FIR #530: Has AI Created a Demand for Human Ghostwriters?

The volume of AI slop appearing as thought leadership on LinkedIn has, according to the Financial Times, led to a surge in interest in human ghostwriters. In this short midweek episode, Neville and Shel discuss whether this makes sense or if it’s just another case of people using AI poorly. **Links from this episode:** - AI is creating ghostwriting jobs on LinkedIn - Using AI to Write at Work Is Basically Fine — Unless You’re a Writer - So AI helped write it. Who cares? - FIR #526: Forget Anthropic’s AI Watermark. Can You Defend Every Sentence? * * * The next monthly, long-form episode of FIR is tentatively scheduled to drop on Monday, October 26. We host a Communicators Zoom Chat most Thursdays at 1 p.m. ET. To obtain the credentials needed to participate, contact Shel or Neville directly, request them in our Facebook group, or email fircomments@gmail.com. Special thanks to Jay Moonah for the opening and closing music. You can find the stories from which Shel’s FIR content is selected at Shel’s Link Blog. You can catch up with both co-hosts on Neville’s blog and Shel’s blog. Disclaimer: The opinions expressed in this podcast are Shel’s and Neville’s and do not reflect the views of their employers and/or clients. * * * **Raw Transcript:** **Shel Holtz:** Hi, everybody, and welcome to episode 530 of *For Immediate Release*. I’m Shel Holtz. **Neville Hobson:** And I’m Neville Hobson. AI was supposed to put ghostwriters out of business. Instead, according to the *Financial Times*, something rather different is happening. LinkedIn is awash with generic AI-generated posts, and the flood of polished but forgettable content is creating new demand for human ghostwriters who can help executives sound human again. There’s a nice irony in that, but it also raises a question I keep coming back to: If I use AI to help write a blog post, a LinkedIn update, or an article published under my name, is AI effectively my ghostwriter? Shel, you and I have touched on this territory before, most recently in FIR 526 in August, when we discussed Anthropic’s text watermarking. We argued that detecting whether AI was involved in writing something tells you remarkably little about who actually did the thinking. A more useful test is whether the author understands the argument, stands behind it, and can defend every sentence. I developed that argument afterward in a blog post titled “So AI Helped Write It. Who Cares?” My point wasn’t that authorship doesn’t matter. It does. But there’s an important distinction between asking AI to do your thinking for you and using it to help express thinking that is yours. Now, a *Bloomberg Businessweek* article adds another twist. It argues that, for many people at work, using AI to write is simply becoming normal. Writing isn’t necessarily their craft. It’s a means of communicating an idea, answering a customer, updating colleagues, or getting something done. The article introduces another tension. Some businesses actively expect employees to use AI for writing because it saves time. But it also describes the opposite problem at one company, where AI-generated documents became polished, lengthy, and difficult to read. The CEO concluded that AI was making writing quicker for the sender but slower for the recipient—a point we touched on in our last discussion—and asked employees for rougher versions containing their actual thinking instead. There’s a novel idea. That brings us back to ghostwriting. We accepted long ago that a CEO might put their name to words written by a communication professional, provided those words express what the CEO thinks. So, is AI-assisted writing fundamentally different? For me, this raises several questions. If the ideas, argument, and judgment are yours, does it matter whether you, a ghostwriter, or an AI produced the sentences? Where is the boundary between assistance and substitution—editing your thoughts versus generating thoughts for you? In an environment where producing polished prose is becoming almost effortless, will distinctive thinking become more valuable than distinctive writing? And would you feel differently upon discovering that someone’s LinkedIn post was written by a human ghostwriter rather than ChatGPT? If so, why? And perhaps more importantly, does it matter? **Shel Holtz:** Yes, it does matter. You have to consider the audience’s expectations. Those matter as much as the intentions of the person who drafted the speech, article, or whatever it might be. Think about the difference between an operational update, a CEO’s personal account of a career failure, and a condolence message to an employee. Those messages could be accurate, carefully reviewed, and approved. On the other hand, readers could reasonably expect a particular level of personal involvement from the person whose name appears on them. I think that makes a difference. But I’d like to step back for a minute and explore ghostwriting itself. I’ve known a couple of people who made their livings, or part of their livings, as ghostwriters. You knew the late Shel Israel, a very good friend of mine. **Neville Hobson:** Sure. **Shel Holtz:** He was a ghostwriter. He wrote books for people who didn’t know how to write, but he didn’t spend most of his time writing. His clients hired him because they liked his writing and his approach to getting to know them and