…also don’t tell lies. But I’m getting ahead of myself already.
I keep running into people online who openly say that they use AI to do their writing for them. Now, technically, they may say they use AI to “help” their writing. It doesn’t matter. The distinction is immediately lost on everyone who sees that assertion, because when AI writes for you it does a lot more than just “help,” and everyone knows it.
My policy is that I never let AI draft anything for me that has my name on it. Not one sentence. Nothing. Ever. (You can download my official AI writing policy here.) When I was a year behind in writing Taking Testing Seriously, was I tempted to use AI to speed things up? No. It would be like hooking a motor to a stationary bike and calling that exercise. It would be like taking a helicopter to the top of Everest and saying that I summited.
I once ghost wrote an article for someone who was a poor writer. As I did so I had tremendous influence over the ideas, not just the style of writing. I chose what to put in and leave out. He could have had more influence, but to do that he would have had to take a very active role. Yet, he didn’t. The article was published in a book, and it had his name on it, but it was substantially my work. I did this because, at the time, I was trying to keep the peace in a small community, but it bothers me, to this day, that I helped burnish the professional image of a man who didn’t earn it. Maybe, today, the same fellow would use ChatGPT.
I have also helped other people write in such a way that we deeply collaborated. The chapter on Prospective Testing, in Taking Testing Seriously, was based on the somewhat jumbled but information-rich notes from, and conversation with, Jeff Nadelman. I respected Jeff when I started writing that article with him, and I respected him more, dozens of Zoom calls later, when we finished. Both our names are on that one. If you look at that chapter, Jeff’s key tangible contributions are stories separated out from the main body of the article. Those are his personal experiences. Apart from that, he contributed by talking with me for many many hours. That chapter is absolutely his as much as mine.
If AI deeply collaborates with you to write something, why am I saying you shouldn’t say you used AI? Because all I have is your word for it that you did any work at all. Unless I know exactly how you did that work– which would have to have been from personal observation– I can’t know the truth. (Yet you can check with Jeff Nadelman to confirm how we worked on that article.) So, in my mind, I am forced to put a big fat asterisk next to anything you do. Is that what you want? Meanwhile, if AI drafts a document for you, and you claim it’s your work, that’s a lie. Don’t tell lies.
Now put two and two together: Don’t tell me that AI helped you write, and don’t tell me that you did all the work when you only did some of it. Therefore, don’t use AI to write things for you that you present as your own work. If you DO, I will treat all your work as if it is slop or spam. And not just me, the entire community of serious professionals will discount “your” work. It used to be that we could take it on faith when someone turned in work that they claimed was theirs. No longer. Now it depends on your reputation.
Consider what you would think if someone said this to you: “I’m a skilled liar. I frequently tell lies. But don’t worry, I wouldn’t lie to you!”
A liar might say “I wouldn’t lie to you.” But ONLY a liar would say “I frequently tell lies.” Similarly, only someone who uses AI to help their writing would say “I use AI to help my writing.” And anyone who uses AI to help their writing may use it to do all of their writing. It’s such a steep and slippery slope to full laziness. We all know that. We are all thinking that.
So don’t say it. Don’t do it. In a world where there are ads with Matthew Broderick exhorting you to take a day off and let AI do your job, we are all perched on the top of a very steep cliff. Will you throw your reputation down the abyss?
I love how tech companies managed to invent a highly advanced digital lie engine and then acted surprised when it started making things up. It feels like giving a toddler a microphone and wondering why it is screaming. At what point do we stop calling these glitches and start calling them features?
Love this, and strong agree. My own personal rule is that I don’t ever let AI write for me.
That said, I certainly do use AI frequently in the _process_ of writing—as a conversation partner, or to give me editing suggestions (not rewrites, just advice like “this part drags”, etc).
I’ve been toying with the idea of building a sort of public declaration mechanism (a la Creative Commons tags) where people can be upfront and granular on exactly how they used (or didn’t use!) AI for a given project or output. For sake of discussion, I threw a mockup up at https://howiused.ai (but I imagine it’d have to be a lot more sophisticated than that).
Do you think that could be a useful capability? Obviously, it doesn’t substitute for honesty or trust (you could still lie with a mechanism like this), but it at least sharpens your assertions to the point where you can give some nuance, rather than defaulting to all-or-nothing (“it was all AI” or “no AI at all”).
[James’ Reply: It seems like a good idea… except, this is not just a technical assertion, it’s a moral stance. The more it’s a moral stance, the only people who will follow this convention are the people who can give the respectable moral answers, or else the people who will tell lies. This is also a gift to the AI industry, since they would want to train their products on sites that have no AI-generated content.]
I’ve been guilty of this myself — casually dropping “I use AI to help with writing” when really I mean it’s doing most of the heavy lifting. Reading your point about the stationary bike with a motor hit different. Makes me think about what I’m actually claiming credit for, especially when I’m putting my name on something. The ghost-writing example is uncomfortable in the best way — it’s easy to rationalize shortcuts when you’re busy, but yeah, there’s a difference between collaboration and just outsourcing your thinking.
I agree. Even terms like Vibe Coding Hurts.
I feel it hurts more when people use these terms with customers and clients.
Thank you for this.
