A growing share of “who should I call” questions are now being asked of an assistant rather than a search box. The homeowner does not see ten blue links. They see a short answer naming two or three businesses, and the businesses that are not named do not get a second chance.
This is a different game from ranking, and most of the advice floating around treats it as the same game with a new name. It is not. Worth being clear about what actually appears to matter.
It is a summarisation problem
A model producing a local recommendation is doing something closer to summarising a body of evidence than sorting a list. It is assembling an answer from what it can retrieve and what it can parse confidently. The practical consequence: being ambiguous is worse than being small.
If your business name, service area, and specialisms are stated inconsistently across your site, your profile, and third-party listings, a model has to guess which version is right. Guessing is expensive, so it tends to reach for a competitor it can describe confidently instead.
What appears to move the needle
- 01Entity consistency. Identical name, address, phone and service-area language everywhere it appears. This is unglamorous and it is the foundation.
- 02Extractable claims. Content that states things plainly — what you do, where, for whom, at what kind of price — in sentences that survive being lifted out of context.
- 03Structured data that matches the visible page. Marking up services, areas and reviews helps, but only when it agrees with what a human would read.
- 04Third-party corroboration. Mentions on sources the model already trusts do more than anything on your own domain, because a business describing itself is weak evidence.
- 05Question-shaped content. Pages that answer the question a homeowner would actually type, in the words they would use, rather than in industry phrasing.
What does not appear to work
- Keyword stuffing, in any of its modern costumes. Models are summarising meaning, not counting terms.
- Thin location pages generated at scale with the town name swapped. These were already weak and they are weaker here.
- Trying to instruct the model in hidden text on your page. It does not work, and it is the kind of thing that ages badly.
- Assuming this replaces local SEO. Map pack and organic still drive the majority of calls in most markets. This is additive.
An honest assessment of the opportunity
Two things are true at once. AI-mediated discovery is still a minority of how home services work gets found, and the businesses establishing themselves there now are doing it against almost no competition. Those are both reasons to act, but they are reasons to act proportionately.
The sensible allocation for most contractors is to keep doing what already works, and to make the specific changes — entity consistency, extractable claims, structured data, question-shaped pages — that improve conventional search anyway. Almost everything on that list pays for itself twice.
The work that gets you named by an assistant is largely the work that should have been done for search. It has simply stopped being optional.
Want this running in your business?
Thirty minutes on the phone. We’ll find where the work is leaking and what it would cost to fix — or tell you it isn’t worth it yet.
Book a discovery call
