When AI Decides Which Books Deserve Suspicion

Imagine being a school librarian and receiving a list of books that must be reviewed for possible violations of state law. The list includes novels, nonfiction, books about puberty and reproduction, books addressing sexual violence, and even books encouraging teenagers to abstain from sex.

You didn't create the list. You didn't receive individual complaints about each title. Instead, an artificial intelligence system identified the books as potentially problematic. Now you're responsible for reviewing the titles in your library, determining whether they comply with the law, and completing the process by a state-suggested deadline. In the meantime, students cannot access the books.

This is the situation facing media specialists in Palm Beach County Public Schools, according to information obtained from the district via a public records request and shared by the Florida Freedom to Read Project.

The district's use of AI to screen its catalogs has resulted in more than 1,800 new entries in the organization's censorship tracking database. Media specialists have been instructed to review flagged books in their collections by December 18. While those reviews are underway, students are prohibited from accessing the identified titles.

Once the reviews are complete, each media specialist and principal must sign an attestation acknowledging that the review has been conducted and that all available materials comply with state law. The implications extend far beyond one school district.

When artificial intelligence is used to identify books that deserve scrutiny, we must ask whose definition of appropriate content is being applied and what happens when that definition becomes embedded in everyday library operations.

An Impossible Task Under Extraordinary Pressure

Before examining the problems with AI-assisted book reviews, we need to acknowledge the position Florida's school librarians and administrators have been placed in.

Recent state directives have created enormous pressure to scrutinize school library collections for potentially prohibited materials. Vague statutory language, changing interpretations, strict timelines, and the threat of consequences have made an already demanding profession considerably more difficult.

School librarians are expected to maintain collections, assist students, collaborate with teachers, promote literacy, and ensure that materials support their schools' educational missions. Now, they're also expected to anticipate which books state officials might consider objectionable. The workload is staggering.

In Palm Beach County, high school media specialists face a particularly substantial review burden because their collections serve older students and naturally contain more books addressing mature subjects. It's understandable that district leaders would seek technological assistance.

Artificial intelligence can analyze enormous amounts of information quickly. Asking it to search library catalogs for titles that might require additional review may seem like a reasonable way to manage an otherwise overwhelming assignment. We should extend compassion to the educators navigating these circumstances. But acknowledging their predicament does not mean accepting every response to it as inevitable or appropriate.

We must distinguish between understanding why a district adopts a practice and deciding whether that practice should become normal.

When the List Becomes the Punishment

One of the most concerning aspects of the Palm Beach County review is that students cannot access flagged books while the evaluation takes place. That means a book doesn't have to be determined inappropriate to become unavailable. It simply has to appear on a list.

This is an important distinction.

A formal book challenge ordinarily involves an objection to a particular work, followed by an established process for evaluating the complaint. Ideally, the review considers the book in its entirety, its intended audience, its educational or literary value, and the reasons it belongs in the collection.

But when AI generates a list of potentially problematic titles, the process begins before any individual has established that a particular book warrants concern. The technology flags the book. The book becomes unavailable. Only afterward does a human determine whether the flag was justified. For students, the practical result is the same as a temporary removal.

A student looking for a book about puberty, a novel by a favorite author, or a resource addressing sexual violence may discover that it is unavailable, not because a review determined it inappropriate, but because an automated system identified it for further examination.

With more than 1,800 new tracking entries associated with Palm Beach County's review, the potential scope is significant. Even if every flagged title is ultimately returned to circulation, students will have lost access during the review period. And if the process becomes recurring, temporary restrictions could become a routine feature of school library management.

When Abstinence Becomes Suspicious

Perhaps the most revealing detail from Palm Beach County's AI-generated list is the range of nonfiction titles it identified. According to the Florida Freedom to Read Project's tracking, flagged books include materials addressing puberty, reproduction, abstinence, and sexual violence, sexually transmitted diseases, and LGBTQ+ rights.

Consider what that means: A book encouraging teenagers to abstain from sex can be identified for potential review under a system designed to find inappropriate sexual content. That should give everyone pause, regardless of their personal beliefs about what young people should read. The mere presence of a discussion about sex tells us remarkably little about a book's purpose or message.

A health education resource might explain reproduction. A memoir might recount sexual abuse. A novel might explore the consequences of coercion. A book about abstinence might explicitly encourage young people to delay sexual activity. These works are not interchangeable.

Yet automated screening can flatten those distinctions, particularly when its instructions emphasize broad categories of potentially objectionable content. This is one of the fundamental limitations of using AI to evaluate literature: identifying a subject is not the same as understanding how a book treats that subject. And when the legal standards themselves are vague, the problem becomes even more pronounced.

If the system is asked to flag anything that might possibly violate the law, its safest response may be to identify far more books than actually warrant concern. The result isn't necessarily a more accurate review process. It may simply be a much larger list of books temporarily removed from students' reach.

Can AI Determine Literary Merit?

Another troubling element of the Palm Beach County review is the reported use of AI to assess whether books possess literary merit. According to the documents obtained by the Florida Freedom to Read Project, the district's AI-generated list included literary-merit determinations, with some works by bestselling authors (including Stephen King) receiving assessments indicating that they had none.

