How I use artificial intelligence to write for clients

AI is a writing assistant. It’s not a writer.

I began experimenting with AI as a writing assistant when ChatGPT was released in November 2022. Since then, I’ve invested in the paid version of Perplexity.ai so I can compare GPT-4 to other LLMs such as Claude 3 and Mistral Large (Gemini by Google was pulled from the platform). I also use Otter.ai to conduct interviews and generate transcripts for them.

Despite ongoing experimentation with AI, I’ve never been able to yield more than two or three consecutive sentences that meet my standards for my clients—but I still use AI as an assistant. Here’s what you can expect behind the scenes when you hire me:

I start research and outlines with AI.

AI is good at sparking ideas and filling in logical gaps (within limits). If I don’t have an outline from a client, I’ll use AI to start one then fill in its many gaps myself. AI is still too broad with its suggestions—and frankly wrong about certain facts—and I’ll always perform in-depth research via other means so my pieces are original and verified.

I use AI for basic definitions.

As my clients shift away from old-school SEO practices, I’m going after Google snippets less these days. That being said, I’ll use AI to generate the basis for a definition that would rank for one—but I always need to edit to get it exactly right.

I use AI for FAQs.

If your piece requires an FAQ, I’ll write the piece in my own words then ask AI to summarize those words for concise answers. But I often need to edit these heavily, too.

I generate headlines with AI.

When I’m finished writing a piece in my own words, I’ll populate AI with samples from that piece and ask it to generate several headlines with a specific goal in mind. I need to remind the AI that I prefer editorial headlines over SEO ones, but alas, it often can’t comply to my standards.