A board deck lands in your inbox. Slide four announces that almost every multifamily operator is using or planning AI, and the figure sits there in 40-point type, unattributed.
You go looking for the source. A trade article cites a consultancy, which cites a blog post, which cites a survey run by a company that sells conversational AI to multifamily operators. Sample size undisclosed. Question wording unpublished.
The figure might be right. There is no way to check, and the deck goes to your investment committee on Thursday.
This article covers what the verified data says about AI in US real estate, where the widely quoted operator numbers come from, and how to read an adoption statistic before you repeat it.
What the 2026 Data Actually Shows About AI in US Real Estate
Two US datasets will tell you exactly how they were built. Start there, because that is rarer than it should be.
The National Association of REALTORS put its 2025 Technology Survey to a random sample of 49,233 members. The finding is a split, with 20% of agents using AI daily, 22% weekly, 27% a few times a month, and 32% not using it at all.
That last figure deserves a moment. Three years into the most publicised technology launch of the decade, a third of US agents have not touched it.
Now look at what the other two-thirds reach for. ChatGPT leads at 58%, with Gemini at 20% and Copilot at 15%, and 46% apply AI to content such as listing descriptions.
Every one is a general-purpose assistant with a text box. The most rigorous AI adoption data in US real estate is measuring people typing into ChatGPT.
PwC and the Urban Land Institute come at it from the capital side, drawing Emerging Trends in Real Estate 2026 from more than 1,700 investors, developers, lenders and advisers across the US and Canada.
Both name their population, sample size and fieldwork date. Keep hold of that, because it is the standard the next section takes away.
Why Operator Adoption Statistics Are Harder to Pin Down Than They Look
Read the NAR findings again and notice who is in the sample. Realtors, meaning residential sales agents running transactions between buyers and sellers. That is not your leasing team.
An agent using ChatGPT to draft a listing description and a multifamily operator running conversational AI across 4,000 units are doing unrelated work:
- Different job: the agent markets a property to sell it once. The operator answers hundreds of enquiries a week and books tours against live availability.
- Different technology: a general-purpose assistant with a text box, against a system wired into a PMS that reads inventory and writes guest cards.
- Different measure of success: the agent wants a better paragraph. The operator wants a signed lease.
So when a headline announces that real estate has adopted AI, check which real estate it means. The rigorous US datasets cover agents and investors. Neither measures whether a BTR operator in Dallas runs an AI leasing assistant. That gap is where the vendor numbers rush in.
Where the Widely Quoted Adoption Numbers Come From
Trace an operator statistic back and the chain usually runs the same way. A trade publication reports it. The trade piece cites a consultancy's guide. The guide cites a vendor blog. The blog cites the vendor's own executive survey, where the number was born.
By the fourth link it reads as industry consensus. It is one company's marketing research wearing a suit.
Vendors are entitled to run surveys. The gap sits in what those surveys leave out:
- The sample: how many operators answered. Twelve executives and 1,200 executives carry different weight, and an undisclosed sample size means you cannot tell which you are reading.
- The selection: whether respondents were drawn at random or invited from the vendor's own customer base. Customers of an AI platform report high AI adoption, which tells you nothing about the market.
- The wording: what counted as "using AI". A team that trialled a chatbot for one building in 2024 can be a yes.
- The incentive: the company publishing the finding sells the product the finding endorses.
None of that makes a vendor statistic false. It makes it unverifiable, which is a different thing and matters more when the number is going in front of your investment committee.
How to Read an Adoption Statistic Before You Repeat It
Four questions do the work. They take about two minutes.
- Who ran it? Find the original publisher rather than the outlet reporting it. If the trail ends at a company selling the category, you have marketing research.
- Who did they ask, and how many? A credible survey names its population and sample size. If you cannot find an equivalent sentence, that absence is the finding.
- What did the question mean? "Using AI" spans a leasing agent pasting into ChatGPT and a portfolio-wide deployment. Broad definitions produce big numbers.
- Does the number match anything else? A figure with no independent corroboration is a claim.
Apply these to the slide-four figure and it thins out. Apply them to NAR's 32% and it holds, because you can see the method.
What the Neutral Data Does Tell You About Sector Differences
Here is where honesty costs something. No neutral US dataset breaks conversational AI adoption down by multifamily against BTR against PBSA against commercial.
The datasets that exist were built for other questions. NAR surveys agents, PwC and ULI survey capital, and neither was designed to compare a student housing operator with a build-to-rent operator. Anyone presenting that breakdown is extrapolating or selling.
What PwC and ULI do show is where capital is moving, with data centres, senior housing and self-storage drawing attention on the strength of AI-driven demand. That is a signal about AI as an asset class rather than AI as an operating tool. Useful, and a different question from whether leasing teams are using it.
The sector comparison you want does not exist yet in neutral form. Treat any version you are shown accordingly.
The Gap Between Using AI and Getting Value From It
The most useful number in the verified data measures impact rather than adoption.
NAR asked members what AI had done for their business. 17% reported a significantly positive impact, 33% a moderately positive one, and 46% said AI had made no noticeable difference at all.
Roughly two-thirds of agents have adopted AI, and nearly half report it changed nothing. That gap between adoption and impact is the finding worth carrying into a vendor meeting.
Adoption is easy. Someone opens an account and the org chart says you have AI. Impact requires the thing to be wired into the work, which is the difference between a chatbot and a conversational AI platform.
The agent data cannot tell you why the gap exists, only that it is normal and large. When a vendor quotes an adoption rate as evidence their category works, they are citing the metric that moves for free. Ask what happened to the leases instead.
What to Watch Over the Next 18 Months
The trajectory in the neutral data is steep. PwC and ULI's European edition found 75% of respondents applying AI-based solutions, against 51% the year before. That is European and investor-weighted rather than US and operator-weighted, and it still shows a category moving fast.
Three things worth watching:
- Whether a neutral operator survey appears: a trade body running a sound census of multifamily AI use would settle what vendor research cannot. Until one exists, the honest answer is that nobody knows precisely.
- Whether reporting shifts from adoption to outcomes: the interesting number counts how many operators kept AI past renewal, and what happened to their conversion rates.
- Whether definitions tighten: "AI adoption" covers everything from a website chatbot to agents wired into a PMS. Numbers improve as categories narrow.
Adoption statistics will keep arriving with confident decimal points. The operators who do well over the next 18 months will be the ones asking who was surveyed.
What VerbaFlo Publishes and What It Does Not
An article arguing that vendors should show their working owes you the same standard.
The figures VerbaFlo publishes describe its own platform rather than the market. The company powers communication across more than 200,000 residential units in 10 countries and supports conversations in over 180 languages.
Those are operational counts rather than survey findings, and they are no evidence about what your peers are doing. VerbaFlo has not run an industry adoption survey. Running one as a vendor would produce exactly the artefact this article warns about.
What the platform is built to do is make outcomes measurable on your side:
- Reporting on your data: enquiries across voice, chat, WhatsApp and email land in one place, so your conversion numbers come from your portfolio.
- Outcomes rather than activity: tours booked and leads qualified against your own baseline, which is the comparison an adoption rate cannot make.
- Your systems, your record: because VerbaFlo is API-first and connects to the CRM and PMS you already run, results appear where your team already tracks them.
Ask us the four questions above, then ask every vendor on your shortlist the same ones. Weigh the answers against your own multifamily, build-to-rent or student housing portfolio rather than a slide.
See what the numbers look like on your own data. Book a demo.
