I’ll read up on AEO or GEO, get a solid grasp of what it is, take notes, and feel prepared. Then I open a blank document to write a post, and that clarity disappears. Because it’s suddenly not a concept anymore, but a decision: Right now, in this sentence, who am I writing for?

Do I structure this for AI extraction?

Do I let the narrative breathe the way THT content usually does?

If I write the way I normally write—thinking out loud, letting a little frustration leak through, throwing in a line that’s just me—will that hurt my chances of being found?

And if I strip all of that out to optimize cleanly, will the post even sound like it came from this blog?

That’s the second-guessing that takes over me. It’s, honestly, what pushed me to write this post.

We’ve covered the technical side of this shift at THT before. So far, we’ve seen how GEO is the practice of optimizing content so that large language models like ChatGPT, Claude, and Perplexity cite it as a trusted source.

Then we went into AEO, which covers how to structure content so AI-powered search features extract and surface your answers directly.

Both posts get into the what and the how. This post is about the part that came after for me. It’s the moment you close the guide, open a doc, and realize you still don’t know how to write.

Not just for AI. Not just for humans. But for both—without losing yourself in the process.

We’ll also address the concern of sounding robotic and resembling everyone else.

When AI assists in writing across the whole internet, a certain sameness starts to leak through. You’ve probably noticed it as phrases, structures, or a kind of polished vagueness that shows up everywhere. You read something and immediately clock it. There’s nothing unique that you couldn’t find elsewhere.

It’s a fear that your content won’t just sound like a machine, but that it’ll sound like every machine-assisted piece of content out there 🙈

So, when you sit down to write, who are you actually writing for? That’s the question this post answers.

Why Most Writing Advice for the AI Era Misses the Point

Most guides about AEO and GEO focus on the what:

• What schema markup is
• What a featured snippet looks like
• Or what AI Overviews do to your click-through rate

That context matters; you should take time to understand it.

But the confusion most content creators feel is about what happens when those concepts meet the act of writing.

You understand, intellectually, that AI systems prefer direct answers, question-based headers, and structured content.

You also know that your audience connects with your voice—the narrative texture, the way you think out loud on the page, the phrases that sound unmistakably like you.

Both things are true. Both things feel important. So which one wins?

This is where a lot of people freeze. Or worse, they try to serve both audiences at the same time without a clear strategy, and end up with content that’s half-structured, half-narrative, and not fully satisfying to either reader.

The problem isn’t that you have two audiences. The problem is not knowing which one is primary for this specific piece.

Figure that out first, and the rest of the decisions fall into place.

The Two Readers at Your Table

When you publish anything today, two fundamentally different types of readers consume it.

Each reader wants something completely different from you.

1. The Human Reader

Your human reader is who you’ve always written for. They skim your headers, get pulled in by a line that resonates, and re-read a paragraph because it puts words to something they’d been feeling.

They decide whether to trust you based on feeling as much as fact.

They notice when your voice disappears.

They’re the reason personality matters because content without personality is just information, and information alone doesn’t build loyalty or keep anyone coming back 💁‍♀️

2. The AI Reader

Your AI reader operates completely differently. It’s not experiencing your content—it’s processing it.

It doesn’t care about the line that made a human stop and smile.

It’s looking for clear entities, direct claims, and structured information that it can extract and cite with confidence.

It rewards clarity, consistency, and credibility signals. It has no patience for preamble (aka noise it needs to parse through) 🤖

Who’s Your Reader?

Here’s your takeaway: these two readers are not equally important for every piece of content you create.

Treating every post like it needs to satisfy both audiences equally is what creates paralysis.

Some content is fundamentally human-first. Some is AI-first. And a lot lives in the middle, but even then, one has a slight edge.

Knowing which one is doing the most work for a given piece is the thing that actually makes writing decisions feel less like a guessing game.

The Audience Decision Framework: Matching Your Content to Your Primary Reader

Before you write a single sentence, ask yourself: What is the primary job of this piece?

That answer tells you who to write for first and how much to bend toward the other audience.

When Your Human Reader Comes First

Some content is fundamentally human-first. These are pieces where voice, narrative, and emotional resonance aren’t decorative—they are the product.

Pushing that content hard toward AI extraction is like turning a handwritten letter into a bullet-point summary.

It’s technically the same information, yet not even close to the same thing.

