Can a school librarian trust AI's reading list?

A plain-language guide for K–12 school librarians who want to use generative AI for finding resources and helping teachers and students, without losing the verification habit or the privacy line.

What generative AI is actually doing when it answers you, and where does it go wrong?

If you have not opened a generative AI tool for work yet, here is the one paragraph that will hold for the rest of this guide. A generative AI chatbot is not a search engine. It does not go look up known pages and hand them back. It reads your question, then writes back the next plausible words, one at a time, the way a very confident person who has skimmed a million library shelves might write. That is why it can sound right while being wrong. It can invent a book title, an author, a publisher, or a URL that looks exactly right and does not exist anywhere.

So when a 5th-grade teacher emails you and asks for eight books on an immigration unit, and you open ChatGPT or Claude or Copilot or Gemini, the tool is not going to a database. It is predicting what an answer to your question would look like, and the answer will read beautifully. The same capability that helps you find a real reading list is the capability that will hand you a fake one. Researchers studying AI in education have named this the source discovery–fabrication paradox: the same capability that supports source discovery can produce fabricated information.Hossain et al. 2026

You do not need to remember that phrase. You do need to remember the habit that comes from it: every AI-suggested title, author, URL, or reading level is a claim to be checked, not an answer to be trusted. Treat it the way you would treat a sixth grader who really, really wants you to believe a book exists. You would open the catalog.

AI literacy is a small bundle of habits, not one skill

Before any tool lands on your desk, it helps to know what you are trying to become fluent in. Researchers who have spent years pulling apart what "AI literacy" actually means for teachers and librarians describe it as six habits that travel together:Hossain et al. 2026

图注：The six habits that make up AI literacy for educators, arranged around a central idea. None of them stands alone. Skip one and the rest start to wobble.

A quick everyday version of each, so you can borrow the vocabulary without needing to read the paper:

• Technical fluency. Knowing, at an everyday level, what the tool is doing when it answers you. It is predicting plausible text, not retrieving known pages. That is enough.
• Critical evaluation. Triangulating what the tool says against your catalog, your leveling system, your own knowledge. Watching for fabricated titles, authors, URLs.
• Human-AI collaboration. Keeping yourself as the decision-maker. AI is a co-pilot, not a co-author. You sign off, or it does not go out.
• Contextual awareness. Recognizing that the same tool produces different consequences in different schools, different languages, and different learners. Equity is built in, not bolted on.
• Ethical reasoning. Handling bias, privacy, authorship, and academic integrity as connected commitments you model for students.
• Empowered agency. Knowing that declining a tool is itself a literacy outcome. Saying "I am not going to use this for this task" is a professional answer, not a failure.

If you take nothing else from this chapter, take the verification habit. The rest of the tutorial builds on it. The next six chapters will give you tools, prompts, workflows, and a privacy line. None of them replaces the habit of opening the catalog before you hand anything to a teacher or a student.

Which AI tool is the right one for the job at the reference desk?

You will hear a lot of brand names. ChatGPT. Claude. Copilot. Gemini. Perplexity. Elicit. Consensus. MagicSchool (which has its own in-platform chatbot, Raina). SchoolAI. The list changes every few months and that is fine. What does not change is that there are four families of tool, and each one is good at one job and bad at another.

The international vocabulary that educators have settled on for what you do with these tools breaks into four domains:OECD/European Commission 2026

图注：Four domains for working with AI in K–12 settings: Engage (recognize AI, evaluate what it says), Create (use it to draft and explore), Manage (decide when to use it and which tool fits), and Shape (influence how AI systems are built and used in your school). Each domain develops across Basic, Intermediate, and Advanced levels — not tied to specific grades.

You can borrow this vocabulary in parent emails and lesson plans without ever needing to cite the framework. Engage AI is what you do when you ask and check. Create AI is what you do when you draft and revise. Manage AI is what you do when you pick the right tool for the job. Shape AI is what you do when you decide a tool does not belong in your library at all.

