Why it matters
Prompts are how customers now ask for a business like yours. They're longer and more specific than old search keywords, often with a budget, a neighborhood, or a worry built in. Each wording gives the AI different text to work from, so you can be named for one phrasing and missed on another.
Thinking in prompts is thinking like your customer. What would a stranger actually type when they need what you sell?
Example
A roofer used to think about one search, "roofer near me." Now his customers type things like, "Who can repair storm damage on an older shingle roof in my area, and do they work with insurance?" That prompt asks about a job type, a roof age, and insurance in one go. The roofer who has a page answering exactly that gets a better chance of being the one the assistant names.
What counts as a prompt
Google's machine learning glossary defines a prompt as any text entered as input to a large language model to condition the model to behave in a certain way. It can be as short as a phrase or arbitrarily long.
The glossary groups prompts into categories, including questions ("How fast can a pigeon fly?"), instructions ("Write a funny poem about arbitrage"), examples that show the model what you want, and roles that tell it who to answer as. Your customers mix these without thinking about it, such as "Act like a local expert and recommend a plumber."
Why prompts matter for a local business
The model answers from the prompt it's given, so how a customer phrases a question changes what comes back. A prompt that names a neighborhood, a price range, or a specific job pulls in different pages than a plain "plumber" does.
For you, the practical move is to write down the questions your real customers ask, in their words, and make sure your pages answer each one in plain text. Then ask those questions yourself, in more than one phrasing, and see who gets named.