Three Editing Assignments From Hell

18th July 2021

Last week a new client gave me what she generously called a 'proofing' job for an article that had clearly been written by an AI content writing program. I'm fairly sure I know which one it was, since I reviewed several earlier this year. However, I can't prove it, and therefore won't name the platform.

The mistakes that a content-generating AI makes are not the same as those that humans typically make. Informed by SEO principles, the piece opened well and ended well, with a definitive summary and a call to action; it broke down nearly every argument into easy-to-digest bullet points; and it wrote to an excellent standard of English.

If you were to speed-read the AI's output, you'd believe that the article was sound. But if you actually really read it, you begin to notice that one of the two sources cited are non-existent; that the lack of supporting hyperlinks is atypical for such an essay; that the middle section of the text will follow a salient point with an utter non-sequitur; and that strange, emotive language will pop up and as quickly disappear in the context of an article with a neutral tone.

If only to save time, I would rather have 'fixed' the piece; but the more I re-read it, the more I became convinced that it was beyond repair. So I rewrote it, and the client was happy. I'm not sure if this qualifies as 'editing', or 'proof-reading', though; If you can give meatware like myself a GPT-3-style 'prompt', that rough first draft kind of fell into that category.

The experience reminded me of several other assignments I've had over the last 15 years as an editor, where it was politically necessary to throw the original submitted material in the trash, and take one for the team. For your amusement, I recall some of them here.

1: The Podcasting Superstar

Word came down from the boardroom that I was to work with him, and generate content around his specialist subject in tech. Mid-forties, dour and pragmatic, he was a prolific broadcaster. He had no other discernible occupation, though some believed he had achieved financial independence earlier in his career.

He'd arrive at the major tech events and blast through the prominent attendees like flu, plotting up in his hotel room for days, editing hours of new interviews into a stream of podcasts that were followed by many thousands of the most influential and desirable followers possible. Of course, the board wanted some of those followers, and to hitch a ride on his audience.

He was volatile and eccentric. In the couple of years I knew him, he would frequently phone me up with non-actionable, shocking intelligence related to the British state, and to clandestine discussions in the halls of power. "You can't tell anyone!". "Don't use this!". I don't know why he told me those things (or even if they were true), except to relieve the general burden of secrets, or out of some bizarre compulsion to impress me. I wished, after a while, that he would stop doing it, not least because he always phoned instead of writing a text.

Now it was time to get the first feature article out of him. He had no written material published anywhere, except for transcripts of some of the hundreds of hours of his podcast interviews. Not one article that I could find.

In due course, the first submission arrived. Some part of me thought that it must be a mistake, though I couldn't contrive the possible circumstances of such a mistake.

It was a tortured document; 800 or so words that hobbled randomly around an uncertain topic like Jack Nicholson in the hedge maze at the end of The Shining. Likewise, it simply died at the end. It made no sense.

I went back to his podcasts and scrubbed through some hours of various episodes. There he was, erudite and informed, marshalling a powerful capacity to reason, asking tough questions, systematically exploring the subject, and bringing the work to an insightful conclusion.

Then I returned to his written submission, and read it again. Finally, I realized the truth: he had adopted the spoken word as his medium of choice because the written word had been his implacable, lifelong enemy. He was functionally illiterate and fatally dyslexic.

Did he even know? I only knew that I didn't want to be the one to tell him. However, there was no way out; the board wanted this guy on our platform, writing, not speaking.

I set aside all other business and began composing an essay that seemed to be set somewhere in the approximate zip code of his subject matter, as far as I could discern it. Under the guise of merely editing his submission, I had to phone him several times in order to extract the meaning of certain of his sentences, such as they were, so that the new piece had some rough relationship to his unintelligible submission.

I wasn't looking forward to sending him the 'revised' piece, because it tacitly said 'I know. I know, and I've got your back.' I wasn't at all sure about the extent of his self-delusion.

About an hour after I sent it, I received an excoriating email from him, and my heart sank. Predictably, it was short, and suggested that my actions had brought this new relationship to an abrupt end.

By now, I knew to stick to audio; I phoned him and smoothed the breach over as well as I could, reminding him that his sole byline was secure, and deferring to his eminence in his own domain. He was mollified enough to let me put the article up under his name, despite his disillusionment about this new relationship.

I posted it and promoted it; it took off at a referring site, and became a hit that day.

