What Happened When Real Readers Started Using AI Book Companions™?
What the early reader testing showed me about how people actually use them, including a few things I hadn't anticipated.
When I started building AI Book Companions™, I spent quite a lot of time testing them myself. I wanted to know whether they understood the book, whether they could work with the author's methodology and whether they knew what to do when somebody asked a question that went beyond the material.
There's quite a lot involved in getting those things right, particularly when you're working with somebody else's intellectual property. You need to understand how the author thinks, why they've structured their work in a particular way and what they would be comfortable having an AI say on their behalf.
But testing something yourself is quite different from putting it in front of somebody who has never seen it before.
I knew what each Companion was supposed to do because I'd built it. I knew the questions I was trying to answer and the situations I wanted to test. What I didn't know was how somebody else would approach it, what they would ask or whether they'd find it useful in ways I hadn't considered.
That was one of the reasons the early reader testing was so interesting, particularly with the Self-Love Companion I created for Sarah Grace's book, From Shadows to Self-Love.
Sarah's book explores how childhood experiences can influence the way people think, feel and behave as adults. We wanted readers to have somewhere they could explore the ideas in the book, work through questions and make connections with their own experiences.
What happened when people started using it gave us a much better understanding of what that could look like in practice.
Readers were bringing their everyday experiences into the conversation
One of the things I noticed was that people weren't necessarily opening the Companion because they had a question about a particular chapter. They were using it when something happened in their lives and they wanted to understand what they were feeling.
One woman had been out for a drink with friends and woke up the following morning feeling overwhelmed by anxiety and guilt. She told Sarah that she wouldn't normally have spoken to anybody about how she was feeling, but she decided to open the Companion.
She described crying during the conversation, working through a small exercise and beginning to feel calmer afterwards.
Another reader used it to explore why she'd always disliked Christmas. Someone else had been upset by an unkind encounter with a stranger while walking her dog. As she talked about what had happened, she began connecting it with experiences of being bullied at school.
These were all quite different situations, and none of them was something we'd specifically designed a conversation around.
The readers were taking the ideas from Sarah's book and exploring them in relation to something that was happening to them at that moment. They didn't need to remember which chapter they'd read or work out which exercise might be relevant before starting.
I found that interesting because it showed me how somebody might return to a book's methodology long after they'd finished reading it.
One reader used it to help pack her holiday suitcase
This was probably one of the examples that stayed with me most.
One of Sarah's readers was getting ready to go on holiday and was having a difficult time deciding what clothes to take. Choosing what to wear could be quite stressful for her, and she'd been struggling with both the clothes she already owned and some new things she'd bought.
She decided to talk it through with the Companion.
During the conversation, she explored why certain clothes were making her feel uncomfortable and what she actually wanted to wear. She eventually settled on a travelling outfit and three outfits for the holiday.
She told Sarah she'd packed her bag within ten minutes.
Later, she reported that she'd been comfortable with the clothes she'd taken and hadn't experienced the same difficulties while she was away.
Her description of what the Companion had done was that it had “GOK WAN-ND ME”.
I remember reading that feedback and thinking how unlikely it was that we'd ever have come up with packing a holiday suitcase as a use case when we were designing the Companion.
But the conversation still made sense within Sarah's work. The reader was exploring her feelings and reactions, using the approach the Companion had been designed to support.
It was a useful reminder that I don't need to predict every situation somebody might bring to a Companion. I do need to be clear about its job and make sure it can work appropriately with the situations readers choose to explore.
People don't always arrive with a neatly formed question
Something else I found interesting was the way readers described what they were experiencing.
One woman was trying to explain something she was feeling internally and called it a “fuzzy rectangle floaty thing”.
Another said the Companion had helped her explore things she'd been sensing but “couldn't put a name to”.
I think that's quite important when you're looking at how people use a book like Sarah's.
When somebody reads about a particular experience or pattern of behaviour, they might recognise something about themselves without knowing how to explain it. They may have a feeling that something is relevant but not be able to put it into words straight away.
In the testing, people were able to start with whatever language they had and continue the conversation from there.
It also made me think about the way we sometimes design learning materials. We tend to assume people know what they want to ask, or at least know which part of the material they need help with.
The reader feedback suggested that isn't always how people approach things.
Readers were asking the Companion to communicate differently
Another thing that came through during testing was that readers had their own preferences for how they wanted information presented.
