Asking AI What I Should Read Next...

Posted by on
At this point, what can be said about AI and literature that hasn't been said before? For all of its aspirations to the I, it's still very much the A, and that means it cannot capture the essence of human imagination. Its job is to regurgitate what it's been fed, to emulate an amalgamation of the voices and styles it's consumed. 

While I'd never trust it to create (I have no interest in reading an AI-written book), what it can do (and do very well) is collate, compare, and filter data to make evaluations. And that's where I got curious and decided to give it a chance to help me plan my next read.

Prompting AI For Reading Recommendations

I queried the major AI models (ChatGPT, Claude, Copilot, Gemini, Google AI, Perplexity) with a simple prompt. I've removed my details in case you'd like to try it yourself.
I am [age], [race], [marital status], [gender], and [sexuality]. I love well-told stories with action, adventure, drama, romance, and imagination. Growing up, [Author Names] were my strongest horror influences. My fantasy influences were [Author Names]. I like my horror dark, violent, and strongly supernatural. I do not like religious themes such as angels and exorcisms. I like my fantasy epic and hopeful, with magic, warriors, elves, dwarves, dragons, and chosen one or quest themes. I don't like pessimistic grimdark themes. Suggest to me 5 horror novels and 5 fantasy novels I should read today, as a mature reader, that will captivate me and entertain me like the books of my teenage years.
Much to my surprise, while Claude and Copilot best fit my professional needs at the office, it was Gemini and Perplexity that stood out with recommendations that truly felt like they understood the query. I found myself reading over the 'why' of the recommendations, nodding my head, and making a shopping list.

AI's Horror Recommendations

Interestingly enough, horror is where the models were most aligned:

  • The Fisherman by John Langan was recommended by 5/6 models
  • The Ritual by Adam Nevill was recommended by 4/6 models
  • The Only Good Indians by Stephen Graham Jones was recommended by 3/6 models
  • Incidents Around the House by Josh Malerman was recommended by 2/6 models


Some of the other standout recommendations (all from Gemini & Perplexity) included:
  • Black Mouth by Ronald Malfi — Pure Stephen King energy
  • Sister, Maiden, Monster by Lucy A. Snyder — For fans of Brian Lumley and extreme horror
  • The Spite House by Johnny Compton — Barker/Koontz-style dread with a strong hook
  • The Hunger by Alma Katsu — Plague-like historical horror with a grim, supernatural edge 
  • The Watchers by A.M. Shine — Excellent for late-night “one more chapter” compulsion
  • The Lesser Dead by Christopher Buehlman — Feral, dark, violent vampire with momentum and bite

AI's Fantasy Recommendations

Maybe it's the size of the genre, or maybe it's the mix of authors I provided, but there was far less overlap/alignment with fantasy:
  • Theft of Swords by Michael J. Sullivan was recommended by 4/6 models
  • Of Blood and Fire by Ryan Cahill was recommended by 2/6 models
  • The Ember Blade by Chris Wooding was recommended by 2/6 models

Standalone standout recommendations (mostly from Gemini & Perplexity) included:
  • Dragon Mage by M.L. Spencer — A direct love letter to the 90s era of Weis & Hickman
  • The Obsidian Tower by Melissa Caruso — Very Tad Williams-adjacent in tone
  • The Elven by Bernhard Hennen — Old-school high-fantasy immersion 
  • The Stone in the Skull by Elizabeth Bear — Feels like a lost classic from the 90s—but smarter and more inclusive
  • Malice by John Gwynne — The warrior-driven, sword-and-glory spirit of Dragonlance and Shannara 

New Connections To Go With Those New Books

While many of the citations within the AI models were from Amazon, Goodreads, or regional newspapers, I did find some new authors and bloggers to follow, including Jack Curran, The Reading Stray, My Geekology, The Fantasy Review, Aaron D. Schneider, and N.S. Mirage, 

Unfortunately, as solid as its recommendations were, a total lack of citations was a definite weak spot for Gemini.

Pushing AI For More (and Failing)

Since we do a lot of prompt engineering at the office, learning how to get the most out of AI, I took things a step further. I followed up by querying those same AI models with a more detailed version of my prompt (using new chats/threads so that there'd be no carryover), but that just seemed to confuse them. 

They seemed to get lost between what I like and don't like, with recommendations that were so far off the mark as to be insulting. I may try again with a different prompt at another time, but I'm happy with what I have.

Comments

Post a Comment