Artificial intelligence has stopped being something real estate agents try and started being something they use on a schedule. Nearly half of agents now reach for it daily or weekly, and the share who say they are not using it and have no plans to has fallen to 12%, according to the National Association of Realtors’ 2026 Realtors Technology Report, released Sept. 22.
A year earlier, by the association’s own comparison, 32% of members had not yet tried AI at all.
The report, which surveyed NAR members about how technology is shaping their businesses, breaks AI use into four active tiers: 23% use it daily, 25% weekly, 31% are experimenting occasionally, and 9% describe themselves as curious but have not tried it.
What is driving adoption is less about the technology than about the clock. Eighty-one percent of agents said saving time is their primary goal in taking on new tools, up sharply from 66% a year ago. Improving the client experience came second at 71%, up from 64% in 2025. Closing more deals (57%), reducing manual work (54%) and staying ahead of the competition (44%) followed.
“Real estate agents are making practical choices about technology, and the two payoffs they want most are time and a smoother experience for their clients,” said NAR Deputy Chief Economist Jessica Lautz. “Those priorities have grown stronger over the past year. When routine work gets handled faster, agents have more time for the guidance, negotiation, and steady advice that clients count on through a major financial decision.”
Marketing copy first, analysis later
Among the agents who use AI, the tasks they hand it are overwhelmingly promotional rather than analytical. Three-quarters β 75% β use it to write listing descriptions. Social media posting follows at 56% and drafting emails and follow-up messages at 52%. Producing market summaries and drafting marketing content in a personal tone tie at 30%, and 27% use AI to review or summarize documents.
That ordering matters for anyone building products for this market. The work agents are comfortable delegating is the work where a wrong output is embarrassing rather than costly. Document review β where an error has consequences β sits at the bottom of the list, used by roughly one in four AI users.
The stack underneath
AI is still a newer layer on a settled foundation. The MLS remains near-universal at 96%, followed by e-signature tools at 79%, showing-scheduling tools at 68%, comparative market analysis and pricing tools at 59%, drone photography and video at 48% and customer relationship management software at 46%. AI-generated content tools, which NAR’s list names as ChatGPT, Copilot and Gemini among others, are used by 41% of agents, a higher share than reported CRM use.
Cost discipline shows up in what agents actually spend. A plurality, 36%, put $50 to $250 a month into technology. Another 18% spend less than $50, 19% spend $251 to $500, and 22% spend more than $500.
What still slows it down
The barriers agents cite are practical. Sixty-three percent named the learning curve as their biggest challenge in adopting new technology; 59% pointed to cost. Those two answers, taken with the spending distribution, suggest the constraint on the next wave of adoption is not skepticism about whether the tools work.
Clients, for their part, are not pushing back hard. Forty percent of agents said clients responded very positively to technology in the buying and selling process, and another 37% said clients found it helpful while voicing some reservations.
The findings land in a year when vendors have been racing to put AI directly in front of agents rather than behind the scenes. ATTOM this year released three AI agents that let users query its property data in plain English, RealReports launched a tool meant to surface business for agents before they go looking for it, and the country’s largest MLS rebuilt the workspace its members open first. NAR’s numbers suggest those products are arriving into a membership that has already built the habit.
One caution on reading the survey: these are agents reporting on their own behavior, and self-reported adoption tends to run ahead of measured usage. The steadier signal in the data is the direction and the size of the one-year moves β a 15-point jump in the share citing time savings, and a collapse in the share of members sitting the technology out entirely. More coverage of tools and adoption is on our technology and AI page.
