Do we need human editors when AI can write and edit for us?
- Amanda George
- 5 days ago
- 5 min read
Updated: 3 days ago
I should start by explaining my position, that I don't knowingly or willingly use artificial intelligence (AI), for the reasons we are all aware of: environmental harm; the displacement of humans (from land and jobs); intellectual property theft and data protection issues; potential bias and prejudice; hallucinations and inaccuracies; and a (documented) loss of critical thinking skills and creativity.
That said, I understand that many of my clients use AI to research, write, and edit, for reasons of speed, ease, not getting left behind, and an expectation that they will do so by their employer or institution. I edit a spectrum of writing, from beautifully messy human prose to perfectly automated output, and everything in-between.
Since 2023, like so many other editors and proofreaders, I have had to teach myself how to edit AI-assisted writing. And after nearly three years of this new age of editorial intervention, and much existential angst, I feel better positioned to talk about the patterns I have observed, and how best we can serve our clients – while safeguarding the craft of editing and writing – in a fast-changing world.
First, what are the issues with AI-assisted writing?
Ten common issues human editors should be aware of:
Lots of short sentences, often with no linking words or signposting, creating disjointed writing that can become monotonous and even confusing for the reader.
Example: The cat needed a good home. The home should not be close to a busy road. This would be dangerous for the cat.
Solution: The cat needed a good home, ideally not situated next to a busy road, as this could prove dangerous for the cat.
Single-sentence paragraphs: while never-ending paragraphs are problematic, and shorter ones help to keep the reader engaged, in more formal writing paragraphs should ideally have more than one sentence.
Important words missing, for example relative pronouns, that the reader needs to make sense of a complex sentence.
Example: In more complex sentences, omitting relative pronouns the reader needs to parse a sentence can slow them down.
Solution: In more complex sentences, omitting relative pronouns that the reader needs to parse a sentence can slow them down.
Overuse of formulas such as 'not only, but also' and 'not this, but that': these formulas can work well – when used in moderation.
Example: She not only finished her to-do list but she also managed to run around the block before dinner.
Solution: She finished her to-do list, and even had time to run around the block, before dinner.
AI loves adding unnecessary words that can disrupt the flow of the sentence.
Example: How can people actually tell if something has been written with AI?
Solution: How can people tell if something has been written with AI?
AI struggles to ensure consistency and apply style guides across larger documents.
Example: She felt better––she had spent days resting. But she was worried about overdoing it – doing too much exercise – in case she relapsed.
Solution 1: She felt better – she had spent days resting. But she was worried about overdoing it – doing too much exercise – in case she relapsed. [UK English]
Solution 2: She felt better––she had spent days resting. But she was worried about overdoing it––doing too much exercise––in case she relapsed. [US English]
Because LLMs are trained to please the prompter and always provide answers, they sometimes (very confidently) make assertions that are false, fabricated, or misleading. Editors should check anything that doesn't feel right lurking beneath the polished veneer of perfection.
AI doesn't inherently understand grammar and usage; rather, it predicts what words are likely to come next based on patterns in its training. This means editors should be vigilant about, for example, words being hyphenated (in)correctly depending on their position in the sentence (compound or predicate).
Example: She was a well-known actress. Her films were all well-known.
Solution: She was a well-known actress. Her films were all well known.
AI-assisted writing can feel very generic, because LLMs are trained on vast amounts of data, and their default is to reflect the average phrasing from the internet.
AI output often looks flawless, but readers can struggle to engage with it, because it is based on mathematical averages and predictions rather than lived experience, nuance, and human emotion. Even in business and academic writing, readers want to connect with the writer and be taken on a journey. As editors, we can try to restore some of the creativity and human experience that readers crave, for example by showing them what something feels like rather than telling them, by varying vocabulary, and by making the writing feel less predictable and more relatable.
And yes, I do use en and em dashes, because they're great pieces of punctuation, and writers and editors were using them to great effect long before AI appropriated them.
And second, what is our role as human editors in the age of automation?
Five ways we can support our clients, and our roles as guardians of the written word:
We can be vigilant about all of the above, and more, when editing AI-assisted output. Of course, human writing entails different challenges, but those are not the focus of this blog, because we were all trained on human writing.
Whether our clients enjoy the process of writing or prefer to use AI, our job is to offer them peace of mind. This means they can spend less time writing, checking, and self-editing, knowing we will smooth out any issues, and any AI or human blips, so they can publish with confidence.
We can be supportive, providing that vital human contact and collaboration we are all craving in an ever-more impersonal world.
Since writing is intended for human readers, it seems logical that a human editor should ensure it is fit for human consumption.
And finally, we can support and encourage our clients to trust their own writing, however raw it might be. When I read or edit human writing, I feel a spark of electricity, which is exactly what our clients' readers should feel. In an increasingly automated and fragmented world, encouraging our clients to trust their own writing, experiment with it, and enjoy the process of research–write–refine is an act of radical care.
Did you find this post helpful?
Whether you’re working on a case study, a business report, or an academic paper, small details like punctuation, sentence length, and word choice can make a big difference to your writing, ensuring that your reader keeps coming back for more.
If you found this post helpful, you’ll find more practical tips on grammar, punctuation, and writing in my other blogs. Or get in touch if you would like me to worry about this – and other things – for you.