Artificial intelligence in real estate: what it really does today and what is just hype

There is a lot of excitement around artificial intelligence in real estate, and just as much confusion. Some promise an agent that AI will fully replace, others dismiss it as a fad that will fizzle out within a year. The truth is far less dramatic and far more useful: AI is an excellent assistant for routine work today, one that saves agents hours and raises the quality of a presentation — but it is nowhere near a standalone player, and for some tasks it is still more of a marketing promise than a practical tool.

The aim of this article is neither to praise AI nor to dismiss it. It is an overview across the whole of estate agency work — from preparing a listing to valuation, communicating with clients, admin and finding buyers. For each area we will try to fairly separate three things: what AI genuinely does well today, where it has clear limits, and where it is still more of a conference-stage demo than something you can rely on in a real sale.

If you are interested purely in the marketing side and concrete time savings, we have a separate piece on that: AI in real estate marketing. This article is broader and deliberately more critical — it is precisely the distance from inflated promises that separates the agent who uses AI wisely from the one who trusts it blindly.

Presenting the property

This is where AI is strongest and most practical today. Not by chance: preparing photos, videos and text is routine, repetitive and time-consuming work, exactly where a computer excels and where a mistake endangers no one — at worst you discard the output and make a new one.

What it does well

Generating video from photos is among the most useful things AI offers in real estate today. From a gallery of images it assembles a smooth video tour with transitions, music, labels and, if you like, a satellite flyover of the location. A few years ago that meant a videographer and a day of work; today the video is ready in minutes. This is exactly what the ELIDAT tool is built for — upload the photos, enter the address, and get a finished video without knowing how to edit.

Virtual home staging and AI interior visualization are the second strong discipline. You furnish an empty flat digitally, add décor, change the style of a room. A buyer orients themselves better in a furnished space and more easily imagines living there. We cover the differences from physical staging and when each pays off in the article home staging vs. virtual staging.

Editing and brightening photos is routine for AI today. It straightens the horizon, lifts dark corners, unifies the colour tone across a set, removes small distracting elements. The result looks professional even with phone photos. Still, AI does not fix bad composition and does not shoot for you — you have to do the groundwork yourself, as described in the guide how to photograph a property for sale.

Generating listing descriptions is the fourth area where AI genuinely saves time. From a few parameters it produces a structured, readable text with a headline and paragraphs. You have a finished draft to refine, instead of writing from scratch.

Where the limits are

A visualization must stay a visualization. A virtually furnished flat is a marketing aid, not a promise about the property's condition. The buyer must clearly know they will not actually receive the furniture in the photo, and the image should be labelled as such. Likewise, AI must not conceal a defect — retouching a damp wall or a crack crosses a line and, at worst, becomes a legal problem.

With text, there is a risk of blandness. Given a thin brief, AI spits out a description that sounds exactly like hundreds of others — all "sunny layouts" and "quiet locations". That is precisely why the finished text is only a half-product. The concrete details, the atmosphere and what makes the property different — those you have to add. On turning a description into something that sells, there is a separate guide how to write a property listing.

Valuation and market analysis

Automated price estimates — AVMs, or automated valuation models — are tempting: enter an address and parameters and you have a number in a second. This is exactly where the greatest critical distance is needed.

What it does well

As a first orientation, these estimates are useful. They quickly tell you the ballpark the price is likely in, and work as a springboard when you are only just entering a new area or want to check whether the owner's expectation is completely off. They can process more data than a person can go through by hand — listings, historical prices, size, layout.

Where the limits are

An automatic estimate does not replace local knowledge, and that is crucial. The model works with what is in the data, but the price is often made by things that are not in the data: south-facing windows, a noisy pub a floor below, a planned development behind the fence, a great neighbour or the reputation of a specific building. Two flats with the same parameters on the same street can genuinely have different prices, and the model will not spot it.

A bad estimate is also worse than no estimate, because it looks precise and confident. A number from a computer tempts you to take it as fact. If an agent adopts it without correction, they risk an overpriced start — one of the costliest mistakes in the entire sale, as we discuss in the article mistakes in property listings. So treat an automatic estimate as one of several inputs, not a verdict. Always set the final price on the basis of your own market knowledge and a viewing in person.

Communication and lead management

Communication is an area where AI can help enormously, but also where it most easily annoys the client. The difference lies in where you deploy it.

Where it helps

Automated replies and chatbots make sense where it is about quick, repeated information outside working hours — confirming an enquiry, sending basic parameters, offering viewing slots. The prospect gets a response straight away instead of waiting until morning, and you don't lose a lead just because you happened to be at another viewing.

