When AI Produces Knowledge, Do Textbooks Still Matter?
Jul 25, 2026
By Jiang Zhouzi
Last month, the 2026 World Digital Education Conference convened in Hangzhou. Education ministers and experts from 82 countries gathered under the theme: “AI + Education: Transformation, Development, Governance.” Among the eight key outcomes announced, one concept drew particular attention: teacher–student–machine triadic collaboration.
In the classroom of the future, teachers, students, and AI are positioned as three equal participants in the learning process. The first reaction this prompted in the author was simple: if AI is an “equal participant” in teaching, then where does the textbook stand?
The question of how technology reshapes education is not new. In 2011, Steve Jobs once asked why information technology had transformed almost every sector over the past two decades, yet had produced surprisingly little visible change in education. More than a decade later, the “Jobs question” still lacks a convincing answer. In the author’s view, this is because the framing has been misplaced. The implicit assumption has always been that technology changes education by changing how teaching is done. MOOCs, flipped classrooms, adaptive learning systems—all operate at the level of pedagogy.
But something different is happening this time. A neglected fact is becoming clear: generative AI is not merely changing how we teach. It is changing what knowledge is.
💡 Core Strategic Takeaway: The Epistemological Shift
- The Pedagogical Trap: For a decade, we assumed technology would change education by altering teaching methods. We were wrong. AI is shifting the very foundation of "what is taught."
- The Dissolution of Certainty: When knowledge is no longer singular, stable, or exclusively human-produced, the traditional educational publishing system—built on the assumption of stable, verified facts—loses its footing.
1. The Foundation of the Knowledge System Is Being Displaced
More than 2,400 years ago, Plato offered a definition: knowledge is justified true belief. It requires reasons, evidence, and the ability to withstand scrutiny. This definition has underpinned education for centuries. It supports a basic assumption: what schools teach has been validated, is reliable, and is worth transmitting across generations. AI is now dismantling this assumption layer by layer.
In a 2022 paper titled “The Landscape of Knowledge in the Intelligent Era: AI and the Reconfiguration of Epistemology,” researchers from East China Normal University analyzed this shift. Here is the reconfiguration across four core dimensions:
Who produces knowledge?
Early AI systems embedded human knowledge. Expert systems and rule-based engines simply executed human-defined logic. Knowledge was transferred into machines.
Then AlphaGo Zero emerged. With no human training data, it learned three board games in three days and defeated earlier versions that had already beaten Lee Sedol. For the first time, machines were not only applying knowledge—they were generating it.
Today, in fields such as protein structure prediction, drug discovery, and materials science, AI is already producing findings that exceed human capability in specific domains.
The subject of knowledge production has shifted. From human monologue to human–machine co-production. Humans become coordinators and validators; machines become collaborative producers.
But a deeper question follows: when a student opens a phone and asks AI to explain photosynthesis, bypassing teachers, textbooks, and editorial review systems, where is the “gatekeeper”? And who is accountable for the explanation they receive?
What is knowledge, exactly?
Traditionally, knowledge implied verification. AI is producing something different: what scholars call soft knowledge.
Soft knowledge is extracted from data by models. It may be correct or incorrect. It is not yet verified, but is already being used.
It differs from traditional hard knowledge. Hard knowledge emphasizes justification and traceability; soft knowledge prioritizes immediacy and usability.
Hard knowledge will not disappear. But the structure is changing. Hard knowledge provides reliability and accountability. Soft knowledge provides speed and breadth. The division of labor is shifting, not replacing.
What does knowledge look like?
This is one of the most revealing conceptual shifts.
Knowledge has moved through three metaphors:
- In the pre-modern and modern era, knowledge resembled a building: hierarchical, structured, stable. A textbook table of contents was effectively an architectural blueprint.
- In the internet era, knowledge became a pipeline: networked, interconnected, flowing through hyperlinks.
- In the intelligent era, knowledge resembles a dance: continuously generated through interaction among humans, machines, and data, without fixed form.
Knowledge no longer exists independently of interaction. It emerges from it.
How is knowledge presented?
Text, images, audio, and video have long coexisted. Now AI pushes knowledge into multimodal form—readable, audible, experiential, immersive.
| Epistemological Dimension | Traditional Paradigm | The AI-Driven Reality |
|---|---|---|
| Who Produces Knowledge? | Human Monologue. Experts define logic; systems merely execute it. | Human-Machine Co-Production. Models generate novel findings (e.g., AlphaGo Zero, protein folding) bypassing human trainers. |
| What is Knowledge? | "Hard Knowledge" emphasizing justification, stability, and traceability. | "Soft Knowledge" extracted from data. Prioritizes immediacy and usability over absolute verification. |
| What Does it Look Like? | A Building (hierarchical/structured) or a Pipeline (networked/hyperlinked). | A Dance. Continuously generated through dynamic interaction, lacking a fixed, permanent form. |
| How is it Presented? | Digitized text, static images, and linear audio/video. | Immersive Experience. Knowledge is no longer "read"; it is virtually entered and interacted with. |
2. Who Is Thinking on Your Behalf?
If knowledge is changing, the more urgent question is this: as access to knowledge becomes instantaneous, what is being lost in the process of learning? A recent paper by Professor Zhang Liang of Southwest University examines this issue through the lens of procedural knowledge versus substantive knowledge.