Write primarily for humans when you’re creating:

Opinion and Perspective Pieces

Don’t think research alone. This is your take on a trend, a conclusion you’ve landed on after sitting with a problem, a position you’ve formed through experience.

Here’s mine: 10 Things People Get Wrong About My Developer Job

AI can summarize what others think. It cannot generate a perspective that didn’t already exist somewhere on the web.

A real perspective isn’t just “here’s my opinion.” It’s grounded in something you’ve actually lived.

At THT, that might look like: I spent three hours chasing a layout bug that turned out to be a single misplaced flex property, and here’s what that taught me about how I was mentally modeling CSS.

That’s a perspective. It’s specific, it’s earned, and it’s not something an AI can fabricate from pattern-matching.

Another example: you’ve been writing technical tutorials for two years, and you’ve noticed that readers don’t drop off at the hard parts but right before them, in the setup sections.

That observation, from your own data and your own blog, is a perspective. It has a point of view that came from doing the thing.

That’s what human readers click through to find. They want the mind behind it, not just the conclusion.

Narrative-Driven Tutorials

When you’re walking a reader through a project, and the story of how you figured it out is woven into the instruction, that narrative texture is doing real work.

It’s what makes the reader feel like they’re building alongside you instead of executing a manual.

In practice, that looks like a line mid-tutorial where you admit: “I tried doing this with a forEach loop first, and it immediately fell apart—hair-pulling event, honestly. Here’s why.”

You’ve seen me doing it all the time.

Check out: The Ultimate Guide To Re-Engineering My Portfolio’s RAG Chatbot

Or a quick aside with an emoji that signals you’re still a person in here: “That’s the fix—and yes, I did spend 40 minutes ruling out everything else first 😅”

That internal dialogue leaking onto the page is what separates a THT tutorial from a documentation page. Both can teach. Only one builds a reader who comes back.

You can—and should—structure these well:

• Use clear headers
• Make sections extractable

But the human thread running through the instruction is the reason someone chooses your post over the one above it in search.

Personal Reflection and Career Content

These are posts about what you’ve learned, what surprised you, and what you got completely wrong.

They build the kind of trust that turns a visitor into someone who actually follows the blog.

No AI overview creates that relationship.

Community-Building Content

This includes anything where the goal is connection over information transfer.

If you want someone to comment, share, or feel genuinely seen, then write like a human, for a human.

Note 🔥
“Human-first” doesn’t mean ignore AI entirely. Your voice and narrative are the priority, and you layer in structural clarity without sacrificing them by using clear headers and direct section openers, without letting the optimization eat the personality.

When Your AI Reader Comes First

Other content types are naturally better served by prioritizing AI-readability. These are pieces where the user’s primary need is a fast, clear answer.

In such cases, being the most structured, direct source for that answer is your real competitive advantage.

Write primarily for AI when you’re creating:

Definitional and Comparison Content

It looks like this:

• “What is X?”
• “What’s the difference between Y and Z?”
• “How does A compare to B?”

These queries are now heavily dominated by AI answers. If you want visibility here, your content needs to be extractable. This means it needs a direct answer in the first sentence or two, question-based headers, and structured data where possible.

You’ve probably already seen what this looks like without realizing it.

Search “what is AEO” right now, and the AI Overview at the top will likely read something like: “Answer Engine Optimization (AEO) is the practice of structuring content so AI-powered search tools can extract and present your answer directly, without requiring a user to click through.”

Clean, complete, standalone.

The page that earned that citation opened that section with exactly that kind of sentence—not a paragraph of warm-up.

Step-by-Step How-To Guides

When someone searches “how do I center a div with CSS Grid” or “how do I set up environment variables in Node.js,” they want the answer with minimal friction.

Being the clearest, most structured source gets you cited even when the user never visits your site.

This is also worth thinking about if your content is anything like recipe writing.

Recipe sites have long been notorious for burying the actual recipe at the bottom of a long page, underneath a personal story about the dish, a history of the ingredients, and three paragraphs about a summer in Tuscany.

Maybe that’s why I do not cook 🙄

Readers scroll straight past all of it. Every time.

AI does the same thing. It finds the recipe and extracts it, and the story above it mostly disappears.

If that’s your content model, AI-first structure is a survival question, not just an optimization one.

FAQ-Style Reference Content

This one’s self-explanatory: documentation, glossaries, and troubleshooting guides.

