Why it matters
ChatGPT tells a buyer that your distribution covers a discontinued Allen-Bradley PowerFlex drive you never carried. The buyer calls to order it. Your counter staff spend the morning explaining that the part does not exist in your catalog, the phantom inventory was invented by a model, and you cannot ship it. That is a brand hallucination, and it carries real support cost. The model did not misquote a source. It fabricated a fact about you.
For industrial sellers the damage is concrete. False stock claims trigger wasted sales calls. Invented specs put a wrong cross-reference in front of an engineer. A made-up return policy sets an expectation you never agreed to. Each one erodes trust in the answer layer your buyers now start from.
Brand hallucination vs hallucinated attribution
Both are model errors, but they break different things. Brand hallucination is about what was said: a false claim regarding your products, pricing, or terms. Hallucinated attribution is about who said it: the model credits a quote or fact to the wrong source. One corrupts the content. The other corrupts the citation.
In practice
Run your own brand and top SKUs through the major engines on a fixed schedule. Ask the prompts a buyer would: does this distributor stock part X, what is the pressure rating, what is the lead time. Log every fabricated spec, phantom stock claim, or invented policy. The fix is supply-side. Publish clean, structured PIM data and machine-readable catalog facts so the engine has a correct source to ground on instead of guessing.