This raises a serious question: Why are we asking artificial intelligence to make judgments that belong to trained educators, librarians, and literary professionals?

Literary merit is not a simple fact that can be extracted from a database. It involves questions of craft, originality, cultural significance, thematic complexity, historical influence, and the relationship between a work and its intended audience. Readers and critics regularly disagree about the literary value of particular books. Those disagreements are part of a healthy literary culture.

A novel doesn't lack merit because it contains profanity, depicts violence, or explores disturbing subjects. Nor does commercial popularity automatically establish literary merit. The question requires thoughtful human evaluation.

AI can summarize reviews, identify awards, or provide background about an author's work. But its conclusions depend on the information and criteria it receives. When an AI system declares that a book has no literary merit, it can give a subjective assessment the appearance of objective authority. And when that assessment appears on a district-generated review list, it may influence the people tasked with deciding whether students should have access to the book. The technology is no longer simply organizing information. It is helping establish the terms under which literature is judged.

Whose Standards Are We Normalizing?

The people behind organized book challenges are not simply raising isolated concerns about individual titles. Many are working to shift society's understanding of what content is appropriate for young people. Books discussing sexuality, LGBTQ+ identities, racism, sexual violence, and other complicated aspects of human experience have become frequent targets of coordinated challenges.

When school districts use lists of previously challenged books to identify titles for further review, they risk allowing those campaigns to influence local collection decisions. Artificial intelligence can accelerate that process.

A title is challenged in one district. Its appearance on a challenge list becomes a reason for scrutiny elsewhere. An AI system incorporates that history into its assessment. Another district flags the book. Eventually, the existence of previous objections begins to function as evidence that the book itself is questionable.

But being challenged does not establish that a book is inappropriate. It establishes that someone objected. Those are fundamentally different things.

If we allow challenge histories and vague statutory language to become the foundation of automated screening, we risk incorporating the assumptions of censorship campaigns into routine administrative practice.

The consequences can be subtle. Librarians may become more hesitant to purchase books that address controversial subjects. Teachers may avoid recommending them. Districts may conclude that preventing future complaints is easier than defending professional decisions. Over time, the range of books available to students can narrow without a single sweeping prohibition.

This is how censorship becomes normalized: not only through books that are formally banned, but through the quiet acceptance of systems designed to identify, scrutinize, and restrict them.

Compliance Is Not the Same as Collection Integrity

School librarians have always evaluated the books in their collections. They consider age appropriateness, educational relevance, student interests, accuracy, literary quality, and community needs. That is legitimate collection development.

But a review process driven primarily by fear of legal consequences is something different. When districts describe these efforts as matters of "collection integrity," we should examine what that phrase has come to mean. A collection with integrity should offer students access to accurate information, diverse perspectives, meaningful literature, and resources that support their development. It should reflect professional standards rather than political pressure.

A collection is not necessarily stronger because every potentially controversial book has been subjected to suspicion. Nor is a school library more educationally sound because an AI system has screened its catalog for possible objections. The responsibility to comply with the law is real. But so is the responsibility to protect students' access to information. Both deserve serious consideration. And educators should not be forced to navigate that tension without clear standards, adequate resources, and meaningful institutional support.

What Transparency Requires

If districts are going to use AI to identify books for review, communities deserve transparency about the process. What instructions were given to the AI system? What sources informed its assessments? Did it rely on book descriptions, excerpts, full texts, or lists of previously challenged titles? How was literary merit evaluated? What safeguards exist to prevent inaccurate flags? And perhaps most importantly, why must students lose access to books before a qualified professional has determined that a restriction is warranted?

Districts should clearly distinguish between a book flagged by technology, a book formally challenged, and a book found to violate an applicable standard. Those categories should never be treated as equivalent.

The public should also have access to information about how many titles were flagged, how many were ultimately retained, and how long students were unable to access them. Transparency isn't an obstacle to compliance. It's a necessary safeguard against allowing temporary administrative decisions to become permanent limitations on intellectual freedom.

What We Choose to Accept

The greatest danger of AI-assisted collection reviews may not be that artificial intelligence occasionally gets something wrong. It's that we may become comfortable with a system that treats books as suspicious until someone proves otherwise.

We can acknowledge the immense pressure facing Florida's school librarians and administrators. We can recognize that many are doing their best under extraordinarily difficult circumstances. And we can still question the practices that emerge from those circumstances. Because every new procedure establishes expectations about what is normal.

When we normalize using AI to identify potentially objectionable books, we normalize the idea that large portions of school library collections require ongoing suspicion. When we normalize restricting access while those books are reviewed, we normalize removing materials before establishing that they violate any standard. And when we normalize allowing AI to assess literary merit, we diminish the role of professional judgment in decisions that profoundly affect students.

Technology can help educators manage information. It cannot replace their responsibility to evaluate literature thoughtfully, consider context, and protect the educational mission of school libraries. The question isn't simply whether AI can make book reviews faster. It's whether we're allowing the machinery of censorship to become an ordinary part of how our schools operate. What we accept as normal today will shape the libraries students inherit tomorrow.