Verification starts here, not later

Most librarians who start using AI at the reference desk report the same pattern: the first few answers feel magical, and the habit of checking forms after the magic fades, when something embarrassing has already gone out. Reverse that. From the first prompt, build the habit of asking: does this title exist in my catalog? Does the author match? Is this URL alive? That habit is the difference between a librarian using AI and a librarian being used by AI.

By the end of this chapter you should have one tool you reach for as an "AI search helper" (Perplexity is the easiest one to point new librarians at, because the citations are right there in the answer) and one chat assistant you are comfortable enough with to draft and revise. The exact brand matters less than the habit: every answer gets a verification pass before it leaves your desk.

How do I turn a teacher's or student's request into a usable reading list?

This is the core of your day: a teacher walks up (or emails) and says, "I need eight books for a 5th-grade immigration unit, no more than 200 pages each, ideally with a protagonist who recently arrived in this country." Or a 7th grader slumps at the reference desk and says, "I want something like Percy Jackson but harder." You know what to do next. You do the reference interview. You ask a few quiet questions, you narrow the request, you pull from the catalog. The only thing that changes when an AI is at the desk is that you can do the "brainstorm a shortlist" step in fifteen seconds instead of fifteen minutes, as long as you verify the shortlist before you print it.

The five-step pipeline

图注：The five-step pipeline from teacher or student request to verified reading list. The middle three steps are where AI does its real work; the first and last are where you do yours.

Each step has a job and a check. The reference interview is yours, full stop. The prompt turns the interview into a question the AI can answer. The verification step is yours again. The catalog confirmation is where the AI's shortlist meets your real shelves. The reader-facing output (a printed list, a display, an email) goes out only after steps three and four are done.

The prompt itself: Role + Task + Context + Constraint

The pattern school librarians have settled on has four parts. TCEA TechNotes (the Texas Computer Education Association's K–12 edtech blog) frames it as the "Role + Task + Context + Constraint" pattern, and once you see it, you will see it in every good prompt library for school work.Qualls 2025

• Role. Who the AI is. "You are a school librarian responsible for…" or "You are a school librarian working with a 5th-grade social studies teacher on an immigration unit."
• Task. The verb and the deliverable. "Create a shortlist of eight books."
• Context. The bracket placeholders you swap in. [grade level], [topic], [catalog system], [reading level band].
• Constraint. The quality bar. "Each title must actually exist; verify against our Destiny catalog. No more than 200 pages. Protagonists who recently arrived in this country when possible."

Here is the kind of paste-in prompt this produces for the immigration unit request:

You are a school librarian responsible for supporting a 5th-grade social studies unit on immigration. The teacher has asked for a shortlist of eight books, each 200 pages or fewer, ideally with a protagonist who recently arrived in the United States. Create a shortlist that spans picture books, middle-grade novels, and memoir, and for each title include the author, page count, and a two-sentence student-friendly hook. Each title must actually exist; if you are not certain a title is real, mark it "to verify" rather than presenting it as confirmed.

Run that through any chat assistant. You will get a draft. The draft will sound great. Some of it will be real and some of it will be invented. Your job in the verification step is to find out which is which before the list goes to the teacher.

The AASL practice guide on your shelf

If you want a single book to keep next to your keyboard, the American Association of School Librarians (a division of the American Library Association) announced one in late 2025 (formal first printing 2026): Leveraging AI in School Libraries: From Basics to Best Practices, by Delandrus Seales.Seales 2025

图注：The AASL/ALA Editions practice guide for K–12 school librarians. Identification anchor; it is the only AASL-published AI practice book for school librarians as of this writing.

What it gives you is a K–12-specific complement to the generic Role + Task + Context + Constraint pattern: a framework the book calls CLEAR. Each letter is one of the questions you would already ask a teacher or student. C is clear language (what words does the AI need to use and avoid). L is length and complexity (how long, how dense). E is explaining or examples (do you want the AI to show its work). A is audience (who is this for). R is relevance (how tightly it must match the request). When in doubt, layer CLEAR on top of your Role + Task + Context + Constraint prompt. The two patterns together produce drafts that need less rewriting on the back end.