their styles, so what he wrote sounded like them. How did he do that? Through a lot of interviews—hours and hours, dozens of hours, on the phone and in person. He challenged the generalities he heard from clients and drew out experiences they might not have thought mattered. As an experienced ghostwriter, he understood which experiences would lend color to the manuscript and what readers would care about. The *Financial Times* story opens up a more interesting possibility than simply saying humans write better. Perhaps the valuable service of ghostwriting was never just sitting there typing sentences. One person featured in the article said in a public response that AI remains integral to his business, but people still handle the conversations and relationships. Another person I know ghostwrites for an online service. Someone signs up and says, “I want to write a memoir. It’s only going to go to 10 people in my family, but I want somebody to write it.” The service matches that person with a writer. There’s a prescribed number of meetings, but they talk for hours. The writer takes notes and produces a first draft. If I take all those notes and everything I’ve learned about the subject, understand their writing or speaking style, and incorporate that into what I ask AI to produce or help me produce, am I still doing the lion’s share of the work as a ghostwriter? I think I am, especially if the person tasked with doing this isn’t a writer in the first place. We’ve talked before about people in organizations who might be assigned this work but can’t write their way out of a paper bag. Writing wasn’t what they studied, and it isn’t their area of expertise. If AI can help them use the material they’ve gathered by doing the work of a ghostwriter, I have less of a problem with it. **Neville Hobson:** I have a similar view. The FT article includes a good quote from Bernie Hogan, an associate professor at the Oxford Internet Institute. The journalist asked whether LinkedIn ghostwriters could help cut through AI slop. The part that caught my attention was his point that the message behind a post can still be sincere, even if you didn’t write it. He compares it to buying a greeting card: Someone else wrote the words, but the sentiment can still reflect what you mean. I agree with that. His further point is where the rubber meets the road. Because we can’t tell when a LinkedIn post is ghostwritten, is it authentic in readers’ eyes, even if it broadly reflects what the person posting it believes? That sums up the situation we’re in. I don’t think there will ever be agreement on the broad question of whether it’s wrong for someone else—or an AI—to write something you publish under your name. We’ve discussed the details many times: Is it your thinking? Did you contribute to it? Or did you give AI some material, ask it to write something, and post the result without further involvement? It isn’t a simple question. I keep coming back to this: Would you feel differently if you discovered that someone’s LinkedIn post was written by a human ghostwriter rather than ChatGPT? If so, why? I’d particularly like critics of AI-assisted writing to answer that. **Shel Holtz:** It would be interesting. I like the greeting-card analogy. The greeting was written by a copywriter at Hallmark, American Greetings, or somewhere else, but you don’t just grab the first card off the rack. People spend 15 or 20 minutes looking. **Neville Hobson:** You might. **Shel Holtz:** You might, but I’ve never seen anybody in the greeting-card aisle just grab a card and go. They pull one off the shelf, open it, read it, put it back, and try another. They’re looking for the one that expresses the sentiment they have. Then, when they get home, they write a little more. I would consider that the human in the loop. Human ghostwriting deserves the same scrutiny as AI ghostwriting. Someone assigned to write a LinkedIn post can manufacture the persona of the person they’re writing for. They can avoid the hard work, fabricate a profound message, or give someone credit for expertise they don’t have. An AI-assisted post grounded in a leader’s actual experience, with that experience shared as part of the process, could represent the person more faithfully than a human ghostwriter does. So, are we defending authentic communication, or are we defending an outsourcing arrangement we’re already familiar with? **Neville Hobson:** I think we’re defending authentic communication, but the definition of “authentic” seems to be a movable feast. I don’t believe we’ll resolve this argument in the near future. Where I land—and I think we discussed this in episode 526—is that you, as the human, need to be confident in your thinking and in how the text expresses it. If you are, you can comfortably continue what you’re doing. Apply that to a human ghostwriting a LinkedIn post for a CEO, senior manager, or someone else. The post will appear on that person’s profile under their name. That’s common practice; I know many people who do it. The question isn’t simply whether the manager or CEO is happy with it and praises you. Are you satisfied with it? Does it meet your standard for authenticity? You might outline your ideas, share them with your AI assistant, explain the assignment, and ask it to suggest the best way to tell the story. You might then ask for a proposed draft. It comes back, and you think, “This is a