Two dental practices sit on the same high street with similar ratings and similar websites. Ask an assistant which suits a nervous patient and the first is named, with a sentence about its sedation options and gentle reviews. The second isn't mentioned. It isn't worse; there is simply less about it that an assistant can read and repeat.
The previous lesson showed what assistants retrieve. This one turns that into four jobs in the order that pays back fastest: Profile, reviews, consistent details, your own pages.
Start with the Profile, because everything else is checked against it
Your Google Business Profile is the record other sources get reconciled against. Google's own Business Profile help says to enter complete information and names the address, phone number, category and attributes as what makes a business easier to match to searches. An assistant uses the same fields to decide whether you are a candidate at all.
Completeness means fields with content, not a tick. Set the primary category to what you most want to be named for, add secondary categories for the rest, and list every service with a short plain description, because a service you never listed is a question you can't be the answer to. Fill in hours and every attribute that applies, and write the description as sentences a person would say: what you do, where, for whom.
Keep it inside Google's guidelines for representing your business: your real name with no added keywords, your real address, one phone number. An assistant that finds "Smile Dental" on your website and "Smile Dental Best Dentist Norwich" on Maps sees two weak entities, not one strong one. Claim and complete your profile walks through every field.
Then reviews that say something, not just stars
Assistants write sentences, and reviews are where the sentences come from. "Great service, five stars" gives an assistant nothing to say. "They fitted me in the same afternoon, explained the fault and the price matched the quote" gives it three things, and it will repeat them almost word for word.
You can't write the reviews, but you can choose the moment and the prompt. Ask when the job is done and the relief is fresh, and ask the customer to say what they had done rather than how they rate you; most people then mention the service, the town and what mattered to them, unscripted. Google's Maps user-contributed content policy forbids paying for reviews and asking only the happy customers, so ask everyone, the same way, every time.
Reply to every review. Replies are public text on the same page, and can restate a fact the reviewer left out ("glad the Saturday slot worked for you"). Getting reviews consistently sets up the routine; Reviews and AI recommendations goes deeper on what assistants pick up.
Then the same details everywhere
When an assistant pulls back five sources it has to decide whether they describe one business, and your name, address and phone number are how it decides. Matching details across Maps, review sites, directories and your own footer confirm a single entity. A different phone number on a directory, a trading name here and a company name there, an old address still live somewhere: each splits you or makes you look unreliable, and assistants are observed to drop a business rather than guess.
This is a one-off audit. Search your name and phone number, list every place you appear, and fix the ones that differ; NAP consistency and citations gives the method and which listings to fix first.
Then pages an assistant can quote
Your website fills the gaps the other sources leave, and an assistant uses it by lifting passages. The question for each service page is whether there is one worth lifting in the first screen.
State the facts plainly and early: the service, the area, who it suits, the terms. "We repair gas boilers across Bristol, including same-day emergency call-outs, and we quote before we start" is a liftable passage; "Welcome to our website" under a hero image is not. Put hours and prices in text, not an image. Use headings phrased the way customers ask ("Do you treat children?") with a direct answer underneath; Google no longer shows FAQ rich results, but the content still serves an assistant. Add LocalBusiness structured data, which Google's local business documentation still supports, so the facts are machine-readable. Service pages that rank and LocalBusiness schema cover the detail.
Two technical points. Make sure your robots.txt isn't blocking the crawlers you want reading you; Google's crawler documentation states that its Google-Extended token controls Gemini training and grounding use and does not affect Google Search. And an llms.txt file is cheap to add, but no major assistant documents reading it, so treat it as a courtesy, not a lever. llms.txt and AI crawlers explains both.
Why this order
The Profile comes first because it is the reference record and takes an afternoon. Reviews come second because they supply the reasons assistants give and accumulate slowly, so start early. Consistency is third because one audit makes the first two count. Pages come last because they take the most writing and only fill what the other sources leave.
Resist jumping to schema or llms.txt first; they are finishing touches on a business that is already readable, and do little for one that isn't.
What to take away
- A complete Profile, with real categories, listed services and a plain-sentence description, decides whether an assistant treats you as a candidate at all.
- Reviews that describe what was done give assistants quotable reasons; ask everyone at the moment the job ends, and never pay or filter.
- Matching name, address and phone across every listing confirms you as one entity instead of several unreliable ones.
- Service pages need a plain, liftable passage in the first screen, with crawlers allowed in and structured data as a finishing touch.
Next
With the four jobs under way, Tracking your AI visibility locally shows how to check what assistants say about you without fooling yourself.