// glossary entry

Hallucination (AI)

What is an AI hallucination?

An AI hallucination occurs when a language model generates content that sounds plausible but is factually incorrect or fabricated, such as false sources, made-up numbers, or hallucinated library functions. Current state-of-the-art models (GPT-5, Claude Sonnet, Gemini 2.5) have hallucination rates ranging from 3 to 15 percent, depending on the benchmark, and significantly higher for complex tasks. For SMEs, this means that AI output without editorial review poses a liability risk. At Waterproof Web Wizard, every AI-generated draft undergoes fact-checking, source verification, and E-E-A-T assessment before it goes live.

// synonyms

  • AI hallucination
  • Hallucination
  • AI Hallucination
  • LLM Hallucination
  • Hallucinations (AI)
Back to Glossary
Is the term still unclear?

Let's talk for 15 minutes

If a term in this glossary isn't entirely clear in the context of your project, or if you want to know whether the concept is relevant to your situation—just ask us. The initial consultation is free and non-binding.

Book an initial consultation Call: +49 751 95899217
Free No obligation 15 minutes