AI search and a small business website

Assistant-driven search is real, it is concentrated on a small number of phrases, and in the Australian data gathered for this site it is much smaller than the volume of discussion about it implies. This page states what was measured, then what a build should actually do.

What was measured about AI search and GEO demand

Two datasets, both gathered on 31 July 2026 against Australian location settings, both registered in this repository.

Dataset	Terms	Returned zero	Returned a figure
Web design keywords	677	584	93
Buyer questions	74	60	14

Eighty-six per cent of the keywords and eighty-one per cent of the questions measured no assistant demand at all. That is the finding, and it is the opposite of the impression the topic usually creates.

Where demand does exist it is concentrated. The head terms website design and web design carry the bulk of it. Among questions, how much does a website cost in australia and what is a wireframe carry most of the rest.

Three cautions about that data, stated because they are real

One figure is an outlier and is not used here. A single term in the question dataset returned a figure an order of magnitude above every other measurement and is flagged in this repository as possibly not Australian in scope. No plan on this site depends on it and it is not quoted as an Australian number anywhere, including here.

Some phrasings return identical figures to their head term. A malformed variant returning exactly the same number as the phrase it derives from suggests normalisation inside the measurement rather than two independent readings. Where that pattern appears, the variant is not counted separately.

Assistant volume and Google volume are different measurements. They come from different sources, they count different behaviour, and they must never be added together. A phrase can be strong in one and absent from the other. In this project’s data several phrases carry real assistant volume and no Google volume at all, which is the signature of a question people ask an assistant rather than type into a search box.

What changes for search visibility and SEO, and what does not

Almost nothing about the build changes. The things that make a page legible to an assistant are the things that already make it legible to a search engine and usable by a person — which is also why “generative engine optimisation” (GEO), the label attached to this topic, is mostly traditional SEO wearing a newer name rather than a genuinely separate discipline with different underlying signals.

Practice	Why it matters to a machine reader
Text as text, not inside images	An image of a price list contains no price
Accurate heading structure	The hierarchy is how a passage’s scope is understood
Facts stated in one sentence	Extractable without inference
Tables for comparisons	The relationship is explicit rather than implied
Structured data that matches the page	Facts stated unambiguously
Dates and sources on claims	A statement that can be attributed

Every row is already on the build-side list on what a new website owes search engines. None of them is new work, and none of them requires a separate engagement — the underlying signals an assistant reads are the same ones a search engine has always read.

The one genuine emphasis change is toward definite, self-contained sentences. A paragraph that only makes sense after reading the two before it is harder to lift accurately than one that states its fact completely. That is a writing habit, not a technology.

What cannot be promised

There is no submission process for an assistant, no verified position report, and no stable ranking to track. The systems change without notice and different assistants cite different sources for the same question.

So a supplier who offers to get a business cited by an assistant is offering an outcome they do not control and cannot demonstrate afterwards. Treat that the same way as a promise about a search position: ask what happens if it does not occur, in writing. The general position on promises about outcomes is on the boundary between a build and SEO.

There is a second, quieter problem. Even where an assistant answers a question well, it frequently answers it without sending anyone anywhere. A business measuring success by visits will see nothing from an answer that satisfied the person completely.

What a small business should actually do

Do the build work properly. The six-row table above, on every page. It is included in a competent build.

Be the source of a fact worth citing. Specific, checkable, local statements are more liftable than general ones. An Australian price band with its GST treatment stated is a citable fact; “affordable web design” is not. That principle is why this site publishes its measurement method at all — see how this site measures things.

Keep the listing and the site consistent. Systems assembling a picture of a business use both, and they disagree more often than owners realise. That is on the business listing.

Do not restructure the business around it. On the measured Australian evidence above, the demand does not yet justify it for most small businesses.

When to revisit

This page reports a measurement taken on one day in July 2026, and it is the kind of measurement that will change. The honest position is that it should be re-measured rather than reasoned about, and that any figure quoted from it should carry the date it was taken.

If the proportion of Australian buyer questions returning assistant volume rises substantially from the eighty-one per cent that measured zero, that is the signal to revisit — not a change in how much the topic is discussed.

What to do next

Do nothing differently on the strength of this page except the six rows in the table, which you should be getting anyway. If a proposal arrives quoting assistant statistics, ask where the figures were measured, in which country, and on what date. Most cannot answer. The rest of the build-side obligations are on the search readiness guides, and the commercial scope of a build is on website design services.