reasonable start. I can work with this.” You might edit it extensively and create something fresh, then pass it back to your AI assistant. It might suggest three changes. You could decide the second suggestion is useful and incorporate that. That process resembles what you might do with a human ghostwriter. So, my question is: What’s the difference? Does it matter? I’ve been accused of being terribly naïve for asking that, but I don’t believe it matters if you’ve satisfied those questions about authenticity, honesty, and confidence in what you’ve created with AI’s help. Others will disagree strongly. I still see absolutist comments on LinkedIn from people saying you can always tell when something was written by AI. I used to ask how they could tell, but I’ve stopped because the responses often suggest they don’t really know. I don’t believe you can reliably tell. I particularly enjoy examples where somebody submits a passage by Charles Dickens to an AI detector and it comes back as 98% AI-written. Or somebody tests an article they wrote for a magazine in 1995, and the detector says the same thing. What can you do with that? Let me ask you, Shel: Would you feel differently if you discovered someone’s LinkedIn post was written by a human rather than an AI? **Shel Holtz:** Generally, no, assuming they used the AI well. I’ll disagree with you on one thing, but only to an extent. You can sometimes recognize AI writing when it was given a weak prompt and defaults to a familiar approach. But that’s different from giving it detailed information about the person whose byline will appear, helping it understand their writing style and thinking, sharing other things they’ve written, and providing guidance on what the piece should say. You might even interview the person and share the transcript. We’ve had this conversation since the early days of blogs, when people debated whether communicators could acceptably ghostwrite for CEOs. I remember where we came down on it. If the CEO calls and says, “Write a thought-leadership post for me on this topic,” then hangs up and leaves you to invent it from scratch, that isn’t acceptable. If you go into the CEO’s office, interview them, and turn your notes into a post, that’s different. Bill Marriott used to record his thoughts and give the recording to his communicators. They didn’t transcribe it word for word because spoken remarks needed editing to work in writing. But the substance was his. Did he sit there and type it? Was every word one he had chosen? No, but it was his post. In that case, ghostwriting is perfectly acceptable. Today, AI can assist with that process. If we maintain the practices that make ghostwriting effective, authentic, and ethical, I have no problem with it. It doesn’t matter to me whether a human or AI helped produce the wording, as long as it reads well and reflects what the person would have said. That person also needs to approve it. We haven’t mentioned that yet. **Neville Hobson:** I agree. The process you described, using Shel Israel as an example—interviewing the subject, spending hours preparing, and understanding their thinking—would still apply. I particularly like the idea of using voice-recording tools that connect to Claude, ChatGPT, or another AI assistant. You could ask the CEO to do a brain dump on the topic, as unstructured as they feel comfortable making it. The recording uploads, you get a transcript, and AI can help interpret or organize it and suggest where you might take the story. That can save a tremendous amount of time compared with doing everything the traditional way. There’s a benefit to AI assistance, and I hope professional communicators make the most of those tools. Of course, some people will skim the output or post it as it is. We’re talking about people who bring their professional skills and judgment to the work. Based on the criteria you outlined, I wouldn’t feel differently about a post written with AI assistance. If those steps had been followed, I’d be comfortable with it. Yet I can think of half a dozen people who would strongly disagree. This seems to be a never-ending discussion. **Shel Holtz:** It is. One other thing we haven’t discussed is disclosure. I don’t think most ghostwriting requires disclosure. I don’t think Shel Israel’s name appeared on the books he ghostwrote. My friend Bill, who writes memoirs through the online service, doesn’t have his name on them either. Those memoirs have a small distribution: “This is my memoir for my children and grandchildren.” But you may remember an instance when a company wanted a PR agency to tweet on the CEO’s behalf while he was at a conference. The agency initially refused, saying, “We’re not you. We’re not going to pretend to be you.” They eventually agreed that roughly every fifth tweet would remind readers that the agency was tweeting on the CEO’s behalf. You have to think about where people might feel deceived and where the writing arrangement simply doesn’t matter. **Neville Hobson:** I agree. That’s our latest 50 cents’ worth of opinion on this topic. We’d like to hear your thoughts. If you’re a ghostwriter listening to this, we’d especially love to hear what you think. It’s an ongoing conversation. **Shel Holtz:** And that’ll be a 30 for this episode of *For Immediate Release*. The post FIR #530: Has AI Created a Demand for Human Ghostwriters? appeared first on FIR Podcast Network.