Type "AI" into any proptech vendor's homepage search bar and you'll get more hits than you have time to read. Site sourcing, valuations, lease drafting, tenant comms, planning risk — there isn't a corner of the property lifecycle that hasn't had "AI-powered" bolted onto it somewhere in the last two years. JLL's 2025 Global Real Estate Technology Survey found that the number of corporate real estate companies running AI pilots jumped from 5% to 92% in just three years, so on paper, the industry looks like it's sprinting toward an AI-first future.
Scratch the surface, though, and adoption tells a more cautious story. In the 2026 AI in Real Estate Survey, run by Remit Consulting with the UK PropTech Association, 93% of respondents said they had access to an AI service, but only 7% considered it fully integrated into how they actually work. That gap between "we have the tool" and "we trust the tool" is the whole story of AI in proptech right now, and it's worth understanding before you add another one to your stack.
Key Highlights
- Real estate has near-universal access to AI tools, but integration lags badly behind, with only 7% of UK real estate professionals describing AI as fully embedded in their workflow.
- "AI-powered" is not a meaningful differentiator on its own. What matters is whether a tool's output is verifiable, and how it was built.
- Even purpose-built professional AI tools hallucinate. A Stanford study found error rates of 17-33% in commercial legal research platforms that had marketed themselves as "hallucination-free."
- RICS has introduced its first global professional standard on responsible AI use in surveying, putting professional judgement and client transparency at the centre of any compliant use of AI.
- Good AI in proptech should reduce, not remove, the need to double-check its own work, and it should make checking easy rather than optional.
"AI-Powered" Doesn't Actually Tell You Anything
Part of the confusion is that "AI-powered" gets used to describe wildly different pieces of engineering. Sometimes it means a purpose-built model trained on structured, verified data. Sometimes it means a general chatbot with a property-shaped prompt wrapped around it. Sometimes it means a plug-in that hands your query off to an external AI model entirely outside the platform you're using, via something like an MCP connection. All three get marketed with the same three letters, and the differences between them matter far more than whether the feature list includes the word "AI" at all.
That distinction becomes urgent once you consider what happens when an AI tool is confidently wrong.
When AI Gets It Wrong, It Gets It Wrong Confidently
The clearest illustration of this risk doesn't come from proptech at all - it comes from a New York courtroom. In 2023, two lawyers representing Roberto Mata in a personal injury claim against Avianca filed a legal brief citing six court decisions that simply did not exist. They had asked ChatGPT to find supporting case law, and the model invented plausible-sounding citations, complete with fabricated quotes and case numbers. When challenged, the AI even insisted the fake cases were real. The presiding judge fined the lawyers and their firm $5,000, and the case became the reference point for an entire industry's caution around unverified AI output.
You might reasonably assume that professional-grade tools, built specifically for high-stakes work, don't have this problem. The data says otherwise. Researchers at Stanford's RegLab ran the first independently verified, preregistered study of commercial legal AI research tools, several of which had been marketed as producing "hallucination-free" citations. Hand-scoring the tools against real legal queries, the study found hallucination rates of between 17% and 33% - meaningfully better than general-purpose chatbots, but nowhere near the "solved" claim being sold to paying professionals.
The lesson isn't that AI is unusable for professional work. It's that marketing claims about accuracy are not the same thing as verification, and the two get confused constantly.
What the Profession Is Already Saying About This
Property isn't waiting for someone else to work this out. In March 2026, RICS introduced its first global professional standard for the responsible use of AI in surveying practice, covering valuation, construction, infrastructure, and land services. The standard is built around a few core principles: firms must maintain governance and risk registers for any AI use with material impact, clients must be told in writing when and how AI is being used, and surveyors are required to apply professional scepticism and remain accountable for outputs they rely on, no matter how confident the AI sounds.
That last point is important. The standard is explicit that AI can support a surveyor's judgement, but it can't replace it, and the professional signing their name to a piece of work remains responsible for it. Whatever sector of property you work in, that's a reasonable bar to hold any AI tool to, whether or not you're bound by the RICS standard directly.
So What Does Good AI in Proptech Actually Look Like?
Pulling the threads above together, a handful of principles start to emerge for what separates a genuinely useful AI tool from a confident-sounding liability.
It Shows Its Working
If a tool tells you a site sits within a flood zone, or that a particular policy applies, you should be able to click through to the actual source and check it yourself, in seconds. Verifiable, source-level citations turn an AI's output from a claim you have to take on faith into a claim you can stand behind.
It's Anchored to Trustworthy Data
General-purpose AI models are trained on the open internet, which is a poor foundation for anything involving current planning policy, since that policy changes and the model's training data doesn't. A tool that's anchored exclusively to a maintained, institutional-grade dataset, kept current centrally, starts from a fundamentally different accuracy floor than one reasoning from whatever it happened to learn during training.
It Doesn't Require You to Cecome a Professional Prompt Engineer
If getting a usable answer out of a tool means learning specialist prompting technique, the effort can end up outweighing the benefit, particularly for anyone already stretched across a demanding pipeline. Planning logic that's already been built into the tool by people who understand the domain means you can describe your project in plain English and get something workable back.
It Stays Editable
Site data changes, schemes evolve, and a flat, static first draft that has to be rebuilt from scratch every time your requirements shift isn't much of a time-saver. Good AI output should be something you can refine in place, iterating as your thinking develops.
It Knows What it's for
No AI tool currently on the market should be trusted to sign off a multi-million pound acquisition on its own, and any tool implying otherwise should raise a flag. The honest use case for AI in property right now is accelerating the early, high-volume stages of the process, triaging sites and drafting planning documents, so that the professional judgement in your team gets applied to the decisions that actually need it, rather than spent confirming basic constraints.
None of this means AI should be flawless before you'll touch it. Nothing eliminates the need to double-check important work, AI included. The realistic bar is whether a tool radically reduces that checking burden by making verification fast and built-in, rather than something you have to do independently, from scratch, every time.
Try It For Yourself
LandTech's AI Assistant was built around exactly this standard. It's native to LandInsight, anchored exclusively to LandTech's own institutional-grade property data, and every planning document it generates carries source-level citations straight back to the record it came from. Tell it your project's goals in plain English, and it matches them against a site's real constraints, no separate AI subscription and no prompt engineering required.
If you're on LandInsight Unlimited, the AI Assistant is already live in your account. Head to your Saved Sites tab and try it on a site you're currently assessing, and see how it holds up against the standard above.
Related reading:
- Introducing the AI Assistant: Triage Sites Instantly with Citations You Can Trust
- AI Land Sourcing Tools Compared: What Actually Matters
- Which Land Sourcing Platform Actually Uses AI to Analyse Planning Constraints?