A prompt is the question or instruction someone types into an AI assistant. Whereas people often use two or three loose words in a classic search engine, in ChatGPT, Gemini or Perplexity they usually write full sentences, including context, preferences and constraints. That changes how you, as a business, need to think about being found.
For marketers, a prompt is therefore the AI-world equivalent of a search query. If you understand which prompts your customers use, you can align your website and content with them and track whether your business appears in the answers.
A classic keyword is short and general, such as "accountant London". A prompt is often much richer: "I run a small online store in London and I'm looking for an accountant who knows how VAT works for sales across Europe. What should I look out for?" It contains a location, a sector, a specific problem and a request for advice.
As a result, the search intent in a prompt is usually much more explicit. Prompts are also conversational: the user asks a follow-up question, refines their wishes or asks for a comparison. So the answer can still change over the course of that conversation.
When an AI assistant answers a prompt, it sometimes names businesses, products or sources. Whether your business is among them largely determines your AI visibility. Because every prompt is different and answers vary, there's little point in testing one question once. You only get a reliable picture when you track a set of representative prompts repeatedly.
Start with the questions your customers actually ask you. A practical approach:
A bakery in Leeds, for example, might track prompts such as "where can I buy a gluten-free birthday cake in Leeds?" or "which bakery in Leeds delivers breakfast hampers to businesses?".
Bear in mind that you never know exactly which prompts people type. AI platforms don't share that data the way Google does with search queries in Search Console. Your prompt list is therefore always an informed estimate, based on customer conversations and search behaviour.
In many AI search environments, a prompt isn't run literally as a single search. The system often splits the question into several sub-searches, a technique known as query fan-out. Content that clearly answers those individual sub-questions therefore has a better chance of being used as a source.
If you want to measure systematically which answers your business appears in, it helps to track your visibility in AI answers using a fixed set of prompts, so you see changes over time instead of one-off snapshots.
Request your free AI visibility plan. An expert analyses your business and contacts you within 1 working day to present your plan in person.