Substantive knowledge treats knowledge as a finished object. Students receive it and reproduce it in exams. AI is now restoring this oldest misconception with unprecedented efficiency. Students no longer need to search, derive, or test hypotheses. They ask a question, and the answer appears instantly. Professor Zhang calls this “instant-answer knowledge welfare.” But this welfare carries three severe risks:
| The Risk Factor | Cognitive Impact |
|---|---|
| Cognitive Outsourcing | Thinking is delegated to AI. Unused cognitive processes weaken silently. Students believe they have learned, when in fact they have only received outputs. |
| Cognitive Substitution | AI simulates the entire reasoning pathway. Students no longer feel the need to understand how an answer is derived. Input a question, output an answer. No confusion. No "why?". |
| Information Enclosure | AI returns what aligns with user expectations. Knowledge is shaped into preference satisfaction rather than exposure to intellectual challenge. |
3. Where Publishers Go From Here
If the subject of knowledge is changing, and the structure of knowledge is changing, and the channels of learning are changing, what remains for publishers? The answer is: they still have a role. But not in their current form.
Tao Xingzhi once wrote: “Knowledge is like grafting branches.” Existing knowledge is the rootstock. New knowledge is the graft. Only when they are integrated does genuine understanding emerge. AI can provide the branch instantly. But it cannot perform the grafting. It cannot integrate knowledge into lived experience. This is where publishing retains its grounding across five new strategic directions:
| The New Role | Strategic Imperative |
|---|---|
| Knowledge Curator | In an era of abundance, scarcity shifts to selection and sequencing. What should an 8-year-old learn first? What core concepts matter most for a high school student? Publishers no longer win by covering everything. AI can already do that faster. Their role is to build orientation in the knowledge landscape. The Harvard Project Zero framework of "teaching for understanding" emphasizes core concepts rather than coverage. Similarly, recent curriculum redesign work highlights a shift toward conceptual and epistemic mastery, which AI alone cannot structure meaningfully. |
| Learning Experience Designer | From content delivery to experience design. A textbook should not be a list of knowledge points. It should be a structured inquiry journey. AI answers "what is it." Publishers must design "how to think about it." Good learning design creates cognitive tension, mistakes, pauses, and discovery. For example, instead of explaining buoyancy and then assigning exercises, a student could first confront a paradox: why does a metal block sink while a metal ship floats? Confusion becomes the starting point of understanding. AI cannot fully replicate this, because it does not experience human cognition. |
| Developer of Educational Agents | China’s national AI strategy explicitly calls for integrating intelligent systems into teaching and learning processes, and building interactive education environments. For publishers, the most valuable asset is not paper, but structured, validated knowledge systems. These can become the foundation of educational AI agents—traceable, auditable, curriculum-aligned assistants embedded in learning environments. Selling books is selling containers of knowledge. Building agents is selling systems of knowledge. |
| Custodian of Cognitive Safety | This may be the most irreplaceable role. AI-generated knowledge carries structural limitations: it is probabilistic, not definitive; it is opaque in provenance; and it lacks accountable authorship. Publishers can provide something AI cannot: institutional responsibility for accuracy. In a world saturated with plausible but unverified knowledge, someone must still be accountable for what is presented as "true." Governments set standards. Schools use materials. Technology companies prioritize speed. Publishers remain one of the few institutions structurally positioned to prioritize correctness. |
| Connector of the Ecosystem | AI is reshaping education from a closed system into an open network. In this environment, publishers are not merely content providers but connectors—linking policy, schools, technology companies, teachers, and students. For example, decades of teacher training data and classroom feedback could be integrated with AI systems to improve learning models. Textbook usage data across thousands of schools could be connected to education quality monitoring systems, shifting publishers from one-off distribution to continuous diagnostic roles. This is what ecosystem connectivity means in practice. |
Closing Thought: Returning to the Day of Textbook Review
What unsettled the author that day was not that AI could produce a textbook. It was something deeper: if a generation grows up with instant answers, will they still think? Will they wrestle with questions long enough for insight to emerge? Will they experience the moment of realization that no system can automate: “Now I understand—this is my own thinking.”
That is the real meaning of knowledge. Knowledge is not merely information. It is the process through which a person becomes themselves. If publishing can preserve this process—ensuring knowledge is not only accessed but experienced—it retains a lasting purpose. Textbooks will not disappear. But publishers that only produce textbooks might.
❓ Frequently Asked Questions
Q: What is the difference between "Hard Knowledge" and "Soft Knowledge" in the AI era?
A: Hard knowledge relies on rigorous justification, traceability, and verified facts—the traditional bedrock of textbooks. Soft knowledge is extracted probabilistically from data by AI models; it prioritizes immediacy and usability but may lack strict verification.
Q: How should educational publishers adapt to generative AI?
A: Publishers must pivot from being mere content providers to becoming knowledge curators, learning experience designers, and custodians of cognitive safety. By embedding their validated knowledge into educational AI agents, they shift from selling containers of knowledge to systems of knowledge.
🎓 Deepen Your Strategic Mastery
Are you ready to adapt your business strategy as AI redefines the nature of knowledge and cognitive work? In the SOLOMOAT Mini MBAs, we decode the precise AI toolchains, cognitive frameworks, and strategic business models required to thrive in the intelligent era.