These are built for extraction by design, so every question and answer is already a self-contained unit.

Product and Tool Comparisons

This includes structured tables, feature breakdowns, and side-by-side evaluations.

All is highly extractable, and exactly the kind of content AI surfaces well.

Tip 💡
A quick gut-check for AI-first content is to see if you can answer the primary question in two direct sentences followed by a structured list that would fully satisfy the reader. If it’s a yes, then that’s AI-first content. Wrap it in some human warmth, but don’t let narrative slow down the answer.

When You’re Writing for Both (And How to Find the Balance)

Some pieces of content serve both audiences well, such as a project walkthrough that teaches something real while being structured enough for AI to extract key takeaways.

Or, an in-depth guide that has a distinct voice and uses question-based headers with clear section answers.

This is the sweet spot, and it’s achievable but requires knowing which audience is slightly more primary, so you know what to protect when you have to make a call.

So how do you actually know?

I’d say it’s mostly instinct built from writing consistently over time—especially writing before AI tools were part of the process, when the feedback loop was simpler.

After enough posts, you develop a feel for what a reader needs versus what you wanted to say.

I’ll be honest, I’m still working this out too.

Part of why I wrote this post is that researching AEO and GEO gave me a lot of guidance on technique and left me with a whole new kind of confusion about implementation 😬

So if you’re in that same spot, you’re not behind. You’re just at the part where the reading ends and the figuring-out begins.

Analytics can help over time. If you consistently see readers drop off at a particular section, that’s a signal.

If a post with more narrative consistently outperforms a leaner one on the same topic, that’s a signal too.

But early on, you’re mostly going on intuition, and it gets sharper the more you write.

What I keep coming back to: protect the thing your primary audience would actually miss if it were gone.

What Is “THT Fluff” and How Do You Spot It?

Before I get into the practical test, it’s worth defining this term clearly because I’m about to use it, and I need you to understand its meaning.

“THT fluff” is the writing that accumulates when you’re performing a personality rather than expressing it.

It’s the paragraph you added because it felt satisfying to write, but a reader skimming for the answer could skip entirely without losing anything.

It’s the sentence that transitions from one idea to the next without actually adding information (“With that in mind, let’s move on to…”) or the one that restates what you just said in slightly different words because it felt like a good place to land.

It’s not the same as the THT voice.

A line like “hair-pulling event” earns its place because it tells the reader something real about the experience. It’s a specific emotional shorthand that makes them feel like they’re in the room.

“It’s worth noting that…” earns nothing. It’s throat-clearing on the page.

The distinction matters because the fix is different. While THT fluff gets cut, THT voice gets protected.

How to Tell the Difference in Practice

I recently added FAQs to a post because it felt like the right AEO move.

When I read it back, it felt off—not because the FAQs didn’t fit the topic, but because they felt alien to how I write.

They covered the right questions and were clear. But it felt like I’d made a left turn into a different type of content entirely, like I was suddenly writing documentation when I’d been writing a tutorial.

That foreignness was the signal.

It’s not that the FAQ sections are wrong. It’s that appended FAQ sections feel bolted on, and readers sense that immediately.

The difference between one that works and one that doesn’t is whether it grew naturally from the post or got dropped onto the end of it.

The same goes for all AI-readability tactics. Question-based headers can feel mechanical or genuinely conversational.

For example, compare:

“Content Optimization Considerations”

To:

“Does Structuring for AI Mean Sacrificing Your Voice?”

Direct answers can read like a corporate knowledge base or like a friend who skips the preamble. The tactics are neutral. The key is how you execute them.

When you’re not sure which side of the line something lands on, ask: Does this add something a reader couldn’t get from any other article on this topic?

Something like your specific experience, honest confusion, the line that’s unmistakably how you think. If yes, keep it. That’s not fluff.

That’s the thing that makes your content worth reading over the twelve other posts on the same subject.

If no, it’s a generic connector, a restatement, a paragraph that’s there to make the post feel complete—cut it.

That’s fluff, and clean writing isn’t robotic writing. It’s writing that doesn’t waste the reader’s time.

Remember, personality phrases and internal dialogue—“hair-pulling event,” “I sure did,” “that got interesting fast”—are not fluff, and they’re not automatically filler either.

It depends on whether they’re doing something.

