What does AI ask about when someone is looking for an estate agent?

When you ask AI about an estate agent in your town, you do not get an advert as your answer. You get a summary of what the model can find and confirm publicly in several places. That is the difference, and it explains why some agents appear in the answer while others are not there at all.

AI assistants such as ChatGPT, Perplexity, Gemini and Copilot fetch web sources and weigh the signals they can find about local businesses when someone asks for help finding an estate agent. If the assistant cannot read the website because it is closed to AI crawlers, the agent is not even considered.

SOCi’s Local Visibility Index from January 2026, based on US data, shows that ChatGPT recommends only 1,2 % of the locations included in the data. By comparison, 35,9 % appeared in Google’s classic local results. The field in an AI answer is therefore much narrower than the familiar list of local results.

It depends on which sources the models rely on. Google Maps used to determine local visibility on its own. Now AI models draw more heavily on Trustpilot reviews, entries in the Danish directory Krak, LinkedIn profiles, Wikipedia articles, news articles and the structure of the website itself, while Google Business Profile often acts as a secondary source. It is a broader range of sources, which means it is no longer enough to be visible in one particular place.

So an AI answer about an estate agent is not an advert. It is a summary of what the model can verify publicly in several places. If you are mentioned, it is because the information about you is consistent across sources. If you are not mentioned, it is rarely down to luck.

Which estate agent should I choose: questions an AI can answer

When a customer writes to an AI assistant and asks “which estate agent can you recommend in [town]?”, the answer has to land somewhere. Either the model mentions your business, or it mentions another agent. There is no third option, and the outcome depends on which sources the model can read and confirm about you.

The assistant does not judge who is the most skilled. It picks up what is said clearly and repeatedly about businesses in the area and presents it as an answer. That is why, in practice, the question “which estate agent should I choose” becomes a question of who is readable and credible in the sources the model uses.

Much of what customers ask about is purely practical and can be answered directly from the listing:

• Is the home still for sale?

• What is the asking price?

• When can we view it?

• How big is the home?

• Is there parking?

• Is the price negotiable?

These are the questions that dominate the first contact, and answering them requires no professional judgement.

But that is only part of the picture. Around 20 % of the questions require human expertise, and they are escalated to an estate agent with the full context, so the customer does not have to start from scratch.

And that is the right division of labour. AI cannot replace the estate agent’s professional assessment, negotiation or viewings. It handles speed, volume and administration, while the agent takes care of relationships, trust and professional judgement.

Which sources AI weighs when it recommends an estate agent

The assistants do not all use the same source. Each model has its own base and preferred sources, and that determines where the model even looks for you when someone asks it to recommend an estate agent.

This difference means that an AI recommendation of an estate agent rarely rests on the same basis from one model to the next. If your business is only visible in one place, some assistants will pick you up while others will not.

Next comes the question of whether the information about you is consistent. Your name, address and phone number must be exactly the same across your website, Trustpilot, Krak, LinkedIn and social profiles. Even small discrepancies, a comma or an abbreviation, make it hard for the model to know who the business is. When the fields do not match, the model is left with conflicting information and cannot confirm that it is dealing with the same agent. Consistency is therefore not just a matter of tidiness. It decides whether the model can confirm you as one business at all.

Reviews carry a lot of weight here. Businesses with a Trustpilot profile were mentioned in 53,5 % of relevant AI answers, and among those mentioned, 75,3 % required more than 80 reviews and ongoing replies to them. The figures show something simple: an empty or quiet profile does not count as a signal. It is the volume of reviews and the sustained activity around them that make the profile readable for the model.

The format matters too. Schema.org markup such as Organization, LocalBusiness and FAQPage tells AI models exactly who the estate agent is, where the office is and which services are offered. The model can read it without guessing, and guesswork usually costs you a place in the answer.

Visible in ChatGPT: technical requirements and quotable content

A scan of 465 Danish websites on 21 July 2026 showed that 5 % had virtually no readable text without JavaScript, and 38 % lacked structured data. Both figures point to the same problem. If a page only makes sense once the browser has run a heavy script, or if the machine cannot see what is where, the assistant is left with very little to work with. That makes it hard to be mentioned in an AI conversation about estate agents, however good the page otherwise looks to a human.

