"Near me" was the local search phrase of the last decade. It worked because the phone did the location maths: Google's map pack took your position, found businesses in the category within a radius, and sorted them by distance and rating. You never had to name the neighbourhood or say what "good" meant.
An AI answer works differently. A question typed into ChatGPT or Perplexity is read as language. The engine may know roughly where you are, but it builds its answer from what pages say, so the words in the question carry more of the load. People have adapted by saying more: the neighbourhood, the situation, the kind of practice they want.
That changes what your site has to say. This post covers the two basic kinds of prompt, the patterns people use in place of "near me", and what each one needs from your pages.
Descriptive and Prescriptive Prompts
A descriptive prompt asks the engine to explain: "What should I look for in a good optometrist?" or "How do I know if a physical therapist is any good?" The engine retrieves guides and explanations, pages that teach judgment. A page with a section titled "What to expect at your first eye exam" can be retrieved for it. A page that only says "we are experienced" can't.
A prescriptive prompt asks for a business: "Find me an optometrist in Crown Heights that takes my insurance." The engine retrieves named businesses and the facts that prove they qualify: the address, the hours, the services, the plans accepted.
Descriptive prompts reward depth: explanation, steps, criteria. Prescriptive prompts reward specificity: stated facts, ideally also in structured data. The same page rarely does both well.
Five Ways People Phrase It Now
These are the patterns that tend to replace a bare "near me". The example prompts are hypothetical. Each one makes a different demand on your content.
Neighbourhood-explicit. "Optometrist in Crown Heights." "Chiropractor in Bay Ridge, Brooklyn." The neighbourhood replaces the GPS. Your pages need to name it as a real place, with the street, the nearest station and the neighbouring areas you serve, not as a keyword repeated in the copy.
Quality-qualified. "Best independent optometrist in Crown Heights." "A physical therapist in Brooklyn people actually like." With no star rating to sort by, the user asks the model to judge. It can only judge from specifics: credentials, years in practice, the services you're known for, reviews that say something concrete.
Use-case. "My four-year-old needs a first eye exam in Crown Heights. Where should I go?" "Private BJJ lessons for a complete beginner in Brooklyn." These describe a situation. The page that gets named speaks to that situation directly: what a first visit is like, who the service is for.
Relationship. "An eye doctor who takes time to explain things." "A physical therapist who works around my schedule." "Takes a personalised approach" gives the engine nothing. "Each new-patient exam is booked for [length], and the doctor walks you through the prescription" gives it a fact.
Anti-chain. "Family-owned optometrist in Crown Heights, not a chain." If you are independent, say so plainly in the opening copy, the about page and an FAQ answer: "Independently owned and operated on [Avenue] since [year]" is a complete, quotable sentence.
Specificity Compounds Across Fields
Every specific prompt carries constraints: a day, a time, a place, a price, a plan. An engine can only match your business to a constraint you have stated somewhere it can read. Each fact you state precisely raises how specific a question can get before you drop out of the answer.
In markup, the constraints map to real schema.org properties:
openingHoursSpecification: the days and times you're open, per day, not "call for hours"areaServed: the neighbourhoods you serve, namedpriceRange, or Offers with prices on each Service: what it costs- one
Serviceper thing you offer, with a plain description
State the same facts in the page text, where people and engines both read them. Hours in markup plus a visible hours table plus the same hours on your Business Profile is three sources saying one thing. A site that states its hours, its area and its prices can be matched to "eye exam on a Saturday in Crown Heights under [Plan]". A site that states none of them can only be listed under "optometrists in Brooklyn".
How to Tell Which Kind of Prompt You're Facing
Prescriptive prompts carry constraints: "in Crown Heights", "open now", "tonight", "takes my insurance", "near the [line] train". Insurance is the constraint with the most rules attached. For an optometry practice, the insurance page an AI can quote names each plan. For a physical therapy practice it also means Medicare, no-fault and workers' comp, which is what a PT practice's insurance page has to say, and the same routes apply to a chiropractor's insurance page.
Descriptive prompts carry open-ended language: "what should", "how do I", "signs of", "how to tell". They ask for judgment, not a shortlist.
The Same Business Needs Both
A patient researching an eye exam asks both kinds, usually in that order: first "what should I expect?", then "who can see me in Crown Heights on Saturday?" A site that answers only one is missing from half the conversation.
Split the work. Descriptive content lives in explainers and FAQ answers: "What should I ask my optometrist?" Prescriptive facts live in the opening copy, the service and insurance pages, the hours, and the LocalBusiness markup that repeats them.
The common mistake is one sentence that tries to do both: "We're the best optometrist in Crown Heights because we have 20 years of experience and take Medicare." It isn't a stated fact an engine can check, and it doesn't teach anything. Break it up: an FAQ that explains what to look for in an optometrist, and a services page that states the years in practice and each plan accepted.
Where to Start
Write down ten questions customers ask before they book, in their own words. Sort them into descriptive and prescriptive. For each descriptive question, check that a page explains the answer. For each prescriptive one, check that every constraint in it, the neighbourhood, the day, the plan, the price, is stated on your site in plain text and, where it fits, in markup. Then ask ChatGPT and Perplexity those questions yourself and see who comes back.