Your Visibility Dropped Overnight. Model Update or Your Fault?
A drop across every prompt on one engine is a model or index change. A drop on a few prompts across every engine is a source of yours…
Invisibility in AI answers has four distinct causes, and they do not share a fix. This section helps you work out which one you have.
The checks here are ones you can run yourself, in an afternoon, without buying anything. Most of them will tell you more than a paid audit will, because the thing that makes an audit credible is not the tool, it is holding the prompt set and the run count constant. Do that and a free check produces a defensible baseline.
If your engines describe you accurately but never list you, that is one problem. If they cannot describe you at all, or they confuse you with somebody else, that is a different one entirely.
A drop across every prompt on one engine is a model or index change. A drop on a few prompts across every engine is a source of yours…
Three numbers get misread in every AI visibility report: a mention taken as a recommendation, one run taken as a rate, a branded prompt taken as reach.
A minimum viable prompt set covers four intents. Most audits test only the first and overstate the result. Here are the twelve prompts, ready to copy.
Described accurately but never listed is a content problem. Confused with another company is an entity problem. Two questions separate them.
A competitor in your slot means the engine resolved their entity and not yours. It is a legibility signal, not a verdict on product or market share.
Wrong descriptions come from stale sources. Ask the engine to cite the specific claim, then check the date, instead of arguing with the output.
Ranking and recommendation read different evidence. One page can win a query and contribute nothing to whether a company gets named in an answer.
A free audit is credible if it holds three things constant: the prompt list, the runs per prompt, and the session state. Here is the method.
Being known but never shortlisted is an evidence problem, not a content problem. Engines source comparison answers from domains you do not own.
Three prompts in a clean session tell you whether an engine can describe you, will recommend you, or has you wrong. Five minutes, no tooling, no budget.
Four causes make a brand invisible in AI answers, and each has a different fix. Here is the afternoon-long diagnostic that tells you which one you have.