I used to mention that “used AI for documentation work”; now I am wondering if I should do that or not…
Honestly, I’ve been thinking about this a lot lately. There’s something that feels off when people casually mention using AI for their writing—you’re right that “help” isn’t really the right word when the AI is doing most of the heavy lifting. I’ve been wrestling with where to draw that line for content on Sell Lemons Info, our Roblox Sell Lemons guide, and it’s made me more intentional about what I actually write versus what I let AI touch. Your point about the helicopter to Everest really stuck with me—there’s a difference between using tools and just outsourcing the work entirely.
I recently had a tester put a regression test strategy in front of me that had clearly been produced by an LLM. The test strategy was context unaware, and completely unworkable – (unsurprisingly), but fully buzz word compliant.
I limited the damage by refusing to review the work, beyond my initial cursory inspection – but I am still annoyed by even the small amount of time I wasted looking at it.
What troubles me about this is that I suspect that I am fighting a losing battle here. Testing as a profession had enough struggle with low professional standards already. LLM’s have only worsened these problems.
[James Reply’: What does it mean to lose this battle? Does it mean that bad work becomes normal? Bad work is ALREADY normal. So, I think this is the ordinary process of good people striving to do good work in a mediocre world. The Hittites struggled with this, I’m sure. So did the Romans and the Carolingians. And so do we.]
I don’t lose this battle personally, if I don’t give in by compromising my standards and start accepting poor quality work.
And I am not planning to give in.
But I have been around long enough to remember how testing was impacted by some of the sloppy thinking that organisations embraced when they first started implementing agile (for example). Test leadership, test management and test advocacy were all really badly impacted and many talented testers ended up exiting the industry.
Having lived through that (and survived it) I can see some parallels with what is coming up over the next few years.
Hello! Interesting blog entry. I am wondering if you are suggesting that AI is good as a tool, like to code other tools, etc. But when it comes to writing, then a line has to drawn? I agree and i also think we have to trust only people who doesn’t tell lies and doesn’t use AI for writing.
I believe AI for writing could help some person achieve some goal of theirs, so then why not use it? because it brings more bad than good out of it?. But for the person using AI, it could bring more good than bad?
I once used AI to write an email for me, it was to an agency i had little respect for. It was a sloppy email and i shouldn’t have sent it, dangerously easy it was to generate and they had to spend more time to answer it, then what i spent telling the AI to send it.
As Brandolinis law says, it really is easier to generate bullshit than to refute it, therefore we must know people we can trust and trust that those people speak for themselves and what they truly believe than manipulation through AI to achieve some goal.
Do you think there should be a line like this towards all types of writing? what about AI generating technical documentation? Maybe the line should be limited to text meant to change peoples view on some thing/things. Because to me it doesn’t seem wrong to use AI to explain some technical thing that it’s good at?
[James’ Reply: I edited your text to remove several spelling errors. This makes you look better at English than you actually are. In a way, it misrepresents you, but I think it’s more cosmetic than essential, so I went ahead and made the changes.
When you use AI in a way that creates a deceptive impression of who you ESSENTIALLY are, I think that is immoral. That’s where I draw the line. This is why I choose my words organically when I draft any text that has my name on it. I agree that official letters, written for legal purposes don’t count against you if done with AI (or even drafted by a lawyer), because LEGALLY you are not presenting your personality, only your legal business.]
Thanks James, I think that is a good way to draw the line.
My English is unfortunately not that good. But thanks, I will try to not make to many errors.
I was wondering also about using AI to create a deceptive impression of what a company essentially is. Then it is used only to present a legal business so I suppose the AI part is okay, but deception in general is immoral I think. But maybe less immoral than if it was done by a person ? As Companys are known to deceive and by other means try to get more money so it is kind of expected.
[James’ Reply: Deception is not inherently immoral. The ethical problem occurs when deception prevents someone from the reasonable expectation of acting in their own best interest. If you are playing poker, then you can morally deceive other players because they have no reasonable expectation to know the cards that the other players are holding.]
This is a thoughtful perspective. The distinction between AI-assisted and AI-authored work is becoming increasingly blurred, and your point about reputation being the ultimate currency resonates strongly. In creative fields like AI image generation, the same principle applies — claiming credit for work that was fundamentally generated by a tool undermines the credibility of everyone in the space.
Another great piece James.
I would add that beyond the creation of AI generated documents, what is the purpose of the document that AI has created? Does it have value – for example a test plan or a bug report or tutorial document.
If AI has created a document and it has not been thought about and edited by a human that cares then how can I as the reader of that writing verify that it will be useful to me or contain something truthful(or something ghosted by the AI)?
A bunch of times now I have received bug fixes from developers with explanations that are clearly the output of an AI and made no sense. I had to go back to the developers and they had to reread the issue to understand it well enough to explain it to me so that I could test the fix. How do I trust that the fix is worth me even testing? Am I testing the AI’s work or is it the developers’ work?
I’ve found AI to be a really useful editing tool, like spellcheck or grammar check. I have thought about and wrote this document, now I want AI to act as a copy editor and make sure it is clear and makes sense. I review each suggestion in the same way I do with spellcheck. Spellcheck is still not infallible and still makes mistakes.
If the goal is to create something that kind of looks “right” and ticks a box, then sure use AI but if I want to create a test plan or create a bug report or write a tutorial for my teammates then I care about it and adds value in some way then I want to take the time and effort to think about it and make it useful and AI cannot generate that because as you have said before AI is not human and does not care.