Deck the Tree with Bogus Blossoms

The first thing I want to say about AI confabulation is that it fooled me too — before I knew what to look for. I was fact-checking my daughter's assignment and I nearly let one fabricated citation slide because the journal name sounded real. That is how well-designed this problem is. It is not a question of paying more attention. It is structural.

Generative AI does not retrieve information. It predicts the next most plausible token based on patterns in its training data. This distinction is not semantic — it is structural. When your child asks an AI system for historical facts, medical advice, or legal precedent, the system is not consulting a database. It is generating text that resembles the patterns associated with authoritative responses. The result is often accurate. But when it is wrong, it is wrong with the same confidence as when it is right.

A 2023 study by researchers at Stanford and UC Berkeley found that large language models produce factual errors in 15 to 20 percent of responses when asked about verifiable information, and the error rate climbs to 40 percent for niche or technical domains.^[1] More concerning: the models do not flag uncertainty. They present fabricated citations, nonexistent studies, and invented statistics with the same grammatical authority as verified facts. Your child cannot hear the difference. Neither can you.

This is the bogus blossom. The tree appears full because the language is fluent, the tone is confident, and the structure mimics expertise. But there are no roots — no retrieval, no verification, no epistemological grounding. The illusion is so convincing that even professionals are deceived. A 2024 survey of knowledge workers by McKinsey found that 63 percent of respondents reported using AI-generated content in client deliverables without independent verification.^[2] The assumption is that something this coherent must be true.

Your child is growing up in an illusion economy. Every homework question, every college essay prompt, every career research query can be answered by a system that decorates the tree faster than any human can fact-check it. The skill they need is not the ability to avoid AI — that is impossible. The skill is the ability to strip the blossoms off and examine the branches beneath.

Human cognition is wired to equate fluency with truth. Psychologists call this the fluency heuristic: information that is easy to process feels more credible.^[3] When your child reads an AI-generated paragraph that flows smoothly, uses academic vocabulary, and mirrors the structure of authoritative writing, their brain interprets that fluency as a signal of accuracy. The same mechanism that helps them navigate daily life — trusting coherent explanations over garbled ones — becomes a liability in an environment where coherence is artificially generated.

A 2023 study published in Nature Human Behaviour found that participants rated AI-generated misinformation as more believable than human-written misinformation when both contained the same factual errors, specifically because the AI version was more grammatically polished.^[4] The bogus blossoms looked better than the real ones. This is not a problem your child can solve by paying closer attention. The deception operates below the threshold of conscious awareness.

The danger compounds when AI systems cite sources. A fabricated citation looks identical to a real one if your child does not verify it. In a 2024 analysis of student research papers by the Stanford Graduate School of Education, 41 percent of high school students who used AI research tools submitted at least one citation to a source that did not exist.^[5] The students were not cheating. They were trusting a system that had decorated the tree so convincingly that they never thought to check the roots.

Your child is not uniquely vulnerable to this. You are vulnerable to it. Every adult who has trusted a polished slide deck, a confident consultant, or a well-formatted report without tracing the data back to its origin has fallen for the same mechanism. The difference is that AI scales the deception to every query, every draft, and every decision. The parent who teaches verification is not teaching paranoia. They are teaching the habit of checking for roots before trusting the blossoms.

In the stratagem, the deception is intentional. In the AI era, the deception is structural — a side effect of how these systems generate language. Your child must learn to see through both.

Teaching your child to verify AI output is not about teaching them to distrust technology. It is about teaching them the same discipline that any serious professional applies to any source of information: trace the claim to its origin, evaluate the quality of the evidence, and distinguish interpretation from fact. This is not new. It is the foundation of scholarship, journalism, and scientific inquiry. What is new is that your child can no longer rely on institutional gatekeepers to do this work for them. The AI they use at home generates output that looks as authoritative as anything they will encounter in school or work.

The verification habit has three components. First, your child must learn to identify falsifiable claims within AI-generated content. Not every sentence is a factual assertion, but when the AI states a statistic, names a study, or describes a historical event, your child should flag it as something that can be checked. Second, they must learn to trace those claims to primary sources. This does not mean rejecting secondary interpretation — it means knowing where the interpretation ends and the evidence begins. Third, they must learn to evaluate source quality. Not all citations are equal, and not all sources are reliable, even when they exist.

A 2024 report by the OECD on digital literacy in secondary education found that students who were explicitly taught source evaluation skills were 3.2 times more likely to detect AI-generated misinformation than students who received general media literacy training.^[6] The difference was not intelligence or effort. The difference was method. The students who knew how to verify did not need to be more skeptical — they just needed to know what to check and how to check it.

This is the skill that survives automation. When AI can generate a polished research paper in 30 seconds, the human skill that remains valuable is the ability to evaluate whether that paper is built on evidence or on blossoms with no roots. Your child will work alongside colleagues who trust the output because it looks good. They will compete with peers who submit the first draft because it sounds confident. The child who verifies will be the one whose work is trusted when the stakes are high.

The mistake is thinking that the solution is to ban AI or to teach your child to avoid it entirely. I went through this phase briefly — my instinct after finding the fabricated citation was to say no more AI for homework. That lasted about a week before I realized it was the wrong response entirely. This is the equivalent of teaching them to avoid cars because cars can crash. The problem is not the tool. The problem is the assumption that fluent output is verified output. When you ban AI, you do not teach verification — you just delay the moment when your child encounters the illusion without your guidance.

Some parents make the opposite mistake: they trust AI because it feels like progress, and they assume that skepticism is a rejection of technology. This is equally dangerous. Your child does not need to be anti-AI. They need to be pro-evidence. The parent who raises a child capable of verification is not raising a Luddite. They are raising someone who can use AI as a starting point and still deliver work that holds up under scrutiny. That is the difference between a professional and someone who gets replaced when the stakes get higher.

The other mistake is assuming that schools will teach this. Some will. Most will not, at least not at the depth required. A 2024 survey by the National Education Association found that only 18 percent of U.S. high schools had integrated explicit AI verification training into their curricula.^[7] The rest are still treating digital literacy as a module on identifying phishing emails. Your child will graduate without the habit of checking sources unless you teach it at home. The parent who waits for the school system to solve this problem is the parent whose child decorates their college essays and job applications with bogus blossoms.