§ 33 · Glossary · Hallucination

Hallucination Hallucination Confident, fluent, and wrong.

A hallucination is an AI-generated output that is confidently stated but factually wrong — a fake citation, an invented statistic, a non-existent quote. Hallucinations are the central failure mode of generative AI and the reason grounding, RAG, and citation-friendly content matter: real sources cut hallucination rate.

Updated 2026-05-01 2 FAQs

How Hallucination differs from Grounding, RAG, LLM

A model error is any wrong output. A hallucination is specifically a confident, fluent, plausible-sounding wrong output — which is what makes it dangerous: it doesn't trip a user's skepticism the way garbled output would.

How Mentionwell handles Hallucination

  • Editorial critic enforces evidence-per-claim so generated articles are themselves grounded.
  • Per-article Markdown mirrors give downstream engines clean source material to ground against.
  • Inline citations to authoritative sources so claims are checkable.

Frequently asked questions about Hallucination

What is an AI hallucination?

An AI-generated output that's confidently stated but factually wrong — fake citations, invented statistics, non-existent quotes. The central failure mode of generative AI.

How do you reduce hallucinations?

Ground answers in retrieved sources (RAG), require citations, use a critic-loop to check claims, and run outputs through structured validators when the schema allows.

See also

Ship Hallucination-optimized articles automatically

Mentionwell handles Hallucination on every published article — alongside the other six optimization targets in this glossary — so you don't have to think about it per post. Drop a domain, approve the first headline, watch the pipeline run.

Sign in →

§ · The rest of the alphabet

The rest of the
alphabet.

Hallucination is one of 34 terms in the AI search vocabulary. Mentionwell optimizes for all of them on every article. Browse the full glossary →

§ 06 · Launch

Drop a domain.
Get cited.

Mentionwell opens to early access on June 1. Headless on any framework. No CMS migration.

EARLY ACCESS · INVITE-ONLY AT LAUNCH