Anaks Labs
Article 01

Your patients now ask AI first

Something near the start of a patient's search changed, and it changed quickly enough that most practices have not been given a reason to notice. This article is about what two published surveys actually counted, what they did not, and which part of it lands on your front desk.

Published · Anaks Labs

The first move used to be a search box

For twenty years the shape of this was fixed. Somebody needed a dentist, opened Google, typed a few words with a place name in them, and looked at a list. Everything a practice was ever told to do about the internet followed from that shape: be on the list, be high on it, have a page worth clicking when somebody arrived.

There is a second shape now, and it does not produce a list. Somebody describes their actual situation in a sentence — a molar that cracked on Sunday, a child who will not sit still for a cleaning, a plan they cannot work out whether it covers implants — and something goes off, reads, and comes back with an answer. Occasionally that answer names two or three practices. Occasionally it names none, and explains what to look for instead.

Both shapes are in use at once. Neither has replaced the other. But the second one is growing at a rate that is unusual for anything to do with consumer habits, and it is worth knowing by how much.

What was actually counted

A quarter of US adults — roughly 66 million people — have asked AI for health information or advice.

West Health–Gallup Center on Healthcare in America. 5,660 US adults aged 18 and over, surveyed 27 October – 22 December 2025, published April 2026. Source and sample →

Forty-five per cent of US consumers asked AI to recommend a local business, against six per cent a year earlier. Among adults between thirty and forty-four — the band most likely to be booking the whole family's appointments — it was sixty-four per cent.

BrightLocal Local Consumer Review Survey 2026. 1,002 US adult consumers. Source and sample → · The age-band figure →

Over the same year, the share of consumers reading reviews on Google fell from eighty-three per cent to seventy-one.

BrightLocal Local Consumer Review Survey 2026. 1,002 US adult consumers. This measures where people read reviews, not search generally. Source and sample →

A seven-fold rise inside twelve months is the number to sit with. Consumer behaviour does not usually move like that, and a business plan built on the assumption that it will settle back down is a bet, not a plan.

What none of those figures says

We would rather narrow our own numbers than let you read something into them that is not there, so:

  • A quarter of adults asking AI about health is not a quarter of adults choosing a clinic that way

    The Gallup question was about health information and advice. Somebody asking what a crown is made of is inside that quarter. So is somebody preparing questions for an appointment they have already booked. It tells you the habit exists at scale; it does not tell you it ends in a booking.

  • The forty-five per cent is about local businesses in general

    Restaurants, plumbers, gyms and dentists all sit inside it. We have not seen a clinic-specific version of that number with a publication behind it, so we do not present one.

  • The review figure is not a statement about Google's decline

    It counts where people go to read reviews of a local business. It is not search market share and we do not use it as one. Google is not going anywhere.

  • There is no honest conversion figure here, because we could not find one

    How often an AI conversation about a local business ends in an actual booking is exactly the number a clinic owner wants, and every version of it we found traced back to a blog quoting another blog. Our rule is that a figure with no linkable original comes off the page rather than getting hedged, so it is not on this one.

Every figure on this site, with its sample and its original publication →

Three things that reach a practice

Strip out the forecasting and there are three concrete differences between the old shape and the new one, and all three are things you can observe rather than things you have to believe.

  • The shortlist is now assembled before anyone contacts you

    In the old shape, a patient saw ten results and did the narrowing themselves, on your website, with your photographs in front of them. In the new one, a great deal of that narrowing happens in a conversation you are not present for, and the practice that is named has already been chosen by the time a human being looks at anything.

  • The answer is assembled out of text somebody could read, not out of how anything looked

    An assistant summarising your practice is working from sentences. Whatever it says about your hours, your services or your prices is a paraphrase of words it was able to obtain. The photography, the colour, the typography and the ten seconds of drone footage over the parking lot are not part of what it received. That is not a criticism of any of those things; it is a statement about what a summary can be made of.

  • There is no position report for it

    You can look up where you rank on Google for a phrase. There is no equivalent for "what the assistant says when somebody asks for a dentist near me", because the answer is not fixed: it varies between people, between products, and between one Tuesday and the next. This is the part practice owners find hardest, and it is genuinely hard rather than a gap somebody is about to fill with a dashboard.

What has not changed at all

Quite a lot. Seven in ten consumers still read reviews on Google. The neighbour who recommends you over a fence is still worth more than any of this. And the ordinary fundamentals of search — a site that loads, pages that can be found, a practice that is where it says it is — still decide whether you are in the pool that anything, human or otherwise, is drawing from.

That is the part worth being clear about, because the industry selling AI services has an incentive to describe this as a replacement. It is an addition. The work that made you findable in the old shape is now the entry fee for the new one.

Where to look next

The useful first move is not to read more about the trend. It is to find out what is already being said about your practice, and what your own website is currently handing over to whatever is doing the reading. Both are things you can check yourself in an afternoon.

Read next

We opened 182 US clinic websites to find out what an assistant receives when it goes looking. The findings are on the clinics page → · how the 182 were drawn, and the limits →