I read 100 books a year —and Meta AI struggled to come up with new recommendations

Growing up shy, with my older sibling always choosing what we watched on TV, I naturally fell in love with reading. The world within books was far more captivating than eating lunch alone or watching another episode of "The Wonder Years." Without a doubt, science fiction books like Arthur C. Clarke’s 2001: A Space Odyssey sparked my interest in AI, but to this day, I’ll read everything from Shakespeare to the back of the cereal box with enthusiasm. I often read more than one book at a time, binging characters and storylines the way most of my friends stream their favorite shows.

Besides having stacks of books on my nightstand, I also have a large collection of audible books in my library that I listen to on long runs and in the car. When you’re a book lover like me, suggestions can come from anywhere. Amazon to the local librarian will make suggestions, but in most cases, I’ve already read them. This time, I decided to put AI to the test.

I wondered if I would stump it with the hundreds of books I’ve read, or if it would actually prove useful in choosing my next read. Here’s what happened when I gave Meta AI a list of the last five books I read and asked it to choose my next book.

Results of round one

I used the prompt: “Based on the last 5 books I read, please recommend 5 more books that you think I would find interesting: 1. Funny Story by Emily Henry 2. The Creative Act by Rick Rubin 3. Roctogenerians by Mo Rocca 4. Eat and Get Gas by J.A. Wright 5. Station Eleven by Emily St. John Mandel”

I purposely picked an eclectic collection of books that I thoroughly enjoyed. They are a mix of romantic comedy, historical fiction, science fiction, self-help, and inspiring essays. It made me smile when the AI complimented me on the selection before listing books that it thought might pique my interest.

The good news is, Meta AI has great recommendations. One of my favorite features about this model is the personal delivery of the responses. It’s as though a friend is texting me their favorite books. The mix it gave from my prompt was of new-ish and older publications. The bad news is, I have already read all of the suggestions at least once. I couldn’t wait to tell the AI to try again.

Results of round two

I used the prompt: “Thank you for those suggestions, but I have read all of those. I liked them all very much with The Storied Life of A.J. Fikry being one of my favorites. Can you please recommend five more related to the books I originally listed?”

The AI acknowledged that I liked the suggestions but didn’t seem to care that I’ve already read them. In the next response, it really leaned in to my comment about enjoying one of the books on its first list of suggestions. In fact, it decided to run with more suggestions based on that book, rather than with the list of books I mentioned in the beginning.

When the next list of suggestions populated, I sighed loud enough for my cat to wake from his nap and stare at me. The problem? I’ve read all of these books and the bonus book. Just Kids by Patti Smith is fantastic. I debated whether or not to tell Meta AI that I really enjoyed that book as I did not want it to stray too far from my original list.

Results of round three

Have I mentioned how much I enjoy Meta AI's constant flattery? After telling me that I was a “voracious reader with a great taste in books,” it gave me a new list of book suggestions. Again, all of the books were ones that appealed to me, but this time I had read all but the fifth one on the list.

The AI seemed to aggressively suggest books based on “identity, creativity, and human connection,” all of which are great topics, but I don’t need to read about them all the time. It’s why my original list was so diversified. Essentially, the AI seemed to be taking my comments on its suggestions and using them to come up with more suggestions. That’s why for the next round I decided to not mention any of the books but to simply ask for another list based on my original prompt.

Results of round four

I used the prompt: “Please recommend 5 more books based on my original list.”

I kept it short and sweet and did not identify that I had read all but the last book in the last round. The next list of recommendations had a combination of books I have read, books I have not, and books that do not interest me. So at this point, the AI has not given me a new list of 5 books to read. I only have one book from the last round and two from this round. I need two more books.

Results of round five

I used the prompt: “Nice recommendations. I am looking for a few more suggestions based on the authors and storylines in my original list.”

Is it just me or is Meta AI ready to throw in the towel? It decided to list two books related directly to each book included in my first prompt. Yet, many of these books were already in previous rounds of recommendations. What’s also curious is that it has made suggestions that would probably seem obvious to a human. For example, it suggested the companion book to Big Magic titled Eat, Pray, Love Made Me Do It. I have read it. There are two books in this list that I haven’t read and would probably find interesting. Evvie Drake Starts Over by Linda Holmes and The Golem and the Jinni by Helene Wecker, which I will add to my list.

Final thoughts

I definitely thought I was going to stump Meta AI, but it came through in the end by giving me what I asked for - a list of five books that pique my interest. I’m a little weary and am left wondering how they relate to my first list of books. Will they be too similar? I’m ready to dive in and see. Overall, I would ask Meta AI for advice again on literature. Who knows? Maybe we could even start a book club together.

More from Tom's Guide

• How to use ChatGPT to help write your resume in 9 easy steps
• How to get started with ChatGPT free model
• ChatGPT can now access the whole internet — this is big