How AI assistants choose which businesses to name
When a customer asks an assistant for a clinic in their suburb, several things happen before a name appears. This page walks through them, and marks which parts you can influence and which you can’t.
What happens between the question and the answer
Roughly: the assistant works out what’s being asked, gathers material it can reach, decides which of it to trust, then writes an answer naming a small number of businesses.
The details differ between products and change often. What’s been stable is the shape: gather, then filter for trust, then write. Most of the businesses we measure are losing at the second step, not the first.
The parts you can influence
Five, and none of them are restricted for a regulated practice.
- Whether your pages can be read at all. Content that needs scripts to appear, or that lives inside an image, may never be seen.
- Whether the page answers the question directly. A page that states the answer plainly near the top is easier to quote than one that arrives at it in paragraph nine.
- Whether it’s clear who’s speaking. A named practitioner with qualifications and registration, on the page they’re speaking on.
- Whether anything outside your website confirms you exist. Registers, colleges, professional bodies, journals, conference programs, consistent business details.
- Whether your treatments are described in the words customers use rather than the words the industry uses.
The parts you can’t
Four, and no firm can change them for you.
- Which sources the assistant reaches for. It draws on directories, professional bodies, publications and other sites as well as yours.
- What those sources say about you. We can strengthen what they can find and verify; we can’t control what they publish.
- Whether the answer is the same tomorrow. It often isn’t, with nothing changed on your site.
- What the model does next. Products change models and sources, and work that was right in one quarter can need revising in the next.
Being cited and being named are different outcomes
A page can appear in the source list under an answer while the answer text recommends someone else.
They need different work. Being cited is mostly about being readable and reachable. Being named is about authority — there has to be something that says why you in particular, and it has to be verifiable somewhere other than your own site.
We score them separately for that reason, and a review reports them separately.
What this means for a regulated practice
The material an assistant is looking for and the material you’re allowed to use overlap more than you might expect.
Qualifications, publications, memberships and plain explanations of how treatment is delivered are all checkable by a third party, which is exactly the property the system is filtering for. The restricted material — testimonials and before-and-after photos — isn’t checkable in that way, and wouldn’t carry much weight even if it were permitted.
Common questions
- Should we add structured data to our pages?
- It helps a machine parse a page, and it’s on the list of things we look at. It isn’t a substitute for the page saying something worth quoting.
- Does blogging more help?
- Volume on its own doesn’t. A handful of pages that answer real questions, under a named practitioner, do more than a pile that don’t.
- Do assistants use Google reviews?
- They may draw on whatever’s publicly available. For AHPRA-regulated practices we don’t request, incentivise or manage reviews either way.
- Our competitor gets named and their site is worse than ours. Why?
- It happens, and we’ve measured it. Usually there’s corroboration about them somewhere other than their own website that doesn’t exist for you.