What is a prompt in GEO?
In GEO, a prompt is the question or request a user puts to an AI assistant, and it replaces the keyword as the basic unit of visibility tracking. Rather than asking where you rank for a term, GEO asks which brands the engine names and cites when it answers a given prompt.
Prompts differ from keywords in length and intent. A keyword is a compressed fragment typed into a search box, such as "accounting software uk". A prompt is a full request with context attached: "what's the best accounting software for a UK sole trader who does everything on their phone?" The intent that SEO tools infer from keywords is stated outright in a prompt, which is why AI answers can be so specific about who they recommend.
GEO measurement is built on prompt sets. You define a list of questions your customers plausibly ask, put them to the engines you care about, and record which brands get named, which get cited and how each is described. Aggregated over many prompts and runs, this produces metrics such as share of voice in AI answers.
One honest caveat: AI answers are not deterministic. The same prompt can produce different brand lists on different runs, because models sample their outputs and retrieval results shift, and location or account context can alter what comes back. A single run is a snapshot, not a ranking, and any tool claiming a precise stable "position" for a prompt is overclaiming.
When building a prompt set, start from how buyers actually talk. Sales calls, support tickets and community threads yield better prompts than keyword tools, though keyword data can seed phrasing. Prioritise commercial-intent patterns such as "best X for Y", "alternatives to Z" and direct comparisons, since those are the prompts where being named translates into pipeline.