The image that caught our attention in May 2026
The post has been circulating on Instagram and Reddit over the past few days. It shows the page of a printed computer science textbook explaining the concept of a "row" in a database. And in the middle of the text, circled in red in the original photo, is this sentence that should never have been there:
"If you want, I can also explain columns, primary keys, or other DBMS terms. Here is a clear and simple explanation of a Column in DBMS:"
That sentence is ChatGPT's exact signature. It is the phrasing OpenAI's model uses to offer the user a follow-up topic at the end of a response. Someone copy-pasted a ChatGPT conversation straight into a school textbook, without even taking the time to proofread it. The book was then printed, distributed, sold, and used by real students.
The Instagram post from the @airesearches account that shared the image.
Beyond the amusing side of it, this affair raises a far more troubling question: how many textbooks currently used in classrooms were generated by AI without anyone knowing?
The UCLA case that changes everything
The DBMS textbook affair is anecdotal because the author was betrayed by his own carelessness. But in May 2026, another case, far more official, is up for debate. At the University of California, Los Angeles (UCLA, one of the most prestigious universities in the United States), Professor Zrinka Stahuljak published a textbook entirely generated by AI for her comparative literature course on the Middle Ages and the Renaissance.
The textbook was immediately mocked by the academic world. Its cover (AI-generated) contained incomprehensible phrases such as "Of Nerniacular Latin To An Evoolitun On Nance Langusages". The content was riddled with hallucinations and factual errors. But the professor fully owns it, and even defends her choice:
"I don't see how a traditional $250 textbook, out of date within two or three years, would be preferable to a $25 AI-assisted, personalised textbook based on my own notes."
She even adds that her students' engagement increased with the AI textbook, with some listening to it while walking or exercising. Her colleagues' reaction was ferocious. An English professor wrote on social media: "If you do this, you should have your doctorate stripped and be put in the stocks in the middle of campus."
The UCLA affair is no longer a proofreading mishap. It is a position publicly taken by a senior academic at one of the best American universities. It marks a tipping point: using AI to produce a textbook is no longer a shameful secret, it is becoming a defensible, even proudly claimed, strategy. And nobody yet knows what the right answer to this new paradigm is.
The paradox: long degrees are still sold as a guarantee of seriousness
Here is the contradiction that deserves to be stated out loud. In public debate, long degrees, doctorates and academic expertise continue to be massively valued as the last guarantee of human rigour in the face of AI. "A real thesis is ten years of relentless work." "A university textbook is a career's expertise condensed." "A certified professor is a skill set that AI will never replace."
Except that. When you look at the reality of 2026, the textbooks themselves are starting to be produced by AI. Theses contain undeclared passages generated by ChatGPT. Major academic publishers (Springer Nature, Elsevier, Wiley, Taylor & Francis, Sage) all had to publish formal policies in 2024-2025 to govern this use, because researchers were adopting it en masse.
An audit published in February 2025 in Research Integrity and Peer Review analysed the policies of 162 scientific publishers. The result: more than half have no clear rule on the use of generative AI. And even among those that do have rules, enforcement is almost impossible: current AI detectors have such a high error rate that they are unusable (as Author Media recalled in March 2026: "these tools are not reliable, they guess with false precision").
The real adoption figures in publishing
A study published in Frontiers in Research Metrics and Analytics in April 2025 documented the use of AI in school textbook publishing in Africa. The conclusions are damning: ChatGPT, Grammarly, Quillbot, Jenni and Copy.ai are now used by the majority of publishers and proofreaders throughout the school book production process. The same trend is observed in Europe and the United States, simply less documented because Western publishers generally refuse to communicate publicly on the subject.
| AI tool | Use in textbook publishing |
|---|---|
| Grammarly | Spelling and grammar correction (the most widely used) |
| ChatGPT | Writing paragraphs, plain-language explanations, generating exercises |
| Quillbot | Rewriting existing content |
| Jenni AI | Structured academic writing |
| Copy.ai and Copilot | Marketing content and chapter summaries |
Niall McNulty, a former school publisher who himself tested writing a secondary-school textbook with ChatGPT, sums it up with clarity: "AI does not produce a finished product. But it produces a draft good enough for an experienced editor to work on, and that can save months on a project's schedule." In other words: AI does not (yet) replace human work. But it massively reduces the time needed to produce a textbook.
What this reveals about the AI versus human debate
In public debate, we like to set AI and humans against each other as two watertight worlds. On one side, machines that degrade quality. On the other, humans who guarantee rigour. But the reality is far blurrier, and the DBMS textbook and UCLA cases prove it.
First, humans are already making massive use of AI across every editorial production chain, without declaring it. The printed textbook in the photo is not an exceptional case: it is a case where the concealment failed. How many other textbooks contain AI-generated content without a single suspicious sentence having been left in by mistake?
Second, universities themselves are starting to own up to the practice. The Stahuljak case at UCLA is not isolated: several professors in the United States, Australia and India published AI-assisted textbooks in 2025-2026. None of them lost their job.
Thirdly, and this is probably the most troubling point, the current university system cannot tell the difference between well-made human content and well-edited AI content. AI detectors are unusable, editorial policies are circumvented, and students themselves use ChatGPT to write their assignments. Everyone knows it, no one hides it any more.
We publicly celebrate "real" professors, "real" authors, "real" qualified experts. We scorn "AI writers" and machine-generated content. But behind the scenes, real professors use ChatGPT to prepare their courses, real authors use Grammarly and Quillbot to write, real experts use LLMs to summarise hundreds of articles before signing their own papers. The line between "made by a human" and "made by AI" has not existed for 18 months. We are just pretending.
What this changes for pupils and students
For a pupil or student in 2026, the practical conclusion is unsettling. The textbook they are asked to buy for 50 or 250 euros, presented as the absolute reference produced by a recognised professor, may very well contain passages generated by ChatGPT with no serious proofreading. And the argument "AI makes mistakes, so trust books" no longer holds if the books themselves are produced by AI.
Three practical consequences follow. First, cross-check your sources systematically. If a passage in your textbook seems strange or repetitive, verify the information against other independent sources (Wikipedia, institutional sites, original scientific papers). You will often find that the textbook oversimplified or hallucinated details. Second, use AI yourself as an active learning tool. If your teacher uses ChatGPT to prepare their courses, you may as well use it yourself to go deeper: ask critical questions, request counter-examples, test your understanding. Third, do not overvalue the paper format. A printed textbook is no more reliable than a rigorous website. Reliability depends on the verification process, not the physical medium.
The real question: where is the authority of knowledge heading?
This affair goes far beyond school textbooks. It raises a fundamental philosophical question, in line with what we explored in our article on AI consciousness and in our article on AI control: if AI can produce textbooks, scientific articles, theses, and no one can tell the difference, what still legitimises academic knowledge?
The honest answer is probably that authority over knowledge no longer lies in production, but in verification. A serious teacher or publisher in 2026 is no longer the person who writes every line themselves. They are the person who verifies, corrects, contextualises and takes responsibility for the content produced, whether it comes from a human, an AI, or both. The DBMS textbook in the photo is not a scandal because it was AI-assisted. It is a scandal because nobody took the time to proofread it. It is the human failure of verification that is the real problem, not the origin of the text.
For educational publishers, for universities, for students, and for readers, the lesson is the same. In the age of generative AI, the question is no longer "who wrote this text?". The question is "who verified its quality, and according to what criteria?". And until we rebuild a system of trust around that question, sentences like "If you want, I can also explain primary keys" will keep slipping into our printed textbooks.