Hi there,

Welcome to a new edition of Reedsy’s marketing newsletter! I’m freshly back from the Novelists, Inc. conference in Florida, where I was invited to give a workshop on one of my favorite topics right now: AI search optimization for authors.

If you’ve been following this newsletter for the past few months, you’ll know I’ve written quite a bit about this already. Over a series of four newsletters, I covered:

• Why AI search is the future of how readers will discover books online (read here).
• How AI search engines work in general:
• They turn one prompt into dozens of different searches — a process called “query fan-out” (read here);
• They gather a corpus of documents, which they analyze, score, and rank based on their similarity/relevance to the initial prompt (read here);
• They personalize results based on available user information, meaning no one gets the exact same results (read here).

If you haven’t had a chance to read through these, I highly encourage you to.

The upcoming issues of this newsletter will keep focusing on AI search, but with a more practical angle. They’ll aim to answer a simple question:

What do I need to do to make sure my books get recommended by ChatGPT?

Why specifically ChatGPT? First, it’s by far the largest AI chatbot in terms of user adoption (over 60% of the AI search market share). Based on my experiments, it’s also the most sophisticated when it comes to providing accurate book recommendations.

But more importantly, all AI search engines follow the same principles — so if you learn how to optimize for ChatGPT, you’ll (mostly) know how to optimize for Gemini, Claude, Perplexity, Grok, etc.

Now, one of my favorite things about AI search engines is that if you want to understand how they work, you can simply ask them. Of course, you can’t trust their answers 100% (LLMs are prone to fabrications, and are explicitly instructed not to reveal their system prompts). But this can still give you a good bird’s-eye view of how they operate behind the scenes.

For example, I asked ChatGPT how it would generate an answer if a reader prompted it for specific book recommendations. You can read the transcript here — it’s eye-opening.

Let’s break down the process:

1. Parsing the prompt: ChatGPT first seeks to understand the request, breaking it into “hard” and “soft” constraints. In my example, the genre, tropes, character types, and heat level I specified are all treated as hard constraints — meaning every recommendation must match them.
2. Query fan-out: ChatGPT then kicks off a batch of internet searches, some broad (e.g., "content warnings" fae romantasy), others targeted (e.g., site:reddit.com/r/Romantasy "slow burn" "enemies to lovers"). This mirrors Google’s query fan-out process.
3. Candidate selection: For each search, ChatGPT scans the top results and compiles a list of book candidates, prioritizing those mentioned repeatedly or by authoritative sites. (See the “Sources I’m likely to consult” section of its answer for examples.)
4. Verification: ChatGPT then runs follow-up searches to confirm that candidates match all hard constraints, discarding books that don’t (e.g., if explicit sex scenes are confirmed, the title is eliminated).
5. Scoring and ranking: Remaining books are scored based on how well they fit the prompt and the reliability of the data. A book flagged as a good match but with uncertain details (e.g. ChatGPT is not sure about the “no explicit sex scenes” bit) will score lower than one where ChatGPT was able to fully verify it meets all the hard constraints.
6. Final output: ChatGPT presents the top 3–5 recommendations, usually alongside a summary table.

So, what does this mean for authors who want their books to appear in those final recommendations?

Here are my two key takeaways:

1. Visibility in initial searches: Your books need to surface in the fan-out searches ChatGPT runs. The more it is mentioned on the sources it trusts, the higher your chances of it being scored and ranked high.
2. Clarity of metadata: Make it as easy as possible for LLMs to understand what your book is about — and whether it meets (or breaks) the kinds of constraints readers might specify.

Over the next few weeks, I’ll dive deeper into how to do this, sharing practical examples, case studies, and some hacks I’ve been experimenting with.

Until then, happy writing, and happy marketing!

Ricardo

Copyright © 2026 Reedsy, All rights reserved.