One tester referred to her “ADHD brain” and asked the Companion to break things down into “little chunks”, using headings, concise bullet points and explanations of the reasoning.
Another tester was dyslexic and found being able to speak her questions and listen to the responses particularly useful.
She said:
“I am dyslexic so being able to speak into it and have it speak back was really helpful.”
Those are individual experiences, so I wouldn't use them to make broad claims about accessibility. But they did show me something useful about how people were choosing to work with the material.
The author had written one book, but readers weren't all approaching it in the same way. Some wanted to talk things through, some needed information broken down differently, and others were using the conversation to help them find the words for something they were experiencing.
That gave me more to think about when designing the reader experience for other Companions.
Did it still feel like Sarah?
This was something Sarah had been particularly interested in.
She wanted the Companion to reflect her work and the way she communicates. She didn't want readers to feel as though they'd simply been handed a generic AI chatbot with her book uploaded into it.
Some of the feedback was lovely.
Readers said:
- “I can hear your voice.”
- “It sounds like you.”
- “It's like having Sarah in my pocket.”
I think it's worth being careful about what that means. The Companion isn't Sarah, and it shouldn't pretend to be. But the readers recognised her language, her approach and the way she explains things.
That mattered because we'd spent time during the build working through how the Companion should communicate, as well as what it needed to know.
Seeing readers respond to those details was encouraging. It suggested that the work we'd done to understand Sarah's approach was coming through in the experience.
What I learnt from putting the Companions in front of real readers
By this point, I'd already become quite clear that every AI Book Companion™ needs a specific job.
That job comes from the book and what the author wants readers to be able to do with the material. It's one of the reasons I don't think you can simply take the same Companion structure and use it for every book.
The reader testing added another layer to that understanding.
I'd been concentrating on whether the Companion could do what we'd designed it to do. The readers showed me how they might choose to use that capability in their own circumstances.
With Sarah's Companion, somebody might want to explore an idea from the book. Another person might have had an upsetting experience and want to understand why they'd reacted in a particular way. Someone else might be trying to describe a feeling they couldn't quite explain.
The Companion still needed to stay within the boundaries we'd established.
That's particularly important with Sarah's work because some conversations may touch on sensitive personal experiences. Sarah is clear that the Companion doesn't replace her work as a therapist, and there need to be limits around what it can appropriately help somebody explore.
The testing gave us useful examples of how readers were using the Companion, but it didn't remove the need for those boundaries. If anything, seeing the range of situations people brought to it made me more aware of their importance.
The relationship with the book started to look different
One of the things Sarah had originally imagined was that somebody would be reading her book, come across something they wanted to understand better and open the Companion to explore it.
That was still a perfectly good reason to use it.
But the testing showed that readers might also come back to the Companion when something happened later and they remembered an idea from the book.
They weren't necessarily sitting with the book open in front of them. They were drawing on Sarah's work in the middle of everyday life.
Sarah has also begun using the Companion alongside some of her one-to-one work, giving clients somewhere to explore questions between sessions and bring anything they want to discuss back into their work with her.
I wouldn't suggest that every author will see the same results or that every book needs this kind of experience. Sarah's book deals with material that readers can explore in relation to their own lives, so there's a clear reason for somebody to want to return to it.
What I took from the testing was a better understanding of the relationship between the book, the Companion and the reader.
I'd started by thinking carefully about how to make the author's knowledge available in a conversation. Watching readers use it showed me how much their own circumstances influence what that conversation becomes.
And that's something I now think about much earlier when I'm working on a new Companion.
What might this mean for your book?
If you've written a book that includes a framework, exercises or ideas people can apply, it's worth thinking about when somebody might want to return to that material.
They might need help understanding something while they're reading. They might want to work through an exercise, explore an idea in relation to their own circumstances or revisit part of your methodology weeks later.
It also depends on the book. Some books lend themselves naturally to that kind of conversation, while others may not gain much from having an AI Companion at all.
For me, the starting point is still understanding what the book is there to do and whether a conversation would genuinely add something useful for the reader.
I explore this in more detail in The AI Book Companion™ Field Guide, including what I've learnt from creating Companions for different authors and different kinds of books.
The e-book is £9.99 and the paperback is £24.99. It also has its own AI Book Companion™, so you can see how the approach works while you're reading.