AIEd Knowledge BaseAIEd KB
On this page

Concept

Information and Media Literacy

created 2026-11-05 · updated 2026-11-05

Information and media literacy are adjacent capabilities that AI is rewriting at the same seam. Information literacy centers on finding and verifying accurate data; media literacy centers on understanding who created a message, how it was constructed, and why. They overlap in their shared goal of building the Critical Thinking that lets people navigate a digital world and spot misinformation — and they now share a second problem: generative AI produces fluent, plausible content whose origin and intent are hidden, so the provenance question that media literacy always asked has moved into information literacy too.

Questions to Consider

  • When a chatbot returns a confident paragraph with three citations, is the student exercising information literacy, media literacy, or neither — and does the assignment make that distinction visible?
  • Deepfakes and fabricated citations are different failures: one is a false record of the world, the other a false record of the literature. Can one course teach both without teaching each badly?
  • AI detectors and fact-checkers are themselves models with error rates. What happens to a verification curriculum when the verification tool is probabilistic?
  • Media literacy has been taught through production — students make media to understand it. If a model now makes the artifact, what part of the pedagogy survives?

Introduction

The two literacies have separate institutional homes. Information literacy was codified in library and information science, most visibly in the ACRL Framework, as a set of dispositions around authority, process and value. Media literacy grew out of communication and media studies and out of school media education, and asks about authorship, technique, audience and effect. One judges whether a source deserves belief; the other asks who made a message and what it is doing.

That division has always been cleaner in theory than in a classroom. Both literacies teach students to slow down, check provenance and resist fluency. Both were built on an assumption that the artifact had an author who paid a cost to produce it. Generative AI removes that assumption from both at once: text is synthesized from an untraceable corpus, images and voices are cloned, and automated accounts simulate the audience itself.

This page covers the pair because the corpus treats them as one instructional problem with two outcome sets. The evidence is uneven — a bibliometric map and two systematic reviews on the information side, a small intervention and an expert interview study on the media side — and almost none of it tests a combined curriculum. That absence is itself the finding worth carrying into practice.

Where the two constructs diverge

Information literacy's object is the source, and its core question is whether the claim is true and the evidence behind it credible. Media literacy's object is the message, and its core question is who made it, how, for whom and to what end. A fabricated citation fails the first test; a deepfake may pass every factual check a student knows how to run while failing the second.

The empirical literatures reflect the split. A survey of 236 US university students asked where academic searches begin and found search engines at 45.5%, the library homepage at 27.9% and an AI chatbot at 26.6%, with international students substituting AI for library resources at roughly double the domestic rate (Lund et al., 2026). A systematic review of 32 studies found AI literacy integration in information-literacy instruction at an "infancy stage," delivered through LibGuides and workshops rather than curriculum (Mirza and Arif, 2026). On the media side, interviews with seven experts and a survey of 141 students found technology inclusion in the critical media literacy curriculum "limited and often superficial," with teachers under-trained for tools that produce disinformation and deepfakes (the critical media literacy study, 2026).

How AI complicates information literacy

The first complication is fabrication at scale. A model produces convincing text, plausible references and confident statistics that may not exist; the reader receives a synthesis whose search trace has been erased. Checking becomes a different skill: not "is this a good source" but "does this source exist, and does it say what the answer claims."

The second is that the search layer collapsed into an answer layer. When the tool retrieves, selects and writes, the literate act moves to judging the synthesis. That is the residue the evidence keeps pointing at: an embedded-librarian platform beat conventional library instruction by 0.78 to 0.25 points on an ACRL-based instrument, and the consultations that produced the gain were dominated by source-evaluation questions (Huang, 2026).

The third is capacity. In a survey of 412 academic librarians across 38 Nigerian institutions, self-rated competence was high for database instruction (M = 4.07) and information-literacy delivery (M = 4.02), but composite AI competency was 2.86, weakest on designing objectives (M = 2.70) and assessing outcomes (M = 2.62) (Dada, 2026). A field mapped at 1,669 publications is being asked to teach a competency its instructors report they cannot yet design (Jabeen and Lux, 2026).

