What is brand safety?
Brand safety originally meant keeping a company's ads away from content that could embarrass or harm it, and it now extends to whether an AI system says accurate, non-damaging things about the brand when asked.
The classic version of this problem was about placement: an ad accidentally running next to violent or offensive content, damaging the brand by association. Marketing teams built entire tooling categories around avoiding that, using blocklists and content categorization to control where ads could appear.
An AI answer engine creates a different kind of risk that a blocklist cannot fix. There is no ad placement decision to make; the exposure is the model's own account of the brand, generated when a user asks about it, which can be outdated, wrong, or unfavorable regardless of anything the brand's marketing team chose to publish. A company cannot opt out of being discussed the way it can opt an ad out of a certain publisher, which makes this a harder and newer category of brand safety to manage.
Related
- HallucinationA hallucination is a confident, fluent statement from a model that is not true. In a brand context it is the model getting your pricing, features or ownership wrong.
- EntityAn entity is a thing an engine models as a distinct object in the world, such as a specific company, product or person, as opposed to a string of text.
- Mention vs citationA mention is the model naming you in its answer. A citation is the model linking a source. You can have either without the other, and they mean different things.
Want to know where you actually stand on this? Run a free visibility check or try the free tools.