Matching reading levels before the list goes out

The other quality-control step the AI cannot do alone is reading-level matching. A 5th grader's Lexile (a scale from BR — below reader — up to 1600L+, scored by sentence length and word frequency) is typically around 850L–950L, while a 4th grader's might be 700L–850L, but the same student's Guided Reading level (an A-to-Z scale that matches readers to small-group books) might be Q or R. Lexile measures the text, Guided Reading measures the reader-to-text match. Both can be right about the same student on the same day. Here is the crosswalk you will want open in a tab while you verify:advancebookreaders 2026

Treat this as a sanity check, not a precise translation. The same student's Lexile can be advanced (say, 800L vocabulary) while their Guided Reading tracks lower (P or Q). AR (Accelerated Reader) measures vocabulary and sentence structure; DRA (Developmental Reading Assessment) is a teacher sitting one-on-one with a child, listening to them read aloud and asking comprehension questions. Two systems can assign different levels to the same book. When the AI hands you a title, look up the Lexile (most public library catalogs show it), check it against the student's record, and if the two numbers disagree by more than a band, ask the teacher or parent to confirm. The 5-Finger Rule is your low-tech backstop: have the student read a full page aloud. Two or three unknown words is just right. Five means it is a read-aloud, not independent reading.

By the end of this chapter, the workflow should be: a teacher request enters as a reference interview, you turn it into a Role + Task + Context + Constraint prompt (with CLEAR layered in if you want), the AI gives you a draft, you check every title against your catalog and every reading level against the crosswalk, and only then does the list go out.

How do I help a teacher draft a lesson, a reading list, or a research guide faster?

Most teacher requests you field in a week are not actually asking you to do library science. They are asking you to do the time-consuming drafting work that lives between the curriculum and the kids. A 5th-grade team needs a research-skills mini-lesson for next Tuesday. A 7th-grade English teacher needs an email home about a new book club. A high school social studies teacher needs an inquiry-based research project on local water issues, with discussion questions and a presentation rubric. You can draft all three of those in a chat assistant, in a fraction of the time, as long as you keep the verification habit from the last chapter.

Three paste-in prompts to keep on hand

These are adapted from a working prompt library written for K–12 librarians by the Texas Computer Education Association.Qualls 2025

Prompt 1: Research-skills mini-lesson.

You are a school librarian responsible for teaching 5th graders how to spot reliable online sources. Create a 25-minute mini-lesson plan that uses a mix of one library database and one curated website, with one discussion question and one short hands-on activity. The lesson should fit inside a single library visit and leave room for students to practice the search themselves.

Prompt 2: Teacher collaboration email.

You are a school librarian working with a 7th-grade English teacher who is planning a three-week unit on dystopia. Write a short, friendly email I can send today proposing two specific ways the library can support the unit (one research-skills mini-lesson and one curated reading list with three to five entry points at different reading levels). Keep the tone collegial, not promotional.

Prompt 3: Inquiry-based research project.

You are a school librarian working with a high school social studies teacher on an inquiry-based research project about local water issues. Outline a two-week project for 11th graders that moves them from question-formation to evidence-collection to a five-minute presentation, with at least one step where students must find a source the AI did not surface, and one step where they document how they used AI.

These prompts are paste-in. Swap in the grade, the topic, the teacher name, and your catalog system. Run them, revise, send. The CLEAR framework (as summarized on the ALA Editions page and in the index entries) from the AASL book layers on top when the draft is too dense (clear language), too long (length/complexity), too vague (explaining/examples), off-audience (audience), or off-topic (relevance).Seales 2025

Tagging the lesson to the AASL Standards

One small payoff that compounds: tag every lesson prompt you draft to one or more of the AASL National School Library Standards' six Shared Foundations. Inquire, Include, Collaborate, Curate, Explore, Engage are the words your district already speaks when they evaluate library instruction. You can, for example, tag a research-skills mini-lesson to Inquire; a reading list with diverse protagonists to Include and Curate; a teacher-collab unit to Collaborate. You do not have to tag every prompt, but the habit means when you hand a principal or curriculum coordinator a folder of "AI-assisted lessons you drafted this term," they can see at a glance which standards each one touches.Seales 2025