How AI complicates media literacy

Media literacy's founding questions assumed a maker with a purpose. Deepfakes, voice cloning and automated bots break that assumption in three ways: the artifact may have no human author, the audience may be synthetic, and the intent may be optimization for engagement rather than persuasion by anyone.

The curriculum evidence suggests schools have not caught up. The expert interviews describe teacher training as the binding constraint rather than student resistance, and argue for a Critical Pedagogy framing that treats evaluation as ethical and participatory practice instead of tool training. A commissioned report for public service media found provision for 14–19-year-olds tilting toward functional AI skills and away from critical evaluation, despite reviewing more than 40 frameworks and interviewing 35 experts (the AI literacies report, 2025).

AI as a defensive tool — claims and caveats

The case for using AI on the defensive side is plausible and widely made: automated credibility checks that scan a source's history and language for bias signals, deepfake detectors, metadata analyzers that look for provenance markers, and simulations that let students practice evaluation on realistic synthetic material. These capabilities are real products, and the classroom use is real practice.

The evidence is thinner than the claims. Detectors are probabilistic and degrade as generators improve; a fact-checking model is itself a model, so the student still has to judge the checker; and provenance metadata only helps when a producer chooses to attach it. One of the corpus's firmest results is that AI-supported media literacy instruction can work at small scale: an 18-hour program built on the 5E model produced large gains among 36 fourth-grade students in one Turkish school (Demir and Akar, 2026). Thirty-six students is a proof of concept, not a procurement case.

A library-sector opinion piece pushes the same idea further: AI could carry information-literacy delivery through retrieval, adaptive tutorials, chatbots and automated fact-checking, with librarians rather than vendors owning the integration (Offem and Arikpo, 2026). The paper reports no data, so it is a design brief for where instruction might live rather than evidence that it works.

Read together, the two sides of the ledger suggest a rule: AI can carry the practice load — realistic scenarios, immediate feedback, scale — but the judgment load stays with the learner, and any tool that claims to replace it should be evaluated as a detector with a measured error rate, not adopted as an answer key.

What this means for practice

  • Teach one procedure, two objects. The checkable routine — trace the claim to a source, trace the source to an author, ask what the artifact is doing — works for a citation and for a video. Name which object the assignment is assessing.
  • Treat detection as evidence, not verdict. A detector's output is a probability; teach students to corroborate with a second method and to say what would change their mind.
  • Keep production in the curriculum. Making media was the pedagogy that made critique concrete; generating with AI can support it only if students examine the choices the tool made for them.
  • Put a librarian or media specialist inside the workflow. The strongest instructional result in this literature comes from a design where human consultation was built in, not offered afterwards.
  • Assess staff capacity before writing the syllabus. The measured deficit sits at designing objectives and assessing outcomes, not at using tools.

Limitations of the evidence

Most studies here are small, self-reported or both: 236 students in one survey, 36 children in one intervention, 412 librarians rating themselves. Bibliometric visibility is not practice, and the scientometric authors say so explicitly. No study in the corpus evaluates a combined information-and-media-literacy curriculum, and the detector and fact-checking tools the practice literature recommends are largely unmeasured in these sources. The claim that AI-powered verification improves outcomes should be treated as a design hypothesis until someone tests it against a control.

Connected Concepts

Connected Articles

Metadata

LevelHigher Ed, K 12, Adult Learning
Intended audienceLibrarians, Instructors, Learners, Curriculum Designers
Learning and instructionCritical Pedagogy
Technologies and techniquesGenerative AI, Large Language Models (LLMs)
Equity, ethics, and responsible useTrust Calibration
Foundations of AI in educationInformation and Media Literacy, AI Literacy, Critical Thinking, Academic Integrity
Page kindSynthesis
SourceHistoryEditChat

Embed this page

Copy the code below to embed this page in a learning management system or other website. The embedded version shows only the page content — it hides the site header, navigation, and footer.