“But here’s the thing.” 

The (somewhat circuitous) route to today’s post began when I saw those four words on a LinkedIn list of 10 phrases said to prove AI wrote something. It caught my eye because I use it often — my way of telling the reader, “pay special attention here, I’m about to turn the conversation somewhere unexpected.”

It got me thinking how irritating I find those lists of arbitrary AI “tells.” So I decided to push back by building my own — “10 Ways to Know Frank Wrote This.” Better still, I’d get AI to help me put it together.

So I turned to Claude, took advantage of the brief window of access to its Fable 5 model, and asked it to read all 213 Monday Memos I’ve written and compile the list.

What Claude Came Back With Really Surprised Me

That’s where the “this is getting interesting” part comes in — the gap between what you expect and what shows up. I figured I’d get a vaguely amusing catalog of my quirks and pet phrases.  Maybe it would work as a parody of those “I can tell if AI wrote it” lists. Maybe not.

There’s some of that in what Fable produced. But the list is more than that. It’s a perceptive take on the strengths and weaknesses of my Monday Memos.

Let me share it exactly as Claude produced it. Then we’ll get to what matters for your writing: not the list itself, but what you can do with one of your own.

10 Ways to Know Frank Wrote This

1. It opens with a story that has no business being in a nonprofit memo. Hemingway eating oysters alone in Paris. Kristofferson pitching a song about a stoned Sunday morning. Springsteen rewriting his way out of a dropped record contract. You spend three paragraphs wondering where it’s going. It always ends up at the reader’s next appeal.
   
   
2. There’s an Elmore Leonard quote within 500 words. If not Leonard, then Zinsser, Kay Lautman, or Lisa Cron. “Try to leave out the parts readers tend to skip” has run a dozen times — and it’s never the part anyone skips.
   
   
3. The credentials arrive disguised as anecdotes. He never says “as an expert.” Instead, his mother appears with FDR, JFK, and Jesus in the Garden of Gethsemane on her living room wall. Jesse Jackson waves off his radio script. The Godfather of Gore spares his ASPCA letter. You realize the receipts only after you’ve enjoyed the story.
   
   
4. The word “verisimilitude” is used affectionately. He hates $100 words – except this one, which he keeps like a pet and walks past the neighbors every few months.
   
   
5. The em-dashes are his. He was leaning on them decades before ChatGPT made them a suspect class – and he isn’t giving them back.
   
   
6. Two-thirds of the way in, the ground shifts. The common advice is right “as far as it goes” — then he tells you where it stops going. Getting attention isn’t enough; be an architect of it. If there’s no pivot, Frank didn’t write it.
   
   
7. The greatest hits are on rotation. Tension, not villains. Lead with emotion, back it up with facts. “Who are you going to be,” not “what are you going to do.” Each has run at least six times. He’d call predictability a crime against craft and repetition how teaching works. The line between them sits one more usage away.
   
   
8. A behavioral science study shows up with numbers attached – and a Monday move. Snickers sold 38% faster with an anchor. Hotel towels reused at 26% higher rates. Fruitcake rebranded as “native Texas pecan cakes.” Then, always: here’s what it means for your ask string.
   
   
9. There’s at least one typo. A memo preaching that “rewriting is where the game is won or lost” will occasionally advise you to “right a message.” The arguments get rewritten five times; the spelling rides in coach.
   
   
10. It ends with “Hopefully, these ideas will help…” The one sentence in the memo that the memo itself would tell you to cut.
    

What a List Like This Is Actually For

Here’s the part I almost missed and the whole reason I’m sharing this. I first took the list as an amusement: “ha, the em-dashes, guilty.” What it turned into was far more strategic and actionable.

And notice what Claude did and didn’t do. It didn’t write a memo for me or think for me. It held up an accurate picture of how I already write and let me decide what to keep. That’s the relationship I’d recommend: not a ghostwriter, a devil’s advocate. Not an AI to think for you, one to stress-test your thinking.

In short, treat it like a junior editor’s notes: examine them carefully, keep what helps, ignore the rest.

Of course, the real test is acting on the insights I agree with. Here’s one illustration.

Acting on What the List Reveals: An Example

Let me show you, using the least flattering line on my list — number ten, my weak “hope this helps” endings.  

Claude is right. It’s a loose way to end memos I’ve worked hard to make thoughtful and persuasive. Here’s the fix, from a recent memo on rewriting.

Before Seeing the List: “Hopefully, this memo has convinced or reminded you how essential it is to give rewriting proper weight and attention in your creative process.”

After: “Here’s what to do on your very next assignment: take a hard look at your review and approval process and find where you’re giving rewriting short shrift. 

If it’s squeezed out for lack of time, shift your schedule. If your writer protests every revision, declare that unacceptable. Too many cooks? Decide who’s in charge. Zinsser is right: rewriting is “where the game is won or lost.” Make sure your process and priorities reflect that.

#1: Gather the evidence. You don’t need 213 memos — just fifteen or twenty pieces that represent how your organization actually communicates: a strong appeal, a weak one, a few emails, a foundation proposal, a landing page. Range matters more than volume.

#2: Ask for the profile. Tell Claude to read the material and identify how a reader could tell your organization wrote a message: phrases you favor, ways you engage the reader, patterns in how you present information — and, most important, the flaws that keep recurring.

#3: Read the flaws first. The flattering findings are pleasant, but it’s the weak spots that deserve your attention.

#4: Calendar out how you will act on the insights. Don’t just file them away for someday. Make a specific plan to tackle the weak spots, starting with the one having the biggest impact.

#5: Keep AI as a devil’s advocate and junior editor. Don’t use it to write for you or think for you. Use it to stress-test what you’ve already written — the way a junior editor would.

If you already use Claude as a devil’s advocate, the “10 Ways to Know” exercise is a strong addition to your routine. And if your new to using AI review in your process, it’s a great place to start. 

My prediction: You’ll soon find other uses: editor’s notes on a single piece, a read on what separates your strongest work from your weakest, a side-by-side comparison with communications from other groups in your field.

But every path starts the same way. Don’t just nod at your own 10 Ways list and file it away — pick the flaw that’s costing you the most and fix it in the very next thing you write. Do that, and you’ll have turned an amusing list into a really useful tool to strengthen your writing.