Review language

Assistants repeat what customers write

When an assistant recommends a business, it almost always gives a reason, and the reason usually comes from reviews. “Locals praise the hand-pulled noodles.” “Patients mention how calm the staff keep nervous children.” “Homeowners say the crew left the site spotless.” Those phrases are condensed from what customers wrote. Bay Area AEO's food and drink page says it directly: reviews that mention specific dishes, the atmosphere and the occasion are phrases AI pulls straight into its answers.

This changes what a good review is. For a star rating, any five-star review counts the same. For an AI answer, a five-star review that says “Great!” adds almost nothing, while a four-star review that says “Best khao soi I've had outside Chiang Mai, and they handled our group of eight without a reservation on a Friday” gives an assistant a dish, a quality and an occasion.

What useful review language contains

Useful reviews tend to mention one or more of the same things that make up a customer's question (see chapter 1): the service, the place, the condition, the quality and the occasion. Bay Area AEO's niche pages show what this looks like trade by trade.

Reading your reviews the way an assistant does

Before trying to change anything, read the business's recent reviews as an assistant would. Take the last fifty across Google, Yelp and the main trade platform. For each one, note the specific things it mentions: services, products, staff names, neighborhoods, occasions, qualities. Ignore general praise. Then count.

The result is a picture of how the business is described in its customers' words. A dental practice might find that patients often mention the hygienist by name and the short wait, and rarely mention implants, even though implants are the service it most wants to be known for. A brewery might find that reviews talk about the dog-friendly patio and never mention the food trucks. That gap between what the business wants to be recommended for and what customers write about is the target for the next two chapters.

Comparing with the AI answers

Then compare the review picture with the descriptions in the audit answers from chapter 7. Very often the assistant's description of the business lines up closely with the commonest review themes. If the assistant describes a restaurant as “a casual spot for lunch”, look for reviews that say exactly that. If a competitor is described as “known for gentle care with anxious patients”, look at its reviews and see how many patients used those words.

This comparison is the most direct evidence a business can gather about how review language feeds AI answers, and it costs nothing but an afternoon.

What a business can and cannot do

A business cannot write its customers' reviews, and should never try. It cannot offer rewards for positive reviews, script what customers say, or post reviews under staff or family names. In the United States, the Federal Trade Commission's rule on fake reviews, finalized in 2024, prohibits fake reviews and the buying of positive reviews, and the major platforms remove reviews and can penalize profiles that break their own rules. Bay Area AEO's stated principle is plain: the agency never uses fake reviews, spam or tactics that could backfire.

What a business can do is shape the moment and the prompt. It can ask at the right time, when the customer has just experienced the thing the business wants to be known for. It can make the request specific enough to remind customers what they might mention, without telling them what to say. And it can make the service itself memorable in the ways it wants to be described. Chapter 32 covers the request in detail.

Making the service worth describing

The most reliable way to get specific reviews is to give customers something specific to describe. A roofer whose crew sends a photo of the finished ridge and explains the fire rating of the material has given the homeowner a detail worth writing. A dentist who explains each step to an anxious patient has given that patient a story to tell. A caterer who labels every dish with its dietary tags at a corporate lunch has given the office manager a reason to mention it. This is ordinary good service, and it shows up in the language.

Staff names and specialties

Reviews that name staff carry particular weight for practices where customers choose a person. Bay Area AEO's dentist page notes that review requests can invite patients to mention staff by name, which AI repeats in its answers. When patients write “Dr. Nguyen did my implant and explained everything”, an assistant can connect that dentist, that treatment and that quality. Make staff names easy to remember: name badges, introductions, the name in the follow-up message.

Places in reviews

Customers often mention where they live or where the job was. Those mentions help an assistant place the business, which chapter 21 covers. A painting contractor with reviews from homeowners in Berkeley and Albany has a geographic footprint written by its customers, which is stronger than any service-area list it could write about itself.

Volume, recency and spread across platforms

Language matters most, but volume and recency still count. A business with a handful of reviews, the newest two years old, gives an assistant little confidence that the description is current. A steady flow of recent reviews signals an active business and keeps the language fresh. Spread across platforms matters too, because different assistants read different platforms: Google for AI Overviews and Gemini, Yelp for Apple Maps, the trade platforms for professional questions.

Regulated trades

Some trades have rules about reviews and testimonials. Health practices must protect patient privacy, which shapes how they ask and how they reply. Regulated professionals may have restrictions on how testimonials are used in their own advertising. Bay Area AEO's chiropractor page phrases the goal carefully: reviews that mention the condition and outcome in the patient's own words, gathered compliantly. Check the rules for the profession before designing a review program, and keep everything in the customer's own voice.

Older reviews still count

A business with years of reviews has a record that assistants read as a whole. Older reviews that describe a service the business no longer offers, or a location it has left, can still shape answers. The business cannot remove honest reviews, and should not try. What it can do is make the current picture clear: reply to older reviews where a reply adds useful context (“We moved to Grand Avenue in 2023; the same team is here”), keep the profile's services and description current, and let a steady flow of recent reviews describe the business as it is now.

Measuring progress

Repeat the review count every quarter: the share of recent reviews that mention a specific service, a staff name, a place or an occasion, and which themes are growing. Set it beside the monthly AI answers. When the themes the business cares about start to appear in reviews, they tend to follow in the descriptions, and the monthly test is where that shows up.

Related chapters32. Asking for reviews · 33. Responding to reviews · 4. The five signals · 48. Reading AI answers