An AI hallucination is model-generated information that is presented as if it were grounded in evidence but is unsupported, fabricated, or inconsistent with the available source material.
How it works
Hallucinations can appear as invented numbers, citations, dates, quotations, causal claims, or confident details that were never present in the source. Fluency is not evidence of factual grounding.
A practical audit separates source-backed facts from unsourced numbers, impossible dates, and reasonable-but-unverified inferences, then traces important claims back to primary material.
Why it matters
Trading research often contains numbers, timestamps, corporate disclosures, and rule definitions where a small factual error can materially change a conclusion or calculation.
The goal is not to avoid AI entirely. It is to use it for acceleration while preserving source verification and human responsibility for risk decisions.
A simple market example
An AI summary correctly reports a company's revenue but invents an exact institutional-flow figure with no cited dataset. The paragraph can sound coherent while one critical number is hallucinated.
Common mistakes
Treating a confident tone as evidence that a claim is sourced.
Checking only the final conclusion while leaving dates, units, and intermediate numbers unverified.
Frequently asked questions
Does every AI error count as a hallucination?
Not necessarily. Errors can also come from misunderstood prompts, stale sources, calculation mistakes, or bad input data.
How should financial claims be checked?
Trace material facts, dates, units, and numbers to authoritative source documents or datasets.
Can citations eliminate hallucinations?
No. Citations can be wrong or irrelevant, so the cited source still needs to support the claim.
Educational content only. Definitions describe common market usage and may vary by jurisdiction, instrument, or institution.