Somebody looking for a plumber at nine at night is now about as likely to ask an assistant as to open a search engine, and the assistant comes back with three names instead of ten blue links.
Nobody has trustworthy numbers on how much of this is happening yet, and anyone quoting you a precise share is guessing. What's not in doubt is that being one of the three named is worth more than being fourth on a page of ten.
Where the answer comes from
An assistant answering a question about a business is doing one of two things. It's either recalling what it absorbed during training, or running a search and reading the pages that come back. Most now do the second for anything local or current, because the first goes stale.
Both reward the same thing: pages that state facts plainly in text, and facts that agree with each other across several independent places. The assistant is deciding whether it's safe to put your name in an answer, and corroboration is how it decides.
Which makes this a stricter version of the old discipline instead of a new one. Search setup that was already done properly is most of the work.
What makes a business easy to recommend
- Facts written as text. Hours, service areas, prices and services as words on the page. Anything baked into an image, a PDF or a video might as well not be there.
- Specifics instead of adjectives. Open 24 hours beats always available. Serves Whitby, Ajax and Oshawa beats serving Durham and beyond.
- A page per service. A page that answers one question thoroughly gets quoted. A services page listing nine things in a row gets skipped.
- The same name, address and phone number everywhere. Your site, your Google profile, and every directory still carrying you.
- Reviews with sentences in them. A rating on its own says very little. A customer describing what happened is what an assistant paraphrases back.
Structured data does the boring half
Schema is a block of machine-readable facts in the page source: what the business is, where it works, what it offers, what it's rated. A visitor never sees it, and a machine reading it has to infer nothing from the layout.
It's not a ranking trick and it won't rescue a thin page. What it removes is ambiguity, and ambiguity is why a business gets left out of an answer it should have been in. It's part of every build here and it's the substance of AI search optimization.
Corroboration beats optimization
One page saying you're open on Sundays is a claim. The same hours on your site, your Google Business Profile and two directories is a fact, and an assistant will repeat a fact.
This is the highest-return item on the list and it costs an afternoon. Claim the profile if you have not, get the hours and the categories right, and fix the two directories still showing your old number. Getting on Google Maps covers the profile properly.
What doesn't work
Writing pages aimed at a model. The output reads like it was written for a machine because it was, and these systems are trained on human writing and are unusually good at telling the difference.
Question headings you never answer. A heading is a promise. An unanswered one costs you the paragraph that would have been quoted.
Paying for placement. There's nothing to buy in most assistant answers today. That will change, and when it does it will be labelled the way ad results are labelled.
How to check where you stand
- Ask the assistants directly, in a logged-out or private session so it's not answering from your own history. Use the words a customer would use, including the town.
- Ask a follow-up: where did that come from? Most will cite. Those citations are a list of what your competitors are present on and you're not.
- Run the same question three or four times. Answers vary between runs. One good result is noise.
- Search your own page for your hours and service areas. If you can't find them as text, neither can anything else.