After this, he decided that he had been perhaps a little over-critical of the extent to which I had edited his original work. The following 18 months were punctuated by the recurring event of his latest essay, which would prompt me to clear the day again. After a while he began to adopt my style (of his voice) a little, and the copy became slightly easier to read.

When I left the company, he cut me off; no more annoying, unprintable secrets from Whitehall, true or otherwise. I wondered for some time if the success of his written work would prompt other editors to solicit articles from him, and how those first email exchanges would play out.

2: The Editor's Friend

Around the editorial table at the publishing company, we had rota system in place for his latest magazine article or website feature. When it was your turn, your day was over; whatever work you had been doing was either deferred, if deadlines allowed, or shunted to your colleagues, who would far rather pick it up than be 'it' that day.

He was the chief editor's friend, and as far as we could tell, the editor had been writing this guy's material for years. They went back a long way. Considering the effort needed to transform the submitted copy into something legible, we surmised that this guy must have saved the editor's life at some point, or perhaps had helped to bury a body.

Unlike the podcaster, the reasoning and facts of his work were about 80% present in the submitted copy; but it seemed that he had written a coherent piece, only to cut it up and rearrange the pieces into random Haiku slices, the way David Bowie used to write lyrics in the 70s. Some of those slices appeared to be from entirely different articles.

For him, as for the podcaster, ideas came in a blizzard, and all he could do was hold the paper up to the blizzard and let them blot it like Rorschach stains that we would later have to interpret.

We all wrote hit pieces under his name, and we all thought that it would have been far easier if the relationship had been more explicit: he could just as easily have said 'Write me a piece about XYZ, and put it up in my name'. We would have saved a lot of time, and arrived at the same result.

3: Forensic Editing

A woman of business and standing in the tech world, she was a native of a Romance language, but communicated brilliantly in spoken and written English. Without any doubt, her average standard of English was above the mean standard of an English-speaking native.

Therefore I had no hesitation in accepting her offer to give me occasional proof-reading work for her influencer articles in trade magazines. She paid well, and quickly.

When she sent the first piece, she asked me how long I thought it would take me to proof it. I opened the piece and scanned down the paragraphs. Uh-huh. Uh-huh. Nice.

I told her I could get the piece back to her later that afternoon, and secretly felt I would be overcharging for a perfunctory and effectively unnecessary read-through of a solidly composed feature.

I read it through, about 1,200 words. It was in impeccable English, and laid out a series of well-ordered ideas in a systematic fashion, leading to a satisfying conclusion and a well-earned call-to-action.

However, something was wrong; in spite of these things, the piece did not give me any clear picture of what she was trying to convey.

I re-read the piece, much slower this time.

At the end, I still did not know what it was I had just been reading.

I tried a third time, now moving at a per-word and per-sentence snail's pace. With a growing sense of deflation, I realized that there was a fatal disconnect between her first-class use of language and her formidable ability to reason. There were ideas, and there were words – but they were unrelated.

Before the desktop publishing revolution of the late 1980s, graphic designers charged with mocking up book covers would use 'knitting' to represent text – random up, down and diagonal strokes at word-sized intervals, rendered with Rotring ink pens and fitted precisely between a grid of descenders and x-heights. Superficially, it looked like language, but did not bear inspection.

This woman was knitting – brilliantly. You had to dig deep, deep into her style to understand that she was saying nothing coherent. Yet it was not a fault of reasoning, but, as far as I could tell, a mental paralysis fueled by her lack of confidence in writing (rather than speaking) English. Her cognitive powers were so engaged in the production of readable and grammatically correct English that her ideas were only faint signals in the prose, rather than powering it.

This was something I had rarely seen in my 15 years editing other people's copy: a case for forensic editing.

I rewrote the piece radically and sent it back. She was delighted, and acknowledged in an email that this was effectively a new work. It confirmed my general belief that no-one will ever curse you for making them look good (except perhaps initially, as with the Podcaster).

Therafter her assignments were easy, since any journey is shorter when the road is known.

I think I have more of this to come; editing of AI output. At least for a while, deep edits will be the norm again, which is perhaps worth a ponderous sigh.

Then, as GPT-4, GPT-5 come online, I guess I'll fall silent (even more silent than the fate of the ghost-writer, who is silent, but ultimately paid for his or her silence). After that, I'll live again, I suppose:

Write me an article about difficult edits. [ENTER]

She was a woman of commerce and esteem in her field, a French native with a mastery of spoken and written English...