One question I get asked a lot these days is ‘do you use AI to create your reports?’ At first I found this to be a strange question because beta reading is, by definition, a test read where you give your opinion. Generative AI, or Large Language Models, don’t have opinions, so they cannot really beta read.

But I learned that this is a very real, very justified fear. I’ve heard several stories of writers finding out that people posing as beta readers or editors have uploaded manuscripts to ChatGPT and prompted it to generate a report (which will ultimately be unusable because AI hasn’t got the ability to do what a beta reader is meant to do). And worse, the writer never gave their permission for the manuscript to be shared. A nightmare. There are already so many things that you might worry about, especially if you are new to the publishing world. And as a beta and sensitivity reader myself, this sort of behaviour poses a reputational risk to not only myself and my colleagues, but the whole industry.

I’ve been thinking a lot about this recently and had to go and educate myself a bit. I’ve come up with some ways that will discourage any people posing as real beta readers from trying to use generative AI programmes to do their work for them.

Wordies, I haven’t written this article to shame anyone, but to make our community aware of thoughts and ideas pertaining to this topic, which can be used to help inform your own decisions.

In order to help me write this article, I asked ChatGPT some questions about its capabilities. I also uploaded a partial draft of this blog post to see if it could respond to comments. You can see a screenshot of those comments and the questions I asked ChatGPT later in the article.

Disclaimer: I wrote most of this before the new GPT-5 came out, so nothing I have written pertains to the new model specifically. However, the principles remain the same.

Why Are People Against AI Use in the Writing Process?

If you are already an active member of the book/writing community, you may have noticed that a lot of people are vehemently against AI for creation of ‘new’ product, and of uploading original content to large language models. But why?

Data Safety: Even if the tool promises privacy, uploading work to the systems feels like a risk. Not only for leaks or piracy, but for the information being ‘learned’ and made available to other prompters.

Theft: Since large language models cannot produce original ideas, their responses can only come from people’s ideas. Putting your name on a book that has been created by using a robot to lift other people’s ideas is seen as theft at worst, or a sort of fraud at best. Many members of creative communities want to protect themselves and their colleagues by not using these tools in this way, or at all.

Ownership: It still isn’t clear who owns work that has been created with generative AI involved with a large share of the process. Several publishers have declared that they aren’t considering publishing work that has been co-authored or fully authored by a large language model.

Environmental Issues: Many creatives worry about the impact AI has on the environment and have strong stances of environmental care, and choose not to use it for this reason.

Devalues other creatives: Writers paying for a service want their money to go to other members of the creative community and professionals. Undisclosed AI use erodes trust in the community.

However, none of these opinions might be shared by the people you employ, work with, or interact with on your writing journey. So what can be done?

• The Abundance Reading Curriculum 2026
• Black British Books to Pre-Order for 2026 (Release Dates & Reader Picks)
• Fancy a crime thriller set in London’s teddy boy era?
• Black British Book List Access
• Questions That Strengthen Your Story: A Guide for Writers, Beta Readers, and Book Reviewers

1. Understanding What Beta Readers Do

This might seem simplistic, but the first thing you need to understand is what beta readers actually do, and if you are ready for one. Beta readers act as test reader, not editors. Their responses and reports should be representative of a real audience, and should be more concerned with the experience of the story, rather than technical aspects. AI cannot respond emotionally to text, and therefore cannot do the work of a beta reader. I’m going to talk a little more about how to make sure you are asking the right questions shortly, but for now, if you want to know more about what beta readers do, you can check out my blog post that tells you all about it!

2. Asking Potential Readers About Their Process

It can be hard to work out people’s stance on AI. Many beta readers do make their views known on their websites and social media pages (and you should be checking those to make sure they are a good fit for your project anyway). But you can also ask potential readers how they record feedback or what tools they use to support their notes. Most readers will be very happy to explain their process to you. Vague answers can be red flags.

NDAs are used primarily to make it clear that the author owns their work, outlines an expectation that their work is handled responsibly and provides a legal boundary against piracy and theft. While it can change the tone of your relationship, it is a professional and empowering way to ensure accountability over the hard work you are handing over. An ethical beta reader will have no problem with signing one as it helps to protect you both (and some can provide one if you don’t have your own – me included!).

4. Consider NDAs

Now, ChatGPT can, in fact, reply to comments in Word allegedly, but struggles with PDF. I say ‘allegedly’ because I wrote this article in Word, added some comments as a test, and uploaded it to ChatGPT after asking it if replying to comments is something it can do. It said it can reply as a contributor on my document, but after two tries, found that it could not. It responded:

While I can edit Word documents and add visible content, I currently can’t generate proper Microsoft Word-style margin comments (the kind you’d see under “Review > Show Comments”) because that requires Word’s own commenting engine.

-ChatGPT

So, asking questions as comments and asking for replies not only makes it clear that the reader is reading everything carefully, but might act as a good AI-deterrent too.

5. Ask Emotion-Based Questions As You Go Along

Asking questions as you go along is helpful, simply because it means that it would be more difficult to isolate the information that a Language Model would require to answer fully. You might ask questions in the comments, within the main body of the text, or even in separate surveys on different sites which would be access via a link. All of these will help discourage reliance on AI.

Asking questions about tension, emotional shifts or emotional resonance, subtext, humour, or specific reactions to certain parts of the text are good because since AI analyses patterns, it cannot effectively answer such questions. AI cannot experience surprise or boredom, track emotional arcs, or reflect deeply on cultural experiences. And unless someone is really good at prompting, it will forget details once messages build up in one session. This means that relying on AI without significant reader input will produce polished and complimentary sounding reports but with vague, disjointed, unusable feedback.

6. White Text and Deliberate Mistakes

I know some writers like to use white text to hide instructions for AI that will cause AI to say something unrelated to the story (you may have heard of the Banana Trap). I’ve also had writers tell me that they changed certain things throughout the manuscript, differently for each reader, so that they know if anything funny has gone on, and if everything is being read.

7. Watermarks

Using watermarks won’t really stop someone using AI if they are determined to, but they are a good visual deterrent and create another copy-paste barrier, especially if the marks are in-page as well as in headers and footers.

---

With all things considered, and the way that new AI-powered technology is quite quickly being integrated into all of our technology, I am aware that our creative works interacting with AI in some way might be inevitable at some point in the writing journey. I am also aware that we can’t dictate how other people write their reports or what they do with their own creative pieces. If you are considering reading for someone else, be mindful of protecting their data, being respectful of their beliefs, and being transparent about your process, so that collaborators can make informed decisions.

Likewise, I am not necessarily encouraging all of the methods I have mentioned (and experienced over the years), as they can potentially create hostile environments. I would personally hope that robust vetting would render most of the other steps unnecessary. So I would say, don’t be afraid to ask questions. Your story deserves the best, ethical treatment, and there is nothing wrong in asking for that.  

---

---

• Instagram
• Goodreads
• Pinterest
• Facebook
• TikTok
• LinkedIn