You’re Not Writing for People Anymore
You hate em dashes. I get it, I just don’t think you know what that means.
In 2005, seeing keywords crammed into a blog post might’ve made you cringe. That’s fine. But refusing to use them? That could’ve tanked your search ranking.
Fast forward to 2025, and we’re watching the same mistake repeat, just with new tools. The danger now isn’t ignoring keywords. It’s ignoring how AI understands and retrieves information. Essentially, if you don’t speak AI’s language, your content might never be seen at all.
From Ranking to Relevance: Why SEO Rules No Longer Apply
We’re living through the quiet death of SEO. This year, Google rolled out AI Mode as the default experience for some users. No more blue links. No more ten results. No incentive to click through.
As Mitty Chang recently observed, the traditional search results page is being replaced by an AI-first interface. Welcome to the Answer Economy, a world where SEO looks and functions differently.
Traditional SEO rewarded keywords, backlinks, and clicks. Generative Engine Optimization (GEO) rewards structured clarity. If your content isn’t logically organized, semantically rich, and easy to parse, the machine will skip it.
You’re no longer writing to be found. You’re writing to be quoted.
Andreessen Horowitz put it bluntly in their recent analysis on the rise of GEO:
“We’re not optimizing for ranking anymore. We’re optimizing for retrievability inside a machine’s reasoning process.”
– You Can’t Outwrite the Format: Why Em Dashes Win in an AI World
This isn’t just a shift in formatting, it’s a shift in the fundamental nature of discoverability. You’re not writing for crawlers anymore. You’re teaching the model how to think.
How AI Thinks (and Why It Doesn’t Care About Your Clever Turn of Phrase)
LLMs like ChatGPT, Claude, and Perplexity don’t browse the internet the way Google does. They’re trained on vast datasets, and they generate answers based on semantic structure, not keyword matching or backlink strength.
If you want your content to be retrievable in this new ecosystem, it has to be machine-legible. That means:
. Clear headings that define ideas
. Bullet points that structure information
. Transitional language like “in summary” or “the takeaway is…”
. Paragraphs that deliver complete, standalone insights
LLMs aren’t impressed by cleverness or stylized voice if it obscures the point. They’re not grading your prose—they’re scanning for logic, clarity, and extractable meaning.
Ideas
If you’re worried you’ll run out of things to talk about, don’t be. There’s a wealth of podcast-friendly topics that your audience will want to tune into. Here are a few to get you started:
. FAQ Fridays: Dedicate an episode each month to answering common client questions like, “Do I need a lawyer for this?” or “What’s the worst that could happen if I ignore a lawsuit?” (Hint: Not a great idea.)
. Client Stories: Sharing anonymized case stories can bring your practice to life and show potential clients how you’ve successfully handled situations similar to theirs.
. Special Guests: Bring in other professionals—financial advisors, real estate agents, or even other attorneys. It’s a great way to keep things interesting while providing well-rounded insights.
If It “Sounds Like AI,” It’s Probably Ready for AI
People love to critique “AI-sounding” writing. But maybe what they’re reacting to isn’t something written by AI—it’s something written for AI.
Writers who avoid formatting conventions or plain summaries because they “feel robotic” often end up with content that’s invisible to machines. In trying to sound more human, they’ve made themselves unfindable.
The goal isn’t to write like a robot. It’s to write clearly enough that a robot can quote you.
Being human is good. But being legible to machines is what makes you visible.
The irony? Writers are afraid of sounding like AI when the real risk is being ignored by it.
Remember Keywords? Refusing to Use Them Was a Mistake Too
Back in the early 2000s, marketers who thought they were “above” keywords lost rankings and reach. The algorithm didn’t care how poetic you sounded. It cared whether you matched its structure.
We’re in the same place now. Just with a more advanced engine.
Ignoring the preferences of AI because they feel “impersonal” is the modern version of refusing to use keywords. And just like back then, it’s a losing strategy.
Style without structure doesn’t get read, and voice without formatting doesn’t get quoted.
The Future Isn’t Clicks—It’s Citations
You’re not competing for rankings anymore. You’re competing for quotability.
The model decides what to synthesize, what to surface, and what to ignore. If you’re not training it to use your content, you’re training it to forget you exist.
The best content doesn’t just inform.
It formats ideas so they can be absorbed, echoed, and trusted.
That might mean using em dashes. Starting with “in short.” Leading with the takeaway. These aren’t quirks, they’re signals. Signals that the content is built for generative engines.
It’s not about flair, it’s about function.
So What Now?
If you’re still writing for Google-era SEO, here’s where to start:
1. Audit for machine readability. Restructure your content with clear headings, bullet points, and meaningful summaries.
2. Write to teach, not to rank. Prioritize semantic clarity, not search bait. Think like a trusted source, not a traffic trap.
3. Test your visibility in generative engines. Use prompts or emerging GEO tools to see whether AI is quoting your work, and fix it if not. Don’t just ask: “Is this ranking?” Ask: “Would an AI quote this in a helpful answer?”
Final Thought: You’re Not Just Writing To an Audience—You’re Writing Through a Model
In the past, content connected you directly to a reader. In today’s ecosystem, it connects you to a model, and that model connects you to everyone else.
That means every line, every structure, every format is a decision: Will this be used, or forgotten?
You don’t have to love the em dash.
You just have to use the language the machine understands, because content that isn’t legible to AI might as well not exist anymore.