Transcribing and summarizing calls is a quiet helper that saves a lot of work. From a recording, AI produces notes, extracts what you agreed and reminds you what to send. Similarly with follow-up: the tool keeps track of who you haven't contacted yet and prepares the draft of a follow-up email.

Where it annoys the client

A chatbot that pretends to be human while spinning a prospect in circles can kill a relationship faster than silence. The moment it comes to specific negotiation, emotion or an unusual question, the bot hits a wall — and the client realises they are talking to a machine that doesn't understand them. Automating communication therefore has a clear rule: it suits first contact and routine, not the core of the relationship. The moment trust and money are involved, a human must take over. And no chatbot should pretend to be one.

Admin and documents

Paperwork is a rewarding discipline for AI — it is linguistic, repetitive and full of templates. Yet this is exactly where the strictest rule of review applies.

What it can do

AI writes drafts of emails and basic texts quickly and decently. It can also summarize a long document into a few points, so you get oriented faster in a report, an extract or lengthy correspondence. For internal contract drafts and standard letters, it saves the first pass, which you then adjust.

Risks

Two risks hang over all of this, and you must not underestimate them. The first is hallucinations — AI can confidently invent a clause, a number or a condition that does not exist. In a legal or financial text this is dangerous, because the error looks credible. Any document that touches a contractual relationship must therefore be reviewed by a human, and for contracts ideally a lawyer. AI is here for the draft, not the final version.

The second risk is data confidentiality. When you upload clients' personal data or sensitive deal details to a public tool, you lose control over where they end up. Watch what you enter where, and follow data protection rules under the GDPR. Sensitive data does not belong in a random online tool.

Search and matching of buyers

Matching demand to supply is an area where AI quietly works in the background and where it makes sense. From a database of prospects and their criteria, it can pull out who a new property might suit and flag it to you before you send the offer out into the world. It speeds up what you would otherwise go through by hand and helps a new offer reach the people who are looking for exactly that.

The limit is the same as elsewhere: the model matches on what is recorded, not on what the client actually wants. A buyer's real decision is shaped by the feeling at the viewing, trust in the agent and things that fit no filter. Matching is therefore a good tip on who to start with — not a list that decides for you.

What is still mostly hype

Some promises around AI in real estate sound tempting but do not yet hold up in practice. It pays to know them so you don't buy a pig in a poke.

• Fully automatic selling and the "AI agent". Real estate rests on trust, local knowledge and negotiation — precisely what AI cannot do. A tool prepares your materials and handles routine, but a human runs the sale.
• Inflated promises of miracle results. When a tool promises it will double sales on its own or replace your whole job, that is marketing, not reality. Ask what it concretely does and where its limits are.
• Blind trust in generated content. Neither text nor visualization is a finished output — it is a half-product to review. Without human editing you risk a factual error and blandness.
• The risk of identical-looking listings. When everyone uses the same tools with the same brief, descriptions and visuals start to resemble one another. You stand out through what you add — a concrete detail, the property's story, your own style.
• Data protection as an afterthought. Convenience must not override the rules. Before you upload client data anywhere, check how the tool handles it.

Where to start

The best way to start with AI is not to overhaul the whole office overnight. It is to pick one routine that slows you down the most and entrust it to AI.

1. Start with presentation. This is where the fastest, safest benefit lies. Take one property and let AI assemble a video from photos or furnish an empty room virtually. You can judge the output easily and risk nothing.
2. Compare time and result. Measure how long the task took by hand and how long with the tool. With presentation, the difference is usually most striking.
3. Add gradually. Once the first step lands, bring in listing text, then call summaries or follow-up. Don't rush admin and valuation — there, let AI only prepare materials that you review yourself.
4. Keep a human in the loop. For anything that touches price, a contract or the client's trust, the agent always has the final word.

You'll find practical examples and procedures in the examples and step by step in the guide. Current prices and packages are on the pricing page. And if you'd like advice on where exactly AI will save you the most, get in touch via contact.

Conclusion

AI is an excellent assistant in real estate today, not a replacement for the agent. It is best at presentation — video from photos, virtual staging, image editing and text drafts genuinely save time and raise quality. With valuation, communication and admin it also helps, but only as support that a human controls. And fully automatic selling or the "AI agent" remain, for now, more of a promise than a reality.

What AI does not replace is exactly what real estate rests on: trust, local knowledge and negotiation. The agent who uses AI wisely gains time for these things — and that is precisely their edge. Start with presentation, where the benefit is greatest and the risk smallest, and expand gradually where you can keep an eye on the output.