In 2020, essayist Vauhini Vara started playing with GPT-3, an artificial intelligence bot designed to predict text and finish pieces of writing it was fed by human users. She was astonished by GPT-3’s output, how it mimicked human language and sometimes produced sentences that were stunningly profound.
In an essay for The Believer, she shared an experiment in which she used GPT-3 to write — for the first time — about the difficult loss of her sister. She started by feeding the tool one sentence she wrote herself, and let the virtual robot come up with what came next. The more of her own writing she gave to the tool, the better its output was.
Along with sounding more like something she would actually write, GPT-3’s contribution seemed to pick up on and echo her own emotional experience. What it wrote was not what she would have written, but a reader wouldn’t necessarily know that. And seeing how the AI did or did not write the truth of her experience helped Vara choose her own words.
ChatGPT and other generative AI products have dominated conversations this year among writers, content managers, and seemingly anyone who touches a keyboard. The hype around ChatGPT has been deafening. What can’t technology do! was the implied exclamation from some circles of the internet. But for all the hype, there were plenty of instances that knocked AI down a peg:
World of Warcraft Redditors enacted virtual sabotage to show the pitfalls of gaming news sites relying on generative AI for accurate content. According to The Mary Sue, “Black Mirror writer Charlie Booker … shared the results of an experiment in which he had AI write an episode of the show, and it turned out to simply be a messy mashup of synopses of other episodes the AI found on the internet.”
In the meantime, Google started testing an AI product designed to produce news stories — the idea being that this tool will account for the shortcomings of other generative AI tools and avoid inaccuracies or ethical snafus.
At this point, the jury pretty much agrees: generative AI is not a replacement for writers, at least not good ones. It could, however, help writers save time at different points in the writing process. Whether it does or not depends on how writers use the particular tool.
To get a sense for how AI tools save writers time (and how they don’t), I spoke to a handful of writers who’ve embraced AI and are using it to enhance their work in a non-creepy, non-replacing manner. For these writers, Wordtune, ChatGPT, and other AI gadgets are simply tools for their workshops, like Microsoft Word, spellcheck, and Clippy (RIP) became decades ago.
Each of them offered insights on what isn’t helpful about AI, what writers (and companies) should definitely not do, and what uses they’ve found to be beneficial and improve their productivity.
We’ll start with what not to do.
No, AI is not the writer you need for longform content.
Ry Tidwell serves as the one-man marketing team for Saltmine, a platform that helps businesses design their workplaces. Tidwell often finds himself in need of extra support for content marketing projects and Saltmine’s blog. But he doesn’t have a huge budget to pay freelancers, so generative AI looks like a possible solution.
At the height of ChatGPT hype, Tidwell played around with the tool and had it write a blog post.
“It produced exactly what I thought it would: a very vanilla, not anything special, blog post,” he says. He also tried ContentBot.ai, but was frustrated by how the tool required him to write essentially an entire project brief to get a few generic outputs.
Ultimately, relying on either of these tools for longform content would give Tidwell more work, not less — because in addition to editing and revising the content to make it more coherent and interesting, he’d also have to factcheck it for accuracy. As a result, he’s given up on using AI for longform content.
If you want a blog that regurgitates every other blog on the topic that’s already on the internet, use an AI tool to generate one. Just don’t expect it to be original.
No, AI is not a stand-in for time-consuming research.
Adina Levin, a Barcelona-based freelance copywriter, does not use AI for information-gathering or anything related to the truth. Why?
“Because it hallucinates,” she says. “It’s not looking up … and giving you a copy-pasted answer of what’s in Encyclopedia Britannica, which went through a zillion fact checkers. Or even an article from the New York Times. … It’s just creating probabilities and figuring out what word is most likely to come next — but just because it’s most likely doesn’t mean it’s true.”
Cassidy Ritter, a business journalist, ran into this issue when she was working at Built In, which serves the startup community with business news and networking. The site uses an AI assistant called Writer to write some basic news articles, but every AI-generated article needs to be closely factchecked, because sometimes Writer will just make things up. An article that may have taken a staff writer 45 minutes to compose could sometimes take even longer to factcheck and correct.
As Levin explained, it’s crucial to remember that generative AI tools are focused on words, not meaning. They’re basically super-powered predictive texting tools. Considering how often your phone’s predictive text is wrong about what you want to say, it should be no surprise that generative AI also gets things wrong.
As Calvin Wankhede wrote in Android Authority: “ChatGPT isn’t ready to become your number-one source of information anytime soon.” And that goes for every other generative AI product.
No, AI is not the way to express yourself creatively.
This one should be pretty obvious, but if AI wrote the thing, you’re not the author. And if you identify yourself as a writer, I’m not sure why you would offload that precious work to a machine anyway. The amount of effort you’ll put into prompting ContentBot.ai, Writer, ChatGPT, etc., to get the output you want could easily be spent stringing your own words together.
“For creative, expressive, or exploratory writing tasks, using ChatGPT is like supervising a bumbling assistant who needs painfully detailed, step-by-step instructions that take more effort to explain than to simply do the work yourself,” Laura Hartenberger, a writing professor at UCLA, writes in Noema.
Vauhini Vara’s essay, “Ghosts,” is a case in point. While GPT-3 could help Vara along by showing her what she didn’t want to say, it couldn’t actually write the author’s inner experience of losing her sister. It got things wrong. And when Vara removed the AI-generated text and put in what was actually true, the words were not the same.