When they surface your experience or bring a moment to life, they’re load-bearing. They’re how you bring your internal dialogue onto the page and make something technical feel as if it came from a real person.

When they’re just decoration—filler between ideas, a reflexive “honestly” attached to nothing—they can go.

The question isn’t whether your personality belongs in the post. It always belongs.

The question is whether that particular expression of it is earning its place.

Tip 😌
If it helps, think of the words you choose to put on a page selectively as though they’re tokenized and you pay for every single one of them while on a budget. Would you waste a single token on that unnecessary “honestly” at the end of the sentence?

Your Unique Voice Is an Asset to AI Systems, Not a Liability

Something that surprised me is that the qualities that feel most at odds with AI optimization—your voice, perspective, tendency to think out loud on the page—are actually what make content more valuable to AI systems, not less.

AI can pull structured answers from thousands of sources. What it cannot do is locate your specific account of debugging a problem for forty minutes and discovering the issue was a missing comma.

It can’t replicate the paragraph where you admit you tried the wrong approach first.

It can’t synthesize a perspective that doesn’t already exist somewhere on the web.

This is what E-E-A-T—Experience, Expertise, Authoritativeness, Trustworthiness—is measuring.

The “Experience” pillar specifically looks for evidence that a real person with real involvement wrote this.

In a web full of AI-generated content that’s technically accurate but experientially empty, that signal matters.

The Gray Area Worth Acknowledging

AI can be prompted to write in a person’s voice. It can be trained on someone’s existing content and produce something that sounds, on a first pass, reasonably close.

It can assemble experiences from real accounts across the internet and present them as first-hand.

So when E-E-A-T rewards “first-hand experience,” the question becomes: how does anyone verify that?

Right now, the honest answer is: imperfectly. The detection tools are still catching up.

In a moment when there’s still enough human-written content on the web for AI to train on, the gap between real experience and convincing fabrication isn’t always obvious.

What tends to leak through over time, though, is specificity.

Fabricated experience tends toward the general. It describes categories of feeling rather than particular moments.

Real experience has texture, be it a specific error message, a particular feature that behaved unexpectedly, or the exact reason you tried approach A before discovering approach B.

That level of detail is harder to manufacture at scale. The more particular you are, the harder your experience is to replicate.

This is still a gray area, and I don’t think anyone has fully solved it. Writing with genuine specificity—not just “I ran into a problem” but what problem, when, and what it felt like to get through it—is both good content and the clearest signal you can send that the experience is real.

Your authentic voice, written with specificity, isn’t competing with AI optimization. Done with intention, it is the optimization.

So, protect original observations and first-hand accounts. They’re the hardest thing to replicate and the most valuable thing you have.

What to cut instead:

• generic transitions like “In addition,…” or “Furthermore,…”
• obvious restatements like “The main point is that…” or “In other words,…”

These add length and subtract trust.

A Content Decision Checklist Before You Write (or Revise)

Run through this before you start or before your next revision pass:

Step 1: What is the primary job of this piece?

• Answer a specific question clearly → AI-first
• Share a perspective, build a connection, or teach through a story → Human-first
• Teach through a project while building authority → Both, with human slightly primary

Step 2: What does my primary audience need most?

• Human reader: voice, narrative arc, the sense of a real person behind the post
• AI reader: direct answers, structured sections, extractable claims

Step 3: What can I not cut without failing that primary audience?

Identify and protect that. Everything else is negotiable.

Step 4: Have I given the secondary audience enough to work with?

• If human-first, are my headers clear? Does each section open with something direct?
• If AI-first, is there enough personality that a human reader doesn’t feel like they’re reading a manual?

Step 5: Is there something on this page that can’t be found anywhere else?

Original experience, honest admission, specific data, a take that’s actually yours.

If no—that’s the gap to fill before publishing. Not with more words but with more you.

It’s a Wrap

I won’t promise that the second-guessing will fully go away. But it gets quieter once you stop treating every post as something that needs to serve every reader equally all the time.

Know who you’re primarily writing for.

Protect what that audience needs.

Give the other audience enough to work with.

And when you’re on a revision pass, wondering whether to keep the paragraph that sounds like you or cut it because it feels indulgent, ask whether it’s adding something real.

If it is, keep it. That’s not fluff. That’s the whole point.

Hope this helps you navigate the write-for-humans vs. write-for-AI tension!

See ya next time.