The typical technical blocks are nothing exotic. They crop up again and again:

• A robots.txt file that turns away AI crawlers such as GPTBot, sometimes set up by mistake by a previous agency

• Content that only appears after heavy JavaScript, which the crawler does not always run

• Text hidden in images or PDFs without alt text, which is therefore invisible to machines

The core text should be readable as plain HTML, without scripts having to run to bring it up.

The writing style matters a lot too. Content that can be quoted directly is short and precise. Single sentences that make sense without context are far easier to lift into an answer than a long paragraph where the point only becomes clear at the end. A question-and-answer format helps in the same way, because the model can pick out the individual answer and reproduce it without bringing the rest of the page along. The same goes for concrete descriptions of target group, problem and solution. If it is clear who the page is for, what the problem is and what the solution consists of, there is something to draw on. If, on the other hand, the content is too generic, too sales-driven or too poorly structured, it is hard for AI to recommend the page.

Put your effort in the right places. FAQ pages and interactive calculators are among the page types AI engines use most often when recommending a business in answer to questions close to a purchase. These are the pages where someone is in the middle of a specific decision and looking for an answer, and that is exactly the situation an AI answer is asked to cover.

AI for selling homes: where the assistant helps and where the estate agent is essential

The most concrete example of AI in home selling is the follow-up after an open house. An AI voice agent can call everyone who signed up the same day and ask about their budget, pre-approval and timeframe. If the buyer says yes, the voice agent books a private viewing directly in the calendar, and the lead, together with its transcript, lands in the CRM. This is not a prototype. It is a workflow that is already up and running.

The first questions in a property transaction are the same every time. What is the budget, is the financing in place, and when would the buyer like to take possession? A voice agent can handle that kind of thing because the script can branch according to intent. Buyers are asked about budget, pre-approval and timeframe, while tenants are asked about their move-in date and income instead.

Before calls like these can be used, certain conditions must be met:

• There must be explicit consent, for example obtained through a free valuation form

• It must be clear that this is an AI assistant and not a human estate agent

• Data must be stored within the EU

• The customer must be able to have their data deleted

Those are the ground rules, and they are not negotiable if you want to use AI in the part of the home sale that involves personal data.

Further into the process, there is another type of tool. AI in home selling increasingly relies on automated valuation models such as Geomatic and BoIQ. Geomatic supplies one of Denmark’s most widely used AVM solutions and is used by several Danish banks to price homes, while BoIQ draws data from more than 40 different sources on Danish properties. Both take their data from official registers such as BBR (the Danish building register) and Matriklen (the Danish cadastral register) and weigh parameters such as location, property type, size, energy label and local environment.

At least as important is what the models cannot do. An AI model does not know whether the kitchen is from IKEA or Bulthaup. Condition, views, noise from neighbours and the quality of a renovation are factors that cannot be read from a register, and they are typically the very factors that move a price one way or the other. The valuation models therefore complement the estate agent’s valuation; they do not replace it.

So AI in home selling has a clearly defined remit. The assistant answers the phone on the first ring, asks the same questions every time and makes sure no enquiry falls through the cracks. But it should never show the home, and it should never negotiate the price. Valuation and pricing are and will remain the estate agent’s professional responsibility, and that is not a technical question but a question of being able to stand behind a figure in front of a seller.

Finding the right estate agent with AI: how to bring it all together

The first expectation to adjust when you want to find the right estate agent with AI is that no one can guarantee that ChatGPT or any other assistant will mention a particular estate agent. What you can do is remove the technical and content obstacles so that the model can find and quote the information correctly in the first place. The rest is up to the model.

Timing also plays a part, and it varies. Answers based on live search can change within days or weeks as the website is updated and new mentions appear. Visibility that rests on training data takes months to shift. Knowing this makes it easier to judge when there is something to look at and when you simply need to let the work take effect.

One condition is fixed. ChatGPT, Gemini and Perplexity mention businesses based on public content and sources the models can read, not on payment. You cannot buy your way into being recommended. It is not a limitation that disappears with a bigger budget, and that is why the technical and content measures are the ones that really make a difference.

If you want to know what your home is worth before choosing an estate agent, you can get a free valuation from Brikk. It is a calm starting point, with no obligation and no promises about the outcome.

Also read: Your first meeting with the estate agent: what should you ask?