The teacher-side anti-pattern to avoid

Two shortcuts that look like time-savers and turn into headaches. The first is letting the AI write the parent email without you reading it. Parent emails carry your voice, your relationship, and your professional judgment. Draft with AI; rewrite in your own voice before you send. The second is pasting a teacher's specific request into a consumer AI tool without scrubbing student names or identifiers. That habit belongs in the privacy chapter, but it is worth naming here: every time you copy a real student artifact into a chat assistant, you are making a decision your district's privacy policy has an opinion about. Chapter 7 is where you learn where that line is. For now, the rule is: do not paste a student name, a student writing sample, a class roster, or any piece of information that could pick one student out of a classroom, into any AI tool you have not vetted with your district.

By the end of this chapter, the workflow for a teacher request should feel like a familiar three-step: paste-in prompt with Role + Task + Context + Constraint (and CLEAR where needed), revise in your own voice, verify the output against your knowledge of the teacher and the curriculum before you send.

How do I help a student use AI safely and well at homework time?

The hardest part of your job is not finding resources. It is teaching the students in front of you to think about what they find. AI did not invent that problem, but it did make it more urgent. A student who would never copy a paragraph from Wikipedia without citation can accidentally paste a chat assistant's hallucinated answer into a paper and turn it in, never realizing they are submitting something invented. The durable habit you can give them is the same one you formed in chapter 1: every AI answer is a claim to be checked, not an answer to be trusted.

The six roles you are preparing students for

The international and US education organizations have converged on a vocabulary for what AI-using students should grow into. The most student-friendly version comes from ISTE+ASCD, the merged International Society for Technology in Education and Association for Supervision and Curriculum Development.ISTE+ASCD 2026

图注：Six roles of an AI-Ready Graduate, as a flat taxonomy. None is more foundational than another. The roles travel together.

A one-sentence version of each, in language you can borrow for a lesson plan or a parent email:

• Learner. Uses AI to set learning goals, get unstuck, and ask for specific feedback on their work.
• Researcher. Uses AI as a research assistant, then checks where experts agree or disagree.
• Synthesizer. Pulls information from many sources (including "releveling" content for a different reading level) into a single product.
• Problem Solver. Brainstorms with AI to clarify a challenge and probe the best options, while keeping judgment about which to accept.
• Connector. Uses AI to lower friction in collaboration across languages, cultures, and viewpoints.
• Storyteller. Picks the right format (slideshow, video, podcast, infographic) and uses AI to sharpen the message.

The role that anchors your library instruction is Researcher. That is the role that says, in plain words, "use AI to find starting material, then cross-check at least one source before you trust it."

The single durable habit: cross-check one source

You cannot teach a student to fact-check everything. You can teach them to cross-check one source. The instruction fits on an index card:

When the AI gives you an answer, pick one specific claim from it. Find one real source (a book in our catalog, a database we subscribe to, a website on the approved list) that confirms or contradicts it. Write one sentence about what changed.

That is the whole habit. It works for a 6th grader writing a one-page paper. It works for a 12th grader writing a capstone. The point is not that one source is enough to verify anything; the point is that the student forms the habit of opening a second source at all. AI is the start of a research trail, not the end.

If you want a longer-term practice to give students, the AASL practice guide suggests an AI Research Journal: a template where the student records every prompt they typed, the answer they got, what they kept and what they discarded, and how they cited the AI in their final product.Seales 2025 When a student has a paper trail, "did you cheat?" stops being the question. The question becomes "show me how you used it."

Reading-level matching as a librarian-only check

The crosswalk from chapter 3 is your tool here, not theirs. When a student asks the AI for "a book like Percy Jackson but harder," and the AI hands back a title, you (not the student) check the Lexile, Guided Reading, AR, or DRA band against the student's record. If the two systems disagree by more than one band, ask the teacher or parent before the book goes home with the student. The student can use the AI to find candidates; you confirm the match.

By the end of this chapter, your role with students has not changed in shape, only in tempo. You still teach information literacy. You still match books to readers. You just do it in a world where every answer a student gets from a chatbot is a claim waiting to be checked. The Researcher role and the cross-check-one-source habit are how you teach that.

