The regurgitation trap
To understand why AI copy still needs work before it represents your brand, it helps to understand what AI copy is. All GenAI makes its goods from our old friend, data. For content writing, that data is typically publicly accessible, un-gated words on the internet. Every day, creators everywhere enrich that source by publishing more original content — research papers, breaking news, podcasts, fact sheets, product information. The list is endless. Once published, it’s up for grabs as raw material for an AI tool’s next spiel.
Or, as Bernard Brode — a techier writer than this one, for Inside AI News — puts it (and almost certainly more accurately):
“These apps are powered by Big Data analysis to uncover patterns, trends and correlations found within the 74 zettabytes of data already in existence. They then use this exposure to learn about human speech patterns and create new copy based on the examples they study.”1
Now imagine that every writer in the world got so fixated on AI tools that they decided never to write anything new again. This is the ‘regurgitation trap’: a never-ending circular conversation about nothing particularly new, going round the houses in articles and posts, standardising information, never bespoke or topical, simply clogging the datasphere with more of the same.
That’s why, when generating copy with AI, we should treat what the tool produces not as the finished product, but as raw material. It’s already saved us a truckload of time pulling things together — our opportunity is to use that time to enrich it: make it interesting, current, useful, and make it represent the views and strategy of the brand. Here are five steps, as a minimum, to get it there.
Read it
Obvious, perhaps — but once your copy-bot of choice finishes its word dump and you’re done marvelling, read it. Thoroughly and critically, like an intern brought you their first draft for review. You’re looking for structure and flow; whether the content is strategically right for your audience; areas more research could flesh out; and unnecessary repetition. Step one sets up the next four.
Remove AI-isms
As smart as AI is, its produce exhibits hallmarks that will have your content screaming “AI wrote me” if you ship it as-is. The best spotter of an AI-ism is still a real, live human. They’re many and varied: keyword placement dropped awkwardly into a sentence; a “Conclusion” at the end of every post (we’re not writing uni essays); leading off with “in this blog post”. Removing them is a simple step to making the content less conspicuous and more your own.
Restructure
Outlines let you brief structure up front, but you can’t be sure the story holds until you see what lands under each heading. AI is good at lists but can be average at story cadence — making the same point in consecutive paragraphs, different ways — and prone to structures that read more like laundry lists than stories. A structural appraisal and re-shuffle makes copy flow, resonate and persuade like it was written by Peggy from Mad Men, and moves the piece a giant leap from its AI-generated birth state.
Customise
This is where you transform copy from the regurgitated to the scoop — the inside story, imbued with your brand’s special sauce. Queried about something current in late 2024, ChatGPT admitted: “I currently don’t have access to real-time news or databases.” Unlike AI, you do have access to what’s happening right now, and that’s what your customers find compelling. Customise with a quote from a team member, facts from internal studies, or current case studies and milestones.
Validate
Isn’t it ironic that copy drawn from 74 zettabytes of data arrives with almost no bibliography? The best papers reference their sources; Jasper hands you plenty of facts and not a footnote in sight. So fact-check and validate. Scan for percentage claims and find sources to back them. Then find the paragraphs covered conspicuously lightly and use them as a cue to hunt for expert opinion you can quote and reference. Before you know it you’ll be an expert — and your reader will start to believe you.
So, resisting the urge (unlike Jasper) to add a “Conclusion” — a nod to our AI friends and all they’re brilliant at. This article is 100% human-written, but its writer is a fan and would never disrespect the tools or their makers. If there’s an opinion in these lines, it’s simply that AI, in all its forms, is servant — not master. It’s the rookie; we’re the seasoned pros, from whom it still has plenty to learn.
Notes
- Bernard Brode, Inside AI News.