As published in Forbes
Ask any writer and they’ll tell you: Writing is frustrating. From coming up with ideas to getting the words down, the whole process can turn you into a wreck. So when generative AI burst onto the scene with the release of ChatGPT, some writers and content creators in our network were intrigued by the possibility of speeding up their workflows.
Almost a year and a half since its release, ChatGPT isn’t quite capable of writing an engaging, unique piece of work from beginning to end just yet. (Ask any teacher who’s had to slog through rote, AI-generated student essays.) But more and more writers are dabbling in some AI-assisted writing.
Let’s walk through the writing process from start to finish and identify the parts where AI could come in handy, and where writers should still rely on good old-fashioned human intelligence.
“There is no such thing as a new idea,” Mark Twain once said. “It is impossible. We simply take a lot of old ideas and put them into a sort of mental kaleidoscope. We give them a turn and they make new and curious combinations.”
Though he never lived to give ChatGPT a spin, Twain might as well have been talking about AI when he said “mental kaleidoscope.” (In fact, Mental Kaleidoscope could be a fun name for a writer-friendly AI model!)
Generative AI models are great at spitting out lots of ideas in a matter of seconds. Are they all good ideas? Of course not. But it’s like having a little writing partner on your shoulder, giving you some stuff to chew on during those moments of writer’s block.
No, AI isn’t a cure-all for writer’s block—many of the ideas aren’t completely original. But for now, generative AI is helping writers and content creators get past those dreaded moments when it’s just them and the blank page.
Occasionally, our writers will enter prompts in the brainstorming stage to identify key points and relevant links to sources. This helps to ensure we’re addressing all the main elements in some way (even if it means consciously leaving out certain aspects) and taking recent literature on the topic into consideration.
From tracking down stats to double-checking facts, plenty of people (including some writers in our network) use AI in the research process. Tools like ChatGPT are skilled at boiling down complex ideas into simple prose, especially if you use some creative prompting to limit the amount of jargon in the answer. So naturally, technical writers can use AI to fill in some blanks when they can’t track down necessary subject matter experts for a piece.
But AI isn’t perfect for research. For instance, if you ask ChatGPT to track down more specific stats around AI in the research process, it reminds you that its last update was in January 2022, before the release of ChatGPT itself. So if you’re looking for specific information from the last few years, the current version of ChatGPT may come up short.
Generative AI tools are also prone to hallucination. While this could improve, we’re still at a point where AI’s output must be fact-checked against other sources—because you never know when it’ll decide to be confidently incorrect.
While it might take some creative prompting, AI thrives when it comes to organizing information and giving it a logical flow. Is it remarkably creative? Not really, but it gets the job done. After all, outlines are for structure—you aren’t yet unpacking the meat of your argument.
Structuring a book calls for more complex outlines that AI can’t quite generate with logical consistency or creativity. But when it comes to outlining a short piece of writing, like a blog post, AI can be a valuable asset.
Tip: Don’t give up after a single prompt. Sometimes, you’ll need to prod AI a few times to deliver what you’re looking for (e.g., “more thorough!” “more interesting!” “more humor!”).
AI may be helpful for drafting little snippets of text—a subhead or maybe some alternative options for the title. But it’s hard to imagine many serious writers generating full text from ChatGPT anytime soon. Anyone who has spotted an AI-generated email or LinkedIn message can understand why: AI-generated text is boring. It gets a bit more fun when you tell ChatGPT to “write like Shakespeare,” but that’s just a playful exercise—no one is actually going to put out marketing copy that evokes Hamlet.
To paraphrase Hemingway, “Writing is rewriting.” Editing might be the most important part of the writing process, so is it worth handing the editing reins over to AI?
In some cases, maybe. You can send a written piece to ChatGPT in chunks and ask it to copyedit—though most word processors do that anyway. For shorter pieces, you can feed the entire thing to ChatGPT and have it observe the piece for logical inconsistencies. Or if you’re not happy with some phrasing, you can ask ChatGPT to offer alternatives while you edit.
But no one knows a piece better than the writer who wrote it originally. Much of the editing process is about “killing your darlings” and condensing. For now, generative AI isn’t great at spotting those moments where a piece might drag. Humans, with their lizard-brain need for novelty and entertainment, are still best for that.
Writing, like any craft, is a practice. While it might be tempting to outsource much of the process to AI, coming up with ideas, structuring and editing are valuable skills for any writer to learn. You’ll only learn by doing it, so maybe give it a try yourself first before letting ChatGPT take a crack.