When someone asks an AI who does this locally, the answer can be your name

Something has changed in how people search, and most businesses have not noticed yet.

People used to type "dental clinic Braga implants" into Google, open six tabs and compare. Now they ask the same thing in plain language to an AI assistant, get three names with an explanation of why, and call one of them.

Anyone doing that never sees a results page at all. You are either among those three names, or you do not exist.

The good news is that this is a new space, still lightly contested, and most of the work needed to enter it is work that should already be done for other reasons.

First, the honest size of this

Worth starting here, because the topic attracts a lot of exaggeration.

In absolute volume, it is still small. Traffic sent to websites by AI assistants sits, on average, below 2% of total referral traffic. Anyone telling you Google is finished is selling you something.

What matters is the direction and the quality. A thirteen-month analysis published by Search Engine Land in February 2026 found that this traffic tripled between January and December 2025.

And it converts far above average. An analysis by Microsoft Clarity across more than 1,200 sites recorded a 1.66% conversion rate for visitors arriving from AI assistants, against 0.15% for those arriving from traditional search.

A warning is needed here that almost no article gives: published figures on this subject vary considerably, and most come from companies selling monitoring tools. Some say ChatGPT accounts for 84% of this traffic, others say 56%. Treat exact percentages with suspicion and keep the pattern, because the pattern is consistent: low volume, high intent.

The explanation is simple. Someone arriving that way has already asked a specific question, already read a considered answer, and already chosen to click. They are not browsing. They are deciding.

Why this does not work like Google

This is the point that changes how you work.

On Google there is a position. You are first, fourth, or on the second page, and that position is reasonably stable.

In an AI assistant there is no position. There is an answer generated one at a time, which can change depending on how the question is phrased, which assistant is used, and what day it is. Two people can ask the same thing and get different lists.

That produces a situation which catches many people out: companies ranking well on Google can be entirely absent from AI answers, while smaller competitors appear because they have clearer content and more third-party corroboration.

What counts is no longer domain authority but something else: being citable. And being citable means information about you is easy to find, consistent across places, and answers questions directly.

The test you can run today, in ten minutes

You do not need to buy any tool to find out where you stand. You need to ask questions.

Open two or three different AI assistants and ask the questions one of your customers would ask. Do not ask about your company by name, because that tests nothing. Ask by category:

Repeat each question with different wording, because answers vary. Note three things: whether you appear, who appears alongside you, and where the answer drew its information from when the assistant lists sources.

That last point is the most useful of all. The sources cited tell you exactly where you need to be present.

What makes a business citable

After the test, the work concentrates on five things. None of them is a trick, and all of them benefit the business even if AI did not exist.

Consistent information everywhere. Your company name, address and phone number should be exactly the same on your website, your business profile, social media, and any directory you appear in. When a system finds three different addresses for one company, it does not end up with three options. It ends up without enough confidence to recommend you. This is by far the most common error and the easiest to fix.

Third-party corroboration. A website saying you are good counts for little. What supports a recommendation is being mentioned elsewhere: sector directories, associations, local press, listings, job boards, reviews. It does not take much, it takes existing and being coherent.

Content that answers questions directly. Pages organised around a question with the answer in the first line are far easier to cite than flowing marketing copy. That applies to a frequently asked questions section, to service pages, and to articles.

Explicit service and location. If you work in one city and build online stores, that needs to be written in plain text, not implied in an image or a video. It sounds obvious and it fails often.

Genuine reviews. They remain one of the most used signals for choosing between equivalent options.

How to measure without spending anything

Two free things, both taking minutes.

In your analytics, create a filter isolating visits whose source is an AI assistant domain. You then see that traffic separately instead of having it buried inside "referral" or "direct".

And add a "how did you find us" question to your contact form, with an option for AI assistants. It is low technology and it captures exactly what referral data misses, because many people read the recommendation, never click, and type the company name straight into their browser.

How much to invest in this

Little, and for now without budget urgency.

At under 2% of referral traffic, it would be unwise to redirect spend away from traditional search or advertising. Anyone proposing a monthly retainer solely for this is getting ahead of the market.

What does make sense is different and cheap: run the test, correct the information inconsistencies, reorganise your main pages so they answer questions, and start measuring. That is a handful of hours which remain worthwhile even if this channel stalls, because all of it also improves traditional search and the conversion rate of your site.

The advantage of acting now is not today's volume. It is that your competitors are not there yet.

How Digito helps

We start with the diagnosis, because without it you do not know what to fix.

We run the test with the questions your customers would ask, across several assistants, and record where you appear, who appears in your place, and which sources are being cited. Then we check the consistency of your company information everywhere it exists, which is almost always where the problem sits.

From there we handle what is needed: reorganising service pages to answer real questions, correcting profiles and listings, creating content that can be cited, and leaving measurement running so you can track how it develops.

And we tell you frankly how much this is worth in your particular case, which is sometimes a lot and sometimes very little.

If you want to know whether you appear in the answers when someone looks for what you do, we will find out with you.

Sources