While, as Hartenberger writes, “generative AI tools like ChatGPT offer the seductive possibility that we can optimize this laborious process,” there’s simply no getting around the difficult work of revealing your inner self on the page. Only you, the writer, can do that work.
Q: If AI adds time for writers when used to generate longform content, isn’t a reliable research tool, and isn’t the way writers should express themselves creatively, how exactly does it save writers time?
A: Through unconventional uses and for microdoses of content generation.
Each of the writers mentioned above has come up with ways of using AI tools that help them at a point of the writing process where they tend to get stuck. These uses fall into a few categories:
- Copywriting
- Editing
- Unconventional Uses (yes, this is a catch-all!)
Here’s what they are:
Real-World Use Cases for AI Writing Tools
1. Copywriting
“Our world runs on filler text: avalanches of words and phrases written to optimize Web sites for search engines, to use as tags on social-media posts, and to employ in marketing newsletters that spam in-boxes,” Kyle Chayka writes in The New Yorker.
The filler text he describes — aka copy, often written more for other machines than for the reading pleasure of humans — is where AI generation is perhaps the most useful. (The irony of a machine writing words for other machines doesn’t escape me.)
Here are a few ways Tidwell uses AI for copywriting:
a. Short-form copy for blog promotion
Tidwell writes long, so he often turns to AI tools to help him write short. This especially applies to copy that he can use to promote Saltmine blog content. For each blog he writes, he’ll have an AI tool generate a two-sentence summary — which he edits as needed and then uses on the website, across social, email, and sometimes even as the blog’s meta data.
b. Title and subheading options for an article
Tidwell also uses AI to come up with title and subheading options for the pieces he writes. He might take them as is or combine bits and pieces of this and that for a better option.
c. Email copy
As an over-writer, Tidwell struggles to write concise, to-the-point emails. With AI, he’ll write the initial draft and have his choice of AI tool cut it down to the needed length. Or he’ll take the best parts from a few AI-generated options and cobble them together for a better result.
“Whether it’s emails, headers, or social posts, I’m still going to tweak them,” Tidwell says. “There are definitely opportunities to use your human brain to take something that [AI has] written and make it that much better.”
2. Editing
Editing is the longest part of the writing process for Tidwell — and as a one-man marketing team, he doesn’t have time to spare. AI tools have come in handy for helping him speed up this part of the process.
“After you’ve read copy and edited copy for three hours in a row, your brain starts to give out a little bit and you’re like, ‘I’ll save this for tomorrow,’” Tidwell says. But since starting to use AI tools, “I save a lot less for tomorrow and just do it today.”
AI helps him:
a. Improve clarity
If he’s at a loss for how to improve a sentence or paragraph that he knows isn’t quite clear, he’ll put it into the AI machine and ask for a revision. What the machine spits back out might not be exactly right, but it can give Tidwell direction in terms of what needs to be changed or adjusted.
b. Tighten long copy
Maybe Tidwell’s written 100 words, but only has space for 75 and doesn’t know what to cut anymore. AI can play the pinch hitter and come up with a new sentence construction to cut it down to size.
c. Adjust tone to be more/less formal
Sometimes, the first draft is too casual — or too formal — but you’re at a loss of how to change it while maintaining the right focus. AI can help with that too.
3. Unconventional Uses
Adina Levin should win an award for most creative uses of AI. At least, her go-tos blew my mind. As a copywriter, she’s always on the hunt for a new way to say things that resonate with audiences, catch their attention, and leave an impression. When she’s at a loss for words, she turns to AI.
a. Interactive thesaurus
Can’t quite place your finger on that one particular word that you know would be perfect for what you’re writing? Forget dictionary.com (or thesaurus.com) and ask ChatGPT instead. Type in “a word that refers to” and then a description of what you’re looking for. It might take a few attempts, but you should be able to find what’s on the tip of your tongue.
b. Idioms and plays on words
Idioms and puns are a copywriter’s bread and butter. The best copywriting uses plays on words to create surprise double meanings that stick with their audience. But coming up with original wordplay is tough. Levin turns to ChatGPT for lists of idioms and puns related to the topics she’s working with — and then mines those lists to come up with new ideas and combinations.
c. Generating emojis
Emojis come in handy for light-hearted marketing copy — especially in email subjects and social posts. But it’s easy to use the same ones over and over again. To find emojis she doesn’t even know exist that are specific to the copy she’s working on, Levin will ask AI for emoji suggestions and then choose the best options it turns up.
The Most Helpful Use of AI Depends on You, the Writer
AI is a tool. “And I don’t think it’s a tool you need to be afraid of,” Tidwell says. “Use it to enhance your weaknesses.”
Tidwell’s weaknesses are over-writing and the time-consuming editing process, so he enlists AI to help him shorten copy and edit more efficiently. Meanwhile, Levin struggles to get started, so she often turns to AI to tackle the blank page. “I just want to get something on the page, and then from there, I can work with it,” she says.
Where do you get stuck in the writing process? That’s probably where AI will serve you best — and where you’ll end up saving time.
So if you, like me, enjoy struggling to outline a story, figure out the throughline, identify the right scenes or quotes or topics to cover, don’t look to AI for help with that. Bring it in at a different stage, as an extra set of eyes or a way to help you reword things or draft the short copy you struggle with.
And remember, as Chayka wrote, “there’s value in the slow labor of writing.”
A machine might be able to string words together in a decently coherent way, but it’s the work of humans, you and me, to make meaning. That’s not work that can be rushed, nor should it be. So do your meaning-making, fellow human, and save time with AI when it makes sense to — when it enhances the work you’re already doing.
%20(1).webp)
.webp)
