Last week I was putting a proposal together for a client, the same way I always do. One of my Claude skills writes the first draft as a Markdown file. I copy it, paste it into a Google Doc, tidy up the formatting, and off it goes.

Then Claude did something I hadn't asked it to do. It created the Google Doc itself, straight into my Drive. Same proposal, same result, no copying and pasting.

I actually sat there and thought: "this is awesome. I've just got back hours of time every week. No more messing around with Markdown files."

Now, 2 or 3 minutes doesn't sound like much. But I do jobs like that 20 to 30 times a day, so it's somewhere around an hour a day that's just gone from my week. And that's one small thing. There's a lot more of my business running on Claude than that: client interviews, proposals, website mockups, follow-ups, the lot.

While I was celebrating that, LinkedIn was busy doing the opposite.

LinkedIn gave your readers a button

Back in July, LinkedIn added a button to every post and comment on the platform. It says "Seems like AI slop". By August its Chief Product Officer said it had been pressed more than a million times, and that views on content LinkedIn classes as AI slop were down 40%. That's LinkedIn's own number, so I guess it’s gospel.

Then in September it took away its own "enhance your post" rewrite tool and replaced it with a proofreader, one that suggests changes and leaves your words alone.

So on one hand, I'm running more of my business on AI than I ever have. On the other, the platform my business depends on has handed my readers a button to report me.

Fair question, then: Should I be worried?

Honestly, no. And the reason comes down to the difference between two things that get lumped together as "using AI".

What I actually do with it

I don't use AI to blindly create posts. I use it to interview my clients, get their actual words, and tidy them up. So the content is built from their lived experience and their expertise, not from scratch by a machine.

AI helps. It doesn't invent. Unless you let it, and most people do.

When someone's post gets reported as AI slop, it's most likely because there was nothing fro, them in it to begin with. No moment, no client, no number they could stand behind, no sentence only they could have written. The tool filled the gap with the same confident, generic mush it fills every gap with, and the reader could feel it.

Start from a real moment and a real opinion and the tool has nothing to invent. It only has to help you say it.

This article is the proof

You see, this article was made the way I've just described, and I want to show you exactly how.

On Tuesday afternoon a Claude skill interviewed me for about 15 minutes. The first question was "What's the moment from the last week that's stayed with you?" I gave it a lesson, something about how powerful AI is becoming. It wasn't having that. It said: that's the lesson, I want the moment. Where were you, what were you doing, what had just happened?

That's where the Google Doc story came from. And the number, 2 or 3 minutes, 20 or 30 times a day, came out of my mouth, not out of a model.

Later it asked me a question I didn't understand. "Someone in your audience is about to get this wrong. What do you tell them?" I said, “get what wrong?” It apologised, made a note that the question was too vague, and asked it again properly.

Every quote in this article is mine, word for word from that conversation. The structure, the tidying, and the checking of LinkedIn's numbers against the original announcements: that part was the machine. Then I read the whole thing and changed what I didn't like. (Yes, I’m afraid you still can’t 100% eliminate AI slop no matter how good your system is).

That's the system. Me first, machine second, me last.

Here’s Claude creating the draft Google Doc on the left, and the actual Google Doc it dropped in my Google Drive on the right.

If you're using AI for your own posts

Pay attention to 3 things, (and don’t look for a new tool or app).

First, give it the moment before you give it the topic. "Write a post about client onboarding" gets you slop. "Here's what happened on Tuesday with a new client, and here's what I thought at the time" gets you a post only you could have written.

Second, keep your own words in it. If a sentence comes out of your mouth, it should stay as it came out - even if it’s not “correct” English. The odd repetition and the odd strange turn of phrase are what readers recognise as a person.

Third, make sure you always have the last say. Read it back, change anything that doesn't sound like you, and cut anything you wouldn't say face to face to a client. If you can't find anything to change, that's a warning, not a reason to celebrate.

Bottom line: AI should save you typing. It should never save you from thinking.

The Google Doc from last week is still sitting in my Drive, by the way. It took Claude a few seconds to make. Docs like that used to take me 4+ hours and still went out with more imperfections than my perfectionist eye cares to admit.

And just because every piece of content should have a CTA...

If you'd like to know how visible you actually are on LinkedIn before you write another post, take the LinkedIn Strategy Assessment. It's short, and it tells you where you stand.
​