Why?
Short answer: ChatGPT offered a recommendation and it aligned surprisingly well with my tastes.
Longer answer: for the past few years I’ve wrapped a project around my reading as a way to keep things fresh, address gaps, and broaden my tastes:
- In 2022 I read more women than men
- In 2023 I read more books in translation than in English
- In 2024 I attempted to read more books I already owned than ones bought throughout the year
In 2025 I couldn’t think of anything. Rather than put unnecessary pressure on myself I decided to just read whatever for a year.
Until March, when I asked ChatGPT to clarify something in a book I’d just read. It offered recommendations at the end of that conversation, I read and enjoyed one, and that planted the seed of the project.
That’s awesome / That’s stupid
Maybe. I’d love to know which you think it is. I’m tracking responses and will report on the ratio periodically.
What have you read so far?
The first recommendation I read was White Noise by Don DeLillo. This was off the back of the conversation about what I enjoyed (and didn’t) about Infinite Jest.
ChatGPT: Would you like recommendations based on what you enjoyed in Infinite Jest?
Me: That would be good. I fancy something shorter next 😂
These were the recommendations. The phrase “Hitler Studies” and the deadpan absurdity of the description swayed me.
1. Thomas Pynchon – The Crying of Lot 49 (1966)
- Why: If you liked Infinite Jest’s sense of conspiratorial weirdness, but want something much shorter (~150 pages), this is Pynchon’s most accessible novel.
- What: A woman uncovers a possible underground postal service conspiracy in 1960s California.
2. Don DeLillo – White Noise (1985)
- Why: DeLillo was a huge influence on Wallace, and this novel tackles media saturation, modern anxiety, and consumer culture—all things Infinite Jest explores.
- What: A professor of “Hitler Studies” deals with an “airborne toxic event” while wrestling with existential dread.
3. Lorrie Moore – Birds of America (1998, short stories)
- Why: If you liked Wallace’s humor but want something more emotionally direct, Moore writes beautifully crafted funny-sad short stories.
- What: Stories about relationships, loneliness, and absurd modern life, with razor-sharp wit.
After that we spoke a bit more about my tastes, and I read two of the recommendations that followed:
- Room Temperature by Nicholson Baker — fantastic
- This Census-Taker by China Miéville — fantastic
I ordered five more while reading these:
- Open City by Teju Cole
- The Pillow Book of Sei Shōnagon
- Gilead by Marilynne Robinson
- The Book of Strange Things by Michael Faber
- The Restraint of Beasts by Magnus Mills
Halfway through Room Temperature I’d pretty much commited to the project. I drafted the original version of this post between Room Temperature and This Census-Taker, then published it halfway through Open City.
I officially handed over the reins to ChatGPT on April 15th.
What’s coming up?
Beyond the handful of upcoming recommendations in the full list, I’m not sure. This is because the recommendations come from an ongoing conversation about what I think of each one.
How’s it going?
Subjectively, very well! The books have mainly been great. All of the recommendations sound really interesting. Most of them are new to me.
Statistically it’s looking good, too. My average rating for books read each year is as follows:
- 2022, women beat men (avg 3.3/5 vs 3.1/5)
- 2023, translation beat English (avg 3.6/5 vs 3.3/5)
- 2024, already owned beat bought new (avg 3.7/5 vs 3.3/5)
And for the seven books I’ve finished so far in this project, 4/5.
It’s also leading to interesting thoughts that I might write more about later:
- ChatGPT is great at picking up on themes and through-threads in my tastes that I’m not aware of, and building on them.
- A human could do this (a librarian, a bookshop owner, etc) but it would feel presumptuous to take up so much of their time.
- The conversation about my tastes and what I enjoy is a really enriching and interesting one to be having.
- ChatGPT says it hasn’t read the books, per se, and that its understanding comes from reviews and analysis, but I’m aware that the way AI tools access to copyrighted material (books, articles, etc) is problematic. At this stage the problem feels partially offset by having books and authors brought to my attention that I may not have found otherwise, but I look forward to thinking/learning more about this issue and hopefully talking to authors about it to see what they think. (Small side note: ChatGPT is aware of my book and personally I find that quite cool. Writing isn’t my living though and I imagine many others will feel differently).
How do you get the recommendations?
The first were off the back of discussing what I liked and didn’t like about Infinite Jest, as I mentioned earlier. That book is definitely not representative of my tastes, though, so there’s been lots of further conversation.
Here’s a brief outline of that until now. I’ll be writing more about this process later as well, as some of it is really interesting.
- I mentioned enjoying The Mezzanine by Nicholson Baker and asked for that to be factored into recommendations.
- The recommendations that followed included How Should A Person Be? by Sheila Heti, which I’d then just finished reading and enjoyed. Seeing books I’d found and enjoyed in the recommendations was encouraging.
- I asked ChatGPT if it had a favourite book. The reply mentioned books that “resonated”, and I asked what they resonated with exactly, as it implied ChatGPT had “personal” preferences, which went against my understanding of the tool.
- We spoke briefly about whether other instances of ChatGPT would have similar preferences.
- I asked what ChatGPT considered to be “the best” books, as opposed to its favourites. The response was split into four categories: Most Formally Accomplished, Most Influential, Most Emotionally Profound, Most Enjoyable.
- Then we spoke about the canon, its merits, and the risks of the feedback loop of canonisation. That is, where books remain in it because of self-perpetuation rather than inherent merit. We also spoke about deserving books that might be excluded as a result of this.
- This led to conversation about reading in a way that deliberately seeks out books that have fallen out of print, never been translated into English, or are otherwise in obscurity. ChatGPT asked for a few examples of books in this vein I’d enjoyed. Then for the first time it spotlighted aspects of my taste and asked about them –
ChatGPT: It’s interesting how many of these books seem to have an outsider’s perspective, whether through exile (The Illiterate Trilogy), isolation (The Icelandic Fisherman), or a kind of observational detachment (Nature Stories). Do you find yourself drawn to books with that kind of tone — works that explore alienation, the natural world, or an oblique, lyrical view of reality? Or is there another common thread that pulls them together for you?
- There was a good amount of back and forth on this: ChatGPT spotting patterns and asking questions, me offering other aspects of my taste and mentioning books I like that fitted into its evolving understanding of what I enjoy.
- Then I asked ChatGPT for wildcard recommendations “that don’t tie directly into themes and styles I’ve mentioned,” along with “a couple to go into completely blind, knowing nothing but the title and the author.”
By this point the list of titles was pretty extensive and almost all of them were intriguing in some way. The first batch had arrived in the post, all sorts of ideas were popping into my head about directions the project could take, and I was feeling just as inspired as I had by previous reading projects.
Hence this post.
And there’s more to come, so watch this space!