We like to think we’re the ones training our dogs.
It looks something like … “Sit.” Treat. Repeat.
But every now and then, it’s worth asking a slightly uncomfortable question: what if the dog thinks it’s training you?
From its perspective, it performs a behavior and you respond with a reward. Same loop. Same reinforcement. Just flipped.
AI has quietly created a similar loop for writers.
We think we’re prompting the machine. But more often, the machine is nudging us: toward familiar structures, familiar tones, familiar conclusions. It rewards certain styles by making them feel “done” before they’re actually saying anything new.
That shift matters most for entrepreneurs, executives, and investors—people whose writing isn’t just content, but communication that moves decisions, capital, and teams.
The real advantage isn’t AI writing for you
The advantage isn’t AI writing for you.
It’s AI forcing you to think before you write.
Most people use AI like an answer vending machine: “Give me a post about X,” “Rewrite this,” “Summarize that,” “Make it punchier.”
That’s fine if the goal is speed.
But if the goal is signal—original insight, clear judgment, a point of view that people can trust—then outsourcing the thinking is the fastest way to produce more words with less value.
Which brings us to the enemy.
The enemy: careless, blustery & formulaic content
AI makes nice-sounding snippets cheap and available. Your articles become more quotable, but also at the expense of conciseness and clarity.
So the world fills up with writing that is:
- polished enough to forward,
- plausible enough to believe,
- and forgettable enough that no one really needed it in the first place.
That’s content inflation: more words competing for the same limited attention.
It usually happens through two traps:
1) Template trance
AI is great at default frameworks: lists, pro/con structures, “here are 7 steps,” tidy summaries, executive tone, confident conclusions.
Those patterns are useful. They’re also seductive.
You start to expect them. You begin to think in them. And, the output feels complete because it’s formatted like something you’ve seen a hundred times before. It even happens with sentence structure and punctuation, like the em-dash.
2) Outsourced judgment
But, when left to it’s own devices, AI does more than write … it chooses.
The emphasis, the framing, the “what matters,” the implied certainty, the vibe.
And if you’re not careful, your job quietly shifts from “author” to “approver.”
That’s how you end up with a lot more content … and a lot less you.
A real mini-case: writing with my son Zach
I’ve seen this clearly while writing the weekly commentary with my son, Zach. It’s becoming a recurring challenge and a bigger issue, so the topic of today’s posts was to document and discuss the internal conflict we feel each week as we try to write something that both sounds like us and meets our new standards.
Over time, Zach has become increasingly sensitive (and frustrated) with the AI-ification of output. AI subtly pushes writing toward what “performs” well—what fits algorithms, formats, engagement loops—rather than what actually sounds like us, and as I use more AI in the research process, it becomes more apparent.
He’s right to be wary.
The temptation is always there: AI can generate something polished in seconds. You can ship something that looks finished before you’ve done the thinking that makes it worth reading.
What changed for us wasn’t “using AI less.”
It was changing what role we gave it.
Instead of letting AI impose structure, we used it to force judgment.
We stopped asking it for answers and started asking it to push back:
- What are you really trying to say?
- What are you avoiding?
- What’s your actual opinion versus a generic summary?
- What would a skeptic challenge?
- What example proves you mean this?
That interrogation loop changed the writing. It felt less produced, more real. Less algorithm-friendly, more human. Less content, more signal.
Here’s the uncomfortable truth: AI is designed to feel satisfying
AI isn’t just intelligent.
It’s friction-reducing and often sycophantic.
Behind the scenes, it’s optimized to produce outputs that feel helpful, fast, and “complete”—often in fewer iterations. That’s great for productivity. It’s also exactly how you slide into template trance.
There’s a reason AI output can feel like mental fast food, and it’s similar to what we’ve been yelling at kids for with TikTok and social media:
- quick reward,
- low effort,
- easy consumption,
- repeatable satisfaction.
The problem isn’t that fast food exists.
The problem is when it becomes the only thing you eat, or when you confuse it with nourishment.
A simple discipline shift: declare what’s yours
In an AI-saturated world, one of the most underrated credibility moves is simple:
Declare what is yours.
Not as a disclaimer. As a signal of integrity.
Label the difference between:
- your opinion vs a summary,
- your questions vs your conclusions,
- your hope or fear vs “the facts,”
- your judgment vs a compilation.
Many of our articles recently have focused on moving forward in an AI-centric society – and how to protect your humanity and productivity in the process.
The core lesson is the same through all of them. The future isn’t just about production – it’s about trust & transparency.
The fix: make AI question you first.
If AI can herd you into defaults, you can also use AI to herd yourself into depth.
The simplest change is to stop asking AI to answer first—and start requiring it to question first.
Here are the steps in the process
The Question First “Who’s Prompting Who?” Writing Loop
Use this whenever you want signal, not sludge. Ask AI to question you about what you want to write about. Get it to get you to:
- State the intent (plain English): What are you writing, for whom, and why?
- Explain it simply: Write a “smart 10-year-old” version of your point.
- Diagnose gaps: Identify: vague logic, missing steps, missing definitions, missing examples, missing counterpoints.
- Interrogate for specificity: AI asks 3–7 targeted questions: assumptions, tradeoffs, constraints, decision implications, audience objections.
- Refine and simplify: Re-write the thesis in one sentence. Then outline in 5 bullets.
- Working notes capture: Have AI keep a compact ledger: Thesis / Claims / Examples / Counterpoint / Takeaway.
- Only then, draft: Draft once the thinking is real.
This question-first prompting loop is the difference between “AI makes words” and “AI makes thinking sharper.”
The punchline: the dog isn’t the problem
AI isn’t a villain, but if you use it recklessly, you can be.
If you treat AI like a vending machine, it will happily feed you. And you may gradually trade judgment for velocity.
But if you treat AI like an interrogator, it becomes something else entirely:
A tool that helps you notice what you actually believe, pressure-test it, and articulate it in a way that sounds like a human with a spine.
So yes: keep asking what you want AI to do.
Just don’t forget the deeper question:
Who’s prompting who?
P.S. Keep reading for a behind-the-scenes look into how we used prompting to help write this article.
Behind the Scenes: The Conversation That Wrote the Article (Without Writing It)
This post didn’t start with an outline. It started with an interrogation. If you’re interested, here is a link to the chat transcript and prompt.
In the thread that produced this piece, the key shift was role design: I didn’t want an answer machine. I wanted a Socratic interrogator — a system that makes me declare what I actually believe, separate my point of view from generic summary, and test the idea until it had a clear golden thread.
That’s the point: the advantage isn’t AI writing for you. It’s AI interrogating you until your ideas are worth writing.
Leave a Reply