The Ideas Are Mine. The Typing, Increasingly, Is Not.

Substack just gave you a new toy, and I’d like to be the first to hand you the instructions. As of this month, you can scan any post on this platform (including this one) and see an estimate of how much of it was written by a human versus written with AI. It’s powered by a detection tool called Pangram; it takes about two seconds, and if you run it on my recent posts, the results will raise an eyebrow.

So let me raise it for you.

Yes, I use AI to help write this Substack, including, with full appreciation of the irony, this very post about using AI to write this Substack. I dictated my long-winded thoughts on it the way I’d talk something through with a colleague over coffee, had AI shape that into a first draft, and then rewrote it until it sounded like me. Which is to say: exactly how I write everything here.

I want to explain why, not defensively, but because I think the distinction I’m about to draw is the one this conversation is missing.

I’ve contributed to major media publications (and written my own articles for some of them) long before AI existed. I got an A in AP English and a degree in journalism. I have spent more than a decade building a career on being able to communicate… whether that’s to describe a hotel, a wine, a place, or a feeling, in a way that makes people want to go. If I had nothing but time, I would happily write every word of this newsletter by hand, slowly, with a glass of something good nearby.

But I don’t have nothing but time. I run two businesses. I lead a team of 15 people. I plan travel for clients across multiple continents, often simultaneously. What I have in abundance is the thing that actually makes a newsletter worth reading: opinions, recommendations, lived experience, and a point of view earned over years of doing this work. What I have in scarcity is the two uninterrupted hours it takes to turn those thoughts into polished prose.

AI didn’t give me ideas. It gave me back the time to share them.

Take my recent piece on the economics of a hosted stay. That post came entirely from my own experience: my numbers, my observations, my opinions about how this corner of the industry works. It’s a piece I’m proud of, because it came straight from my years of lived experience: not research or aggregation. I’m not claiming the sentences are works of extraordinary genius I crafted from nothing, but I’m confident in the knowledge behind them that took me years to acquire.

Run it through the scanner, though, and it would likely come back as a majority “AI-assisted.” And technically, that’s true as far as the sentence structure or connective tissue is concerned. AI very much helped me with it. But the scanner is measuring syntax, not substance. It can tell you who arranged the words. It cannot tell you whose ideas they are, whose taste shaped them, or whose reputation stands behind every recommendation.

That’s the chasm. A percentage treats writing as one activity, when it’s really two: having something to say, and then saying it. AI can obviously do a great job when it comes to assisting the second. In my opinion, the issue is when people want it to do the first.

I understand why this tool exists, and I don’t hate it. The internet is filling up with content made by no one (or by humans who lack their own first-hand opinions so they borrow some from AI). That erodes trust for all of us, and a platform built on the relationship between writers and readers has every reason to protect it. If the scanner makes it harder to fake having a perspective, I think that’s good.

My quibble isn’t with transparency. It’s with a number standing in for a conversation. A percentage flattens “this person outsourced their thinking entirely” and “this person dictated twenty minutes of over a decade of expertise and used AI to organize it” into the same score. To me, those are not the same thing. One is a shortcut around having ideas. The other is a shortcut around typing.

Substack, to its credit, seems to know this… which is why alongside the scanner, they’re encouraging writers to explain how they actually work. So, consider this my explanation.

Another point of clarification: the wariness around AI and writing is about media literacy, and I share it. When AI is used to generate news coverage, investigative reporting, or anything where the facts themselves are the product, the risks are real, and the scrutiny is deserved. Nobody should want their understanding of the world assembled by a machine with no one accountable behind it.

I would argue that that’s not what this is. I’m not delivering breaking news or reporting on matters of consequence. I’m telling you the name of the restaurant in Paris I can’t stop thinking about, or trends I’ve observed from my own experience. The stakes here are whether you have a wonderful dinner or share my taste. If AI helps me get those sentences to you faster, with every opinion still mine & every recommendation still personally vetted, the risk profile feels closer to a friend texting you her favorites than to a newsroom cutting corners. Different categories deserve different levels of caution, and I’d put mine firmly in the low-stakes column.

One more distinction, in fairness to the writers this argument doesn’t cover. There are essayists and humorists on this platform whose product is the prose. You subscribe because of how they say things, not just what they say. For them, the scanner measures something that matters greatly. If you read someone for their sentences, you’re entitled to know whose sentences they are.

But I’d guess that’s not why you’re here. You’re here for my judgement. That part (the taste, the vetting, the decade of showing up in person) has no shortcut, and no scanner can measure it either way.

Every recommendation in this newsletter is mine. Every opinion, every hotel, every itinerary, every strongly held view about what makes travel worth doing… all mine, drawn from my own decade (plus some) of experience. I will never publish a recommendation I haven’t personally stood behind, because my entire business runs on exactly that.

What AI does is help me get there faster. I talk; it drafts; I rewrite. The voice you’re reading has been edited back into every paragraph by hand… including this one.

If that arrangement bothers you, I understand, and the scanner is there for you to use. But my hope is that most of you care about the question the scanner can’t answer: is there a real person here, with real expertise, telling you something true? There is. She’s just learned to delegate the typing.

Both things can be true. The ideas are entirely human. The efficiency is not. And I’ve decided I’d rather tell you that myself than have a percentage imply something murkier.

The truth is, AI gave me back the one resource none of us can renew: time. Every hour I don’t spend typing is an hour I can spend out in the world: in the hotels, the conversations, the places themselves… or here, sharing it all with you. What I do with that time has always been the human part. I’m just glad I finally have the hours to share it.