Three exercises you can actually run in your library this week

Reading is one thing. Doing is another. Here are three exercises that fit inside a single school morning, each under twenty-five minutes, each designed to land a habit before the chapter wears off. You will need a chat assistant (any of them will do) and an AI search helper with clickable citations (Perplexity is the easiest to point new librarians at). Everything else you already have.

Exercise 1 (15 minutes): click every citation

Pick one teacher reference question you would have answered yesterday from memory. Open an AI search helper. Paste the question in plain language, no special formatting. Read the answer. Then click every citation the tool gives you. The point of this exercise is not to learn the answer (you already know it). The point is to see how citations behave. Some will resolve cleanly to a real page. Some will resolve to a real page that does not actually say what the AI claimed. Some will resolve to a 404 (page not found) or to a domain that no longer exists. The point is to feel, with your own hands, what "the citation is right there but quietly wrong" looks like. That feeling is what you will remember when the AI hands you a reading list next month.

Exercise 2 (20 minutes): draft and verify a reading list

Take a real teacher request. A 5th-grade immigration unit. A 7th-grade dystopia unit. A high-school research-skills intro. Open a chat assistant. Paste this prompt (or adapt it to your request):

You are a school librarian responsible for supporting a [grade level] unit on [topic]. The teacher has asked for a shortlist of [number] books, each [page count] or fewer, ideally with [specific representation]. Create a shortlist that spans [formats], and for each title include the author, page count, and a two-sentence student-friendly hook. Each title must actually exist; if you are not certain a title is real, mark it "to verify" rather than presenting it as confirmed.

Read the draft. Mark every title "to verify" that you do not personally recognize. Open your catalog. Confirm or reject every title. Note the ones that did not exist. Note the ones that did exist but were wildly off-topic. You have just done one round of the chapter 3 workflow. Time how long it took. Compare that to how long the same shortlist would have taken you from memory.

Exercise 3 (20 minutes): teach one student the cross-check habit

Find one middle or high school student who is doing a research project this week. Sit with them for twenty minutes. Have them paste an AI-generated answer to a question they are researching into a document. Then walk them through the cross-check-one-source habit from chapter 5: pick one specific claim from the AI's answer. Find one real source (a book in the catalog, a database the school subscribes to, an approved website) that confirms or contradicts it. Write one sentence about what changed. Do not turn this into a graded exercise. Frame it as a tool the student can use on every assignment from now on. If you can, do it with a student who is naturally skeptical. Then do it with a student who is naturally credulous. You will need to teach it differently the second time.

By the end of a week, you have done all three. The verification habit is no longer something you read about. It is something you have done, with your own hands, three different ways.

Where is the line I should not cross, and what counts as crossing it?

You have spent six chapters building up positive uses. The librarian who sits down with ChatGPT on Sunday night to draft a 5th-grade immigration reading list, who verifies every title against the catalog, who teaches the cross-check-one-source habit to a 7th grader on Tuesday, who uses a vetted contracted tool for any student-named request, is doing the job you want to be doing. This chapter is about the line that protects that work. The line is not invented by your district; it is written into two federal laws you have been hearing about for years.

The profession-level values you are standing on

In July 2026 the American Library Association Council adopted a formal guidance document on AI in libraries.ALA Council 2026 It organizes its advice around six values: Public Good, Intellectual Freedom, Privacy, Sustainability, Diversity/Equity/Inclusion/Access (DEIA), and Labor. The single most useful sentence in the document, for a school librarian staring at a new AI tool and wondering whether to adopt it, is this one: "A considered decision to delay, omit, remove, or discontinue use of a product is an affirmative professional choice grounded in library values, not a failure of adoption." Principled non-use is a professional answer. It is the answer you give when the tool does not fit the task, when the privacy math does not work, when you do not have time to verify the output. It is not a confession of failure.

The guidance also names, in plain words, what you may not paste into an AI tool unless the tool has been formally approved: "personally identifiable information; sensitive personal information; borrowing histories; reading histories; reference interactions; account information; staff records; prompts containing identifiable details; or other non-public records." That is the red-line list. Print it. Tape it next to your monitor.

