I created a Writing project in ChatGPT. 

I love the model’s project function. I have a Home Maintenance project where I’ve uploaded the make, model, serial number and purchase information of all my appliances (allowing me to burn all the manuals). When I ask it to diagnose why the dryer isn’t spinning it knows what dryer I’m talking about and even how old it is. 

I have a Health project where I store all my medical and broader health data. After catching a misdiagnosis, it offered to write a letter to my doctor requesting specific tests, and it nailed the tone perfectly. 

I have many friends using AI to write in their voice, and they rave about the results. So I thought I would spin up a Writing project to see just how good it might or might not be. 

Prepping The Model

I paid to have my writing voice profiled by answering a series of questions and uploading writing examples. I added this voice profile in the form of an extensive markdown file. I instructed the model on how to use this profile.

I uploaded markdown files of my three books. I referenced my publicly available work (blog posts, with certain parameters, and podcast transcripts).

Then I asked it to write this week’s post for me, on lead generation. It came back with a series of questions that I was able to answer succinctly. They were good questions around the audience, core ideas and tone (even though it has extensive instructions on tone). 

I started broad on purpose, expecting it to ask these types of questions. So far, so good. Once answered it came back with something that was interesting, and readable, but felt more like ChatGPT than me.

But I kept working with it.

“Reference my framework The Ladder of Lead Generation,” and “reference my post Attending The Way.”

“Write longer sentences with more flow.”

After many iterations I got something that someone who doesn’t write for a living might find to be “useful” content. In fact, part of me worries it would be more useful than what I would write.

There are even some great lines I might use after a little tweaking:

“If your positioning is generic, AI will help you scale generic faster.”

“If your conviction is thin, AI will help you broadcast thin conviction more widely.”

I wouldn’t use the word “generic” to talk about positioning, although it’s perfectly valid. And I would choose another word over “conviction” even though it too is used appropriately.

I know I haven’t “trained” the model on my writing, I only have it reference it, but I’m surprised the output doesn’t sound more like me, even after a bunch of prodding and tweaking.

And yet I’m not surprised.

I could treat the output as a first draft and edit it, but that would be almost as much work as writing it myself. Perhaps most importantly, it would be a different type of work: more drudgery (editing vs writing, anyone?) and without the major payoff. 

I would not be smarter for having created the content.

There are two reasons to write. AI is great for one of them. For the other, it’s a trap. 

You Write to Communicate

Communication is the obvious reason you write. And when writing to communicate, AI is your savior.

You can have it draft the near-perfect rejection email in seconds, getting the direct but empathic tone you couldn’t nail in 10 minutes.

You can have it write all your marketing language. It will succinctly get to the point where you could not. It will match features to problems expressed in your clients’ own words — and it will do the research to find those words.

You can have it write that letter to your doctor, that email to your team member, that thank you note to your spouse. It will do all this well and you will be grateful.

But You Also Write to Think

But you also write to think. And to outsource this writing is to outsource your thinking. 

There is no world in which I get to where I am today without writing. 

There is no world in which I get to where I want to go tomorrow without writing. 

(There is a world — this one, right here and now — where every day someone promises me the benefits of writing without writing.)

I know of no other path to building, refining, coalescing or codifying expertise than writing. It might just be me (it’s not, I’m feigning politeness, like all Canadians do) but I see no substitute, no shortcut, no saving me or anyone else from the pain and euphoria of writing. 

Write to Think, Publish to Communicate

Now, once you’ve written to learn, you then publish to communicate. That is how you prove your expertise in the public domain and move from “just another vendor” to “the expert” in the mind of your future client. You drive inbound leads with you already in the power position in the relationship.

AI can help with the editing to more clearly communicate your thinking, but you have to do the thinking first.

But that’s not how AI is being used in content marketing right now, is it?

No, it’s all shortcuts. It’s writing for the sake of communicating, barely disguised as thinking. It’s a morass of pithy, well-written, not-incorrect, vaguely useful communication where the poster learned nothing in the writing, where they did not become one bit smarter in the process but were able to check the box that says, “I did some marketing today.”

And some of it still works to generate leads. But for how long? And who are the laggard clients who still think they’re hiring the human who penned the post?

This isn’t about how good or bad the AI tools are. They’re great and getting better faster.

And it’s not about whether or not you can still generate leads this way. You can. For now. 

It’s about misunderstanding the role of writing in an expert business. You write to think, to learn and to grow. When you publish that writing, you build reputation and drive leads that build your business. If that post you post doesn’t make you smarter at the end of writing it (or make people smile or laugh), why post it?

This whole ponzi scheme is coming down. I don’t know when, but when it does there will be a lot of atrophied “expert” brains. I hope yours isn’t one of them.

I want to thank ChatGPT for the inspiration for this week’s post. You were very helpful. Any errors are mine. All. Mine.

-Blair