How to Use AI and Still Write Like a Human

AI is now part of the writing process for a lot of people, and rightfully so. It helps with research, ideation, editing, and getting unstuck––all the tiresome tasks that slow us down. But it presents a tricky tension: AI can be an incredible writing partner and an incredibly bad writer. As AI-generated content becomes more common, readers are growing familiar with its patterns. That leaves writers to figure out how to use the tool without letting it take over their voice.

Doing that requires attention to two things. The first is the writing itself: the words, sentence structures, and stylistic habits that can make a draft immediately feel AI-generated. The second lies underneath the language, in the logic of the piece and the way its ideas are organized, developed, and connected. If you want to write like a human while using AI, you have to pay attention to both.

Start with the Obvious

Some patterns have become easy to spot. AI leans on certain rhetorical constructions, particularly the familiar it’s not X, it’s Y format. It also likes qualifiers that soften a point without adding meaning. A straightforward sentence can quickly turn into something like “AI can often help marketers improve efficiency in many cases.”

Then there are the sentence patterns themselves. AI frequently produces strings of short, similarly structured statements that give the writing a stilted, mechanical rhythm, even when each sentence works on its own. Transitions can feel just as formulaic, with words like additionally, moreover, and furthermore moving the reader from one point to the next without explaining why those points belong together.

And, of course, there’s the ever-controversial em dash––the grammatical equivalent of spray-painting a wall with the words “AI wuz here.” We’ve defended it in our blog post “The Em Dash Hill We’re Willing to Die On,” and we’re still standing on that hill.

None of these habits is inherently evil, but great writing requires balance, and AI tends to use them in excessive, predictable patterns. Fortunately, what AI lacks in moderation, humans can make up for with close attention. In practice, that means varying sentence length and structure, replacing empty transitions with ones that show how ideas relate, and cutting meaningless qualifiers. Instead of presenting your reader with a string of short, declarative sentences, try swapping a period for a comma and letting your thoughts meander a little. The writing will feel more natural, and readers will notice.

Then, Look at the Thinking Underneath

A draft can avoid every obvious AI writing tell and still feel artificial if the ideas themselves are underdeveloped. Good writing has to make sense before it can sound good, and that takes more than generating a series of relevant points. Think back to writing an essay in school: before you could draft it, you had to decide what you thought, determine which ideas supported your point, and arrange them in an order the reader could follow.

AI works differently. It excels at recognizing patterns in language and predicting what is likely to come next, which means it can generate several accurate points about a topic without building strong connections between them. Ask ChatGPT or Claude to write a blog post, and you may notice this pattern: one idea appears, then another, then another––and although each belongs to the same general subject, the reader is left to do too much of the work connecting the dots.

Writers know that a strong paragraph does more than collect information. Its opening sentence establishes a point, the sentences that follow develop or support it, and the transition to the next paragraph helps explain how the argument is progressing. The reader should understand not only what you are saying but also why this point comes now and what it adds to the larger idea.

The same principle applies to short-form content. A landing page does not need traditional essay paragraphs, but it still needs a clear progression: the hero introduces the central idea, supporting sections expand on it, proof points add credibility, and the CTA arrives once the reader has enough context to act. When each section is generated in isolation, the pieces may sound polished while the full page feels disconnected. That underlying structure is harder to fix after the fact, which is why writers should stay involved in the thinking before AI gets involved in the writing.

Use AI Without Handing Over the Process

The easiest way to preserve both voice and logic is to be deliberate about when and how AI enters the room. Start with the prewriting yourself. Research the subject, work through the more complicated ideas, and decide what the audience needs from the piece. Brainstorm before asking AI to brainstorm for you, and build the outline yourself so you understand what belongs together, what should come first, and how each section contributes to the larger point.

It is also worth writing the first draft yourself, even if it is rough, because drafting forces you to test your ideas. Two ideas that seemed connected in an outline may need more explanation between them, a section may turn out to repeat something you already said, or you may realize the most interesting point is not the one you originally planned to emphasize.

Once that foundation is there, AI becomes much more useful. Ask it to tighten a paragraph, improve a transition, identify repetition, or help vary the rhythm of a section that feels too uniform. You can also use it to pressure-test the logic by asking where the argument moves too quickly or where a reader may need more context. Whether or not AI was involved along the way, what matters is that the final piece reflects human decisions at both levels: how the writing sounds and what it’s trying to say.

Readers may not be able to name every pattern that makes a draft feel machine-made, but they can sense when something is off. Catching the surface-level tells makes a piece easier to read, and keeping the logic in human hands gives it something worth reading in the first place. Both take attention, and neither can be fully handed off. AI can be great for refining your words, as long as there’s a human shaping the thought behind them.