What counts as personally identifiable information in your day

The 2025 update to the federal Children's Online Privacy Protection Act took effect on June 23, 2025, with most requirements in force by April 22, 2026.Belgrade School District 2025 The update added government-issued identifiers and biometric identifiers (fingerprints, voiceprints, facial recognition data) to the list of what counts as personally identifiable information. For a librarian, the practical PII inventory looks like this:

• A student's first and last name, or first name plus last initial
• Grade level combined with a school name (when it identifies the student)
• Student ID numbers
• The content of student inputs (writing samples, answers, voice transcripts)
• Performance data (scores, progress, completion rates) tied to an identifier
• Usage data (time on tool, features accessed, IP address, cookies, device IDs)
• Biometric identifiers (added by the 2025 rule)
• Government-issued identifiers (state IDs, birth certificates, passport numbers, SSN)
• Photos or videos that could identify a student
• Class rosters, even partial; reading lists with student names attached
• Parent contact information tied to a specific student

The practical rule is the one Belgrade's guidebook writes in plain language: if a piece of information could be combined with another piece to identify one specific student in your school, treat it as PII. The test is: would a reasonable person in this district be able to pick this student out of a classroom of 400? If yes, it is PII.

FERPA's four conditions, in plain English

FERPA (the Family Educational Rights and Privacy Act, the federal student-records privacy law) has a "school official exception" that lets schools share student education records with vendors without per-record parental consent, but only if all four conditions are true:Belgrade School District 2025

1. The vendor performs an institutional service or function for which the school would otherwise use employees.
2. The vendor is under the direct control of the school or district with respect to the use and maintenance of education records.
3. The vendor is subject to the rules in §99.33(a) governing the use and re-disclosure of personally identifiable information from education records.
4. The vendor meets the criteria the school has published in its annual FERPA notification for being a school official with a legitimate educational interest.

A consumer ChatGPT / Claude / Copilot / Gemini account that you log into with a personal email satisfies none of these conditions. The vendor is not performing a service the school has contracted for. The vendor is not under your district's direct control. Your district cannot turn the tool off, audit it, or require deletion on demand. That means: a consumer AI account is not a school official under FERPA, and student PII (the list above) does not flow through it without written parental consent. The librarian's practical rule of thumb, borrowed from district AI policies that have already worked through this math:TeachAI 2025

Aggregate, de-identified information goes into any tool. Student-named information goes only into a vetted, contracted tool that has cleared your district's vendor review.

The school-level checklist to bring to your principal

TeachAI, the coalition that includes ISTE, Khan Academy, Code.org, and the World Economic Forum, publishes a free AI Guidance for Schools Toolkit organized around seven principles: Purpose, Compliance, Knowledge, Balance, Integrity, Agency, and Evaluation.TeachAI 2025

图注：The TeachAI seven principles for AI in K–12 settings. No arrows or required order; the framework is a checklist, not a sequence. The toolkit ships with a sample guidance document you can adapt for your school.

The toolkit ships with a sample guidance document you can copy and adapt. It also links to a live policy tracker that, as of this writing, lists AI guidance from 30 U.S. states plus Puerto Rico, plus country-level guidance from Australia, Canada, the European Union, Japan, New Zealand, South Korea, the United Kingdom, and Uruguay.TeachAI 2025 If you want to bring a one-page artifact to your next principal meeting, the TeachAI seven principles image is the artifact. The district policy tracker is the answer to "who else has done this."

A decision tree for "can I paste this into this tool?"

Here is the workflow that resolves the most common librarian question, distilled from the Belgrade district guidebook and the ALA guidance:Belgrade School District 2025

flowchart TD
    Q1["Does this data identify a specific student?"] -->|No| OK1["OK in any AI tool"]
    Q1 -->|Yes| Q2["Is the AI tool on the district's approved list?"]
    Q2 -->|No| STOP1["STOP. De-identify or use a different tool."]
    Q2 -->|Yes| Q3["Is the use covered by FERPA school-official exception, directory info, or written parental consent?"]
    Q3 -->|None apply| STOP2["STOP. Escalate to district IT or Privacy Officer."]
    Q3 -->|Yes| Q4["Has the parent signed a COPPA-compliant consent form if the student is under 13?"]
    Q4 -->|No and required| STOP3["STOP. Get consent first."]
    Q4 -->|Yes or not required| FLOW["Data flows. Log the disclosure. Respect retention."]

