Human editor vs AI

I went head-to-head with AI to test my editing skills.

I asked ChatGPT to create a short introduction to a fictional accounting firm. To mimic a typical client experience, I briefed ChatGPT to be the owner of this accounting firm and not a trained marketer or writer. I loaded the text with technical jargon and corporate fillers.

Then I tried to edit it: for length, structure, and ease of understanding.

And finally, I asked both ChatGPT and Gemini to have a go: with a clear brief to make the text approachable, useful for people without an accounting background and to restructure it if needed.

I also asked them to explain their working.

The verdict?

I was pretty happy with my own first efforts and on reviewing again, there are some further changes I could make. But overall, it achieved the brief and injected a bit more personality to engage readers.

ChatGPT took my instructions quite literally. It shortened sentences, took out some technical jargon and by its own explanation, “removed repetitive phrases and unnecessary details.” But it lacked any personality and didn’t do any restructuring; keeping sentences in place and just editing them based on their literal meaning.

Gemini did a surprisingly good job. It took the intent of the text, restructured it to be more web-friendly and stripped out almost all of the jargon. It also felt friendly and had more personality than the ChatGPT attempt. Unfortunately, in doing so it almost over-simplified the text and lost a few key details that I know the client would want to keep.

(I also asked my Microsoft Copilot preview and it...reduced it to 100 very basic words that missed most of the point, so I eliminated it.)

So what’s to learn here?

AI can do a decent job writing and editing. We know this. With more prompting, I’m sure either tool could have refined their efforts and churned out something acceptable.

But there’s the key – “with more prompting.” 

Optimistically, this reassures me there’s still a lot to be said for the nuance and understanding humans bring to the writing process. 

These tools also needed me to brief them: to explain the role they'd play and request an output in very careful detail. I also had to outline the client's needs and their audience's expectations.

The other key difference is that I had a natural sense of when to stop. The AI tools will work for a second and present one option, waiting for any feedback before trying again. On the other hand, I was immediately ready to iterate and make improvements to my effort based on my own re-reading and assessment of the copy.

In short: the AI tools required me to talk to people, understand their needs and translate this into a request. Then they needed me to decide if the final product was up to scratch. In my opinion, these steps are still uniquely human.

So for now at least, AI is a part of the team – not a replacement.