Original research

The AI concepts people get wrong most

Which ideas about AI trip learners up? This report is computed from real quiz answers on AI Understanding. It publishes only aggregate rates, only above a minimum sample size, and contains no personal data.

Read the methodology

Data refreshed October 12, 2026. Published under CC BY 4.0: please cite AI Understanding.

Status

Not enough data yet to publish rankings

We will not publish a ranking until each item has at least 30 answers behind it, because small samples are noise. Right now nothing qualifies:

  • 111 first-attempt quiz results exist across 67 quizzes, so no single quiz reaches the minimum.
  • Per-question answers were not recorded until now. Anonymous per-question counters (no user, session or IP) were added to the quiz player; 0 answers have been counted so far.

This page fills in automatically each time the report is regenerated and items pass the minimum. No numbers here are estimated or filled in.

Methodology

How this is calculated

  • Per-question data. Each time a learner checks an answer in a quiz, two anonymous counters for that question go up: answers checked, and answers that were wrong. We do not store who answered, which option they chose, or when.
  • Per-quiz data. From finished quizzes we use only each learner's first attempt at each quiz, so retakes cannot flatter the results. We report the average score and the share who did not pass.
  • Minimum sample. An item is published only with at least 30 answers or first attempts. Everything below is withheld, not rounded up.
  • Who is in the data. People who chose to use our free quizzes. This is not a representative sample of any population, and the same person can appear in many questions.
  • Limits. A high wrong-answer rate can mean a hard concept or a badly worded question. We review very high rates before drawing conclusions, and we do not claim causes.
  • Privacy. The published file holds counts and rates only. No names, emails, accounts or IP addresses are included.

Sample sizes at generation: 111 first-attempt quiz results across 67 quizzes; 0 individual answers counted across 0 questions.

Reuse

Cite or embed

Free to quote and reuse with attribution (CC BY 4.0).

Citation
AI Understanding. "The AI Concepts People Get Wrong Most." AI Understanding Research, October 12, 2026. https://aiunderstanding.org/research/ai-misconceptions
Attribution line for an article or slide (HTML)
<p>Source: <a href="https://aiunderstanding.org/research/ai-misconceptions">The AI Concepts People Get Wrong Most</a>, AI Understanding research (CC BY 4.0).</p>

Researchers and journalists: contact us for methodology questions.