A ten-year-old plants herself at your desk: “I finished all the Wings of Fire books. What’s next?” You know this dance. So does the dad behind her who “doesn’t really read but wants something for the beach,” and the book-club organizer who needs six titles by Thursday. Readers’ advisory is the most human thing a librarian does — and on a busy desk, it’s also the thing you never have quite enough minutes for.
Here’s where ChatGPT genuinely helps: not by replacing your judgment, but by giving you a fast first draft of read-alikes to work from. The catch — and it’s a big one we’ll come back to — is that ChatGPT will confidently invent books that don’t exist. Which, honestly, makes you the perfect person to use it. Verifying a title before it reaches a patron is already your reflex. You’d never make up a book on the spot; your job is to make sure the machine didn’t either.
What “AI for readers’ advisory” actually means
You’re not handing the recommendation to a robot. You’re using ChatGPT the way you’d use a sharp library-school intern: “give me ten candidates and your reasoning, fast,” and then you do what you’ve always done — filter for your collection, your community, and the actual kid in front of you. The whole movement here, which the American Library Association now frames as “AI literacy is the new information literacy,” puts librarians at the center precisely because the tool needs a skeptical expert holding the reins.
The trick to good prompts is feeding the AI what you know: the reader’s age, what they loved and why, your reading level, and any content lines you need to respect. Generic prompt in, generic list out. Specific prompt in, a usable shortlist out.
The 7 prompts
Copy these, swap in the specifics, and treat every result as a draft to vet — never a finished list.
1. The read-alike (the bread and butter).
“A 4th grader loved the Wings of Fire series and the Percy Jackson books. Give me 8 read-alikes at roughly a 3rd–5th grade reading level, each with a one-line hook explaining why it fits. Prefer well-known, widely-stocked titles.”
2. The reluctant reader.
“Recommend 6 books for a 12-year-old who says they ‘hate reading’ but loves Minecraft, soccer, and funny YouTubers. Short, high-interest, fast-paced, lots of white space or illustrations. One line on why each might hook them.”
3. The “I finished the whole series, now what.”
“A teen just finished all of Sarah J. Maas. Give me a ‘if you loved this, try this’ ladder of 6 titles — some very similar, some a gentle step in a new direction — with a one-line reason for each.”
4. The adult who “doesn’t read much.”
“Suggest 5 books for an adult who hasn’t read for pleasure since school but enjoys true-crime podcasts and thrillers on TV. Accessible, gripping, not intimidating. One line each on the hook.”
5. The book-club shortlist.
“Build a 6-title shortlist for an adult book club that liked ‘Lessons in Chemistry’ and wants discussable fiction under 400 pages. For each: a one-sentence pitch and one discussion question.”
6. The themed display bundle.
“I’m building a ‘cozy autumn reads’ display for a public library. Give me 12 candidate titles across adult and YA, mixing well-known and backlist, with a 5-word tag for each so I can write shelf-talkers.”
7. The “we probably own these” filter.
“From this list of recommendations, flag which are most likely to be in a mid-sized U.S. public library’s standard collection, and which are niche or out of print.”
For every one of these, the next step is identical: confirm the titles are real and check your catalog. Paste a suspicious result back and ask “are all of these real, currently-published books? Flag anything you’re not certain exists.” — but never skip the human check against your own system. That five-second verify is the difference between a great recommendation and an embarrassing one.
What this means for you
If you’re a public librarian: this shines at the desk during a rush and for building displays on your lunch break. Keep a saved prompt or two in a note on your phone, feed it the reader in front of you, and let it surface candidates while you do the judging.
If you’re a school librarian: the reading-level filter is your friend — ask for Lexile or grade-band ranges, and lean hard on the “verify it’s real and we own it” step before you point a student at a title. You’re also modeling the exact skepticism you want them to learn.
If you’re a library assistant or paraprofessional: this lowers the barrier to confident advisory. You don’t need to have read everything in the building to offer a thoughtful starting list — you need to know how to ask well and check carefully.
If you run a solo branch: the time savings compound. The same prompt skeleton produces beach reads on Monday and a Día de los Muertos display on Wednesday.
What this can’t do
- It invents books. ChatGPT will hand you a plausible title by a real author that simply does not exist. This is the single most important thing to internalize: verify every title before it reaches a patron. Your professional reflex is the safeguard the tool doesn’t have.
- It doesn’t know your collection. It can guess what a “standard” library owns, but it has no idea what’s actually on your shelves or in your consortium. The catalog check is non-negotiable.
- Never paste patron data. A reader’s name, card number, or borrowing history must never go into a public AI tool. Describe the reader in general terms (“a 9-year-old who loves dragons”), never as an identifiable person.
- It can miss the room. It doesn’t know your community’s sensitivities, your displays from last month, or the local author everyone’s asking about. That context is yours.
- It’s not a collection-development tool. It can brainstorm, but purchase decisions still need real reviews, real budgets, and your professional judgment.
The bottom line
ChatGPT won’t replace the librarian who knows that the quiet kid asking for “something scary but not too scary” really wants Coraline. What it does is take the blank-page friction out of the first ten candidates, so you spend your minutes on the part only you can do: matching a real reader to a real book, and catching the one the machine made up. Used with your verify reflex intact, it’s a genuinely useful desk companion.
Want the full toolkit — readers’ advisory, program planning, and teaching patrons to use AI well — in under an hour? Our AI for Librarians course walks through it with copy-paste prompts built for the everyday public and school librarian. First two lessons are free.
