What Is an AI Hallucination?

In one sentence

A hallucination is when a language model states something false or fabricated with the same confidence it states facts — including about your brand.

What Hallucination means

Models generate plausible continuations rather than looking facts up. When the underlying signal is thin, ambiguous or contradictory, the output can be confidently wrong.

For brands this shows up as invented features, wrong pricing tiers, confusion with a similarly-named company, or capabilities you don't have.

Why it matters for AI search

It's a commercial risk, not just a technical curiosity. A buyer told you don't support an integration you do support may never contact you to check.

In regulated categories it's a compliance issue as well — a model stating outdated rates or implying availability in an unlicensed market creates real exposure.

See it in action

What conditions produce hallucinated brand information.

Hallucination risk factors

Diagnostic
$ ron crawl --agents all
Retrievable current informationwaiting
Entity resolves unambiguouslywaiting
Independent corroborationwaiting
Own properties agreewaiting
Resultwaiting

Every risk factor is something you control.

Hallucination about your brand is usually a signal problem rather than a model problem. Thin, ambiguous, contradictory inputs produce confident nonsense.

How to get it right

Reducing hallucination about your brand

  • Make accurate current information retrievable, so the model has something to ground on
  • Resolve your entity so it isn't blending you with a similarly-named company
  • Eliminate contradictions between your own properties
  • Build independent corroboration for the facts you most need stated correctly
  • Monitor monthly for factual errors and trace each back to its probable source

Common questions

Can we stop models hallucinating about us?

Not entirely — it's inherent to how models generate. You can reduce it substantially by removing the conditions that cause it.

Thin signal, ambiguous entity and contradictory sources are the main drivers, and all three are fixable.

Is hallucination a legal risk?

Potentially, in regulated categories. Models stating outdated rates or implying availability where you aren't licensed creates exposure.

We flag factual errors for legal and compliance review rather than treating them as marketing metrics.

How do we find out what models say about us?

Run a fixed prompt set covering your products, pricing and positioning on a schedule, and record what comes back.

Doing it once tells you little; doing it monthly shows you what's changing and whether corrections landed.

These come up alongside Hallucination constantly.

Free 12-point check

Send Ron your site

Two fields. We’ll crawl your site as GPTBot, ClaudeBot and PerplexityBot, benchmark a sample of your category’s prompts, and send you what we find.

  • What each AI crawler actually receives from your site
  • Whether models resolve your brand as a real entity
  • A sample of category prompts and who gets cited
  • The three fixes we’d prioritise first

No commitment, no sales sequence. If we’re not a fit, we’ll say so.

We’ll pull your favicon so you know we found the right site.

We reply from a real address. No sequence, no newsletter.