybe flattened content will win the next era of search. Maybe machines value storytelling and personal perspective more than the checklists suggest. I don’t know yet.
That differs from the hedging our content-review article warns against. That kind of hedging abandons a position to avoid pushback. This is the opposite. I’m saying I don’t have the answer yet because I don’t.
One last thing: Except for the key takeaways, this article has no answer blocks. And even those takeaways pose a question instead of answering one. The stories come first, and the points follow. If AI systems can’t quote it cleanly, that’s fine. I didn’t write it for them.