What AI Taught Me About My Reading Habits

I was working on my yearly book writeup when I realised that I have 6 years worth of reading data sitting around in old notebooks. Maybe there was some fun data in there that I could pull up. Maybe I could learn a little something about myself while practicing some AI skills. 

Before getting into what I did, here’s the short version of what fell out of the analysis:

• My reading has shifted permanently from print to audio, but only for certain kinds of books  

• Psychology is the only genre that shows up every single year  

• I overwhelmingly read female authors, except when I’m learning about systems and business  

• 2025 shows a very obvious parenting pivot, and it’s the only category I consistently read in print

None of this was obvious to me before I saw the data.

How I analysed six years of reading data with AI

One of my favourite uses of AI is structuring unstructured data. I recently used ChatGPT to turn our family Wordle group chat into a spreadsheet with everyone’s daily scores, which later turned into a full-blown Wordle Awards ceremony at Christmas.

Old, handwritten reading lists felt like a perfect candidate. I approached this in three steps.

Step 1: Structure unstructured data

I fed photos of my notebooks into Gemini and asked it to turn them into a Google Sheets table containing just two fields: book title and year of reading. This was essentially an OCR task, and I didn’t try to clean anything up beforehand.

Step 2: Enrich the data

Next, I asked Gemini to add additional columns:

• Author name

• Author gender

• Whether the book was fiction or non-fiction

• Literary genre

Of course, this is where hallucinations can creep in, so I spot-checked everything. The error rate came out at around 4%, which was lower than I expected. Interestingly, it consistently struggled with Sara Pascoe books, repeatedly attributing them to a completely made-up author. No idea why.

I then added one manual column: the format I read the book in. I had a vague sense that I preferred audiobooks in some cases and print in others, but I couldn’t clearly articulate the rule.

The literary genres Gemini suggested were far too fragmented to analyse, so I asked it to consolidate my non-fiction into five themes and reassign every book accordingly. Fiction stayed as a single category.

Because the dataset was only 92 rows, I was comfortable trusting the result after a quick scan. At a larger scale, I would have enforced schemas and validation, likely in Colab, to prevent category drift.

Before moving on, I asked one final question: is there another data point we should add? Gemini suggested something I hadn’t considered: whether each book counted as professional or personal reading.

Step 3: Ask the right questions and analyse the data

Any decent analysis starts with the questions, not the charts. I asked myself what I actually wanted to know about my reading habits.

Time-based questions came first:

• Am I reading more or less than before?

• Do I read more or less non-fiction?

• When did audiobooks take over?

Then I looked at how different variables intersected:

• Do I choose audiobooks for specific genres?

• Does author gender vary by topic?

• Are certain subjects tied to print by default?

Countless =COUNTIFS later, I had little summary tables connecting all the main data points.

My reading habits according to AI

Audiobooks have helped me read more

The clearest pattern in the data is a format shift. In 2020, almost everything I read was in print. By 2025, audiobooks made up roughly two thirds of my reading.

After years of averaging a stable 11-12 books, my volume surged to 16 books in 2024 and 29 books in 2025. This is mainly driven by the rise in audiobook consumption.

I love to learn, and audiobooks help me do this when I’m walking, cooking, taking a shower or at the pottery studio.

Print books are for depth

I’ve ended up with a pretty clear split. Audiobooks are for flow, familiarity, and momentum. Print is for books I expect to use.

A big chunk of my fiction listening is functional, especially the Harry Potter re-reads, which I basically use as a sleep aid. Psychology and business also skew audio-heavy, because I want exposure, not study.

However, I remain loyal to Print books for categories like Parenting & Lifestyle. Because these books serve as active resources rather than passive entertainment, I prefer the tactile format so I can easily underline key points, flip through chapters, and refer back to them as I apply the lessons to my daily life.

I prefer female authors

72.2% of the books I’ve read in the last 6 years are written by women. That wasn’t a conscious choice, and it wasn’t always the case.

In 2021, my reading was almost an even split (7 female vs. 6 male). By 2023, I shifted almost entirely to women (11 female vs. 1 male), and I maintained that high ratio through my high-volume year in 2025.

The exception is business and economics. That’s where male authors re-enter the picture.

Seeing this laid out was… uncomfortable in an interesting way. Apparently, when I want to understand people (identity, psychology, relationships, lived experience) I instinctively turn to women. When I want to understand systems (markets, organisations, incentives) I’m more likely to accept male voices as authorities.

I don’t love that this is true. But it clearly is.

Everything is psychology (and I’m fine with that)

This one made me laugh, because it’s the least surprising result of all.

Psychology shows up every single year, regardless of what else is going on in my life. I studied it. I love it. And apparently I keep re-approaching everything through that lens.

• Business? Psychology applied to markets.

• Behavioural economics? Psychology in a trench coat.

• Memoir? A guided tour of someone’s inner world.

• Gender theory? Also psychology, just louder.

Parenting books are the outlier. They’re more practical, more tactical, and for now they sit outside this bucket. But everything else? Same underlying curiosity.

TL;DR: Using AI can be fun

This didn’t make me better at reading, and it didn’t optimise anything.

It was just… fun.

Learning to use AI can be daunting, and these fun, borderline silly projects are a great way to build confidence and problem-solving skills. Out of this analysis work I got to learn about myself, get more AI-confident and have fun. 

And honestly, I like sharing this little snapshot of how I think, learn, and read with the people who care to look. Thank you for caring to look.