The Paradox of AI Writing

Before you can get a large language model (LLM) to write for you, you must first write for it.

Initially, this may seem like a small distinction, an insignificant variance in the order of operations.

But it’s ultimately the factor that decides what sort of experience you’ll have with an LLM.

It’s almost always the difference between average and excellent, or generic and specific.

As these models get bigger, faster, and more capable, it’s easy to think of them as the solution to everything. After all, there’s not much they can’t do at this point.

You can shape them to be strategists or data analysts or venture capitalists, and they will respond as such, in ways that are truly incredible.

But if you can’t write clearly and capably, or if you’re unaware that there’s some skill and tact required to prompt, then you’re at a real disadvantage.

For most people, having an LLM write content is the easiest use case to deploy—and the fastest.

Perhaps your organization has always wanted to have a robust content strategy, and you see frontier models as the realization of your goal.

You can ask an LLM to write a 600-word blog post on nearly anything you can think of, and it’ll do a pretty good job without a whole lot of direction.

What you’ll realize, though, at some point, is that the content is missing one vital component: your perspective.

We’ve reached a point where it’s no longer useful (was it ever?) to publish information on the internet simply for the sake of publishing it.

What your customers are looking for is not one more forgettable description of what something is, or some generalized advice on how to perform a task.

Instead, they’re seeking your unique perspective, the lessons and insights you’ve learned and applied in all your years of business.

People are increasingly selective in what they read, and if you don’t give people a good reason to read what you’re putting out into the world, then you’re just contributing to the problem.

Yes, you can ask an LLM to write for you, but you first need to write for it, even if all you’re writing is your top three thoughts on a subject.

If you’ve got an idea for an article, you should write it, and then ask the LLM to act as a developmental editor and give you critical feedback on what’s working and what’s not, and how you can shape it into something better and more useful to your reader.

Instead of asking it to write for you, write for it, and then ask it what it thinks.

This is far more useful in the long run for both you and your readers.

You do the thinking and the machine does the refining.

We often describe AI as “human + machine.” not “machine + human.”

That’s important because “human + machine” isn’t just a description, it’s a directive.

It’s the order of operations for a reason.

Resist the temptation to change it.