The first question is the one that rules out most paste-ins at your desk. If a piece of information cannot identify a specific student (even when combined with other fields), it is not PII and the rest of the tree does not apply. If it can identify a student, you move down the tree. If the tool is not on the district's approved list, you stop. If the use is not covered by FERPA's school-official exception, directory information, or written parental consent, you stop and escalate.

What the answer often is

Most librarian work does not require a vetted, contracted tool. Most librarian work is on the librarian's side of the desk: drafting reading lists, drafting emails, drafting lesson plans, brainstorming parent communications. None of that requires student-named data. The verification habit from chapter 1 covers the librarian's workflow. The privacy chapter covers the rare cases where a student-named artifact has to enter an AI tool, and in those cases the tool must be vetted, contracted, and on the approved list.

If your district does not have an approved list yet, you can still do your work. The Librarian's Guide to ChatGPT's Book Scout post is explicit that the prompts it publishes are for behind-the-scenes librarian use only, not for direct use with students, and the privacy caveat is repeated three times in the post.Hatley 2025 That posture is the safe default until your district's approved list is in place.

References

The reading list below is organized by the role each source played in this tutorial.

Conceptual backbone (foundations):

1. Hossain, S., Ahmadi, S., Li, L., Awoyemi, I. D., Huang, W., Zhou, C., Li, C., Haniya, S., Khanam, S., & Subaha, T. M. (2026). Rethinking Generative AI Literacy: An Integrative, Developmental, and Dialectical Framework for K-12 Teacher Education (RAIL-Ed). arXiv:2608.01705 — https://arxiv.org/abs/2608.01705
2. OECD / European Commission (2026). Empowering Learners for the Age of AI: An AI Literacy Framework for Primary and Secondary Education. — https://ailiteracyframework.org/

Librarian practice (workflow and prompts):

3. Seales, D. L. I. (2025). Leveraging AI in School Libraries: From Basics to Best Practices. ALA Editions / AASL. ISBN 979-8-89255-305-6 — https://www.ala.org/news/2025/09/ai-guidance-school-librarians
4. Qualls, S. (2025). What Would a Librarian Ask ChatGPT? Turns Out… A Lot. TCEA TechNotes — https://blog.tcea.org/librarians-ask-chatgpt-15-prompts/
5. Hatley, C. (2025). A School Librarian's Guide to ChatGPT's Book Scout. TCEA TechNotes — https://blog.tcea.org/school-librarian-guide-chatgpt-book-scout/

Student-side and reading-level (verification):

6. ISTE+ASCD (2026). Profile of an AI-Ready Graduate. — https://iste-ascd.org/ai-ready-graduate
7. advancebookreaders (2026). Reading Level Comparison Chart: Lexile, Guided Reading (F&P), AR/ATOS, DRA. — https://advancebookreaders.com/reading-level-comparison-chart/

Privacy, ethics, and policy guardrails:

8. ALA Council (2026). Guidance on the Use of Artificial Intelligence in Libraries (adopted 16 July 2026). — https://www.ala.org/news/2026/07/ala-council-adopts-guidance-use-artificial-intelligence-libraries
9. Belgrade School District (2025). The Belgrade School District Artificial Intelligence (AI) Guidebook (Final Draft, 16 June 2025). — https://resources.finalsite.net/images/v1764021264/bsd44org/jbp1byuy6owsiawf78sr/BelgradeSchoolDistrictAIGuidebook-FinalDraft6-16-25.pdf
10. TeachAI (2025). AI Guidance for Schools Toolkit. — https://www.teachai.org/toolkit (with live policy tracker at https://www.teachai.org/policy-tracker)