48% of U.S. Realtors Use AI Daily or Weekly, NAR Reports
Nearly 48% of Realtors use AI daily or weekly. NAR's data science director urges paid tools, verified facts, image disclosure and a close reading of ethics rules.

By Priya Raman
48% of U.S. Realtors use AI daily or weekly, according to the National Association of Realtors’ 2026 Realtors Technology Report, released in September.
The report puts daily use at 23% and weekly use at 25%. Another 31% experiment occasionally. Only 12% do not use AI and have no plans to try it, compared with 32% who had not yet tried AI in 2025. The source material gives no sample size or revision history for the report.
At the National Association of Realtors’ 2026 NXT conference on Sunday, data science director Mike Woodward addressed agents already using AI for routine work, client relationship management and marketing materials.
Woodward invoked Spider-Man to frame the risk: “With great power, comes great responsibility.” His session covered what AI can do for agents and the wider housing industry.
Why does a free AI tool put client data at risk?
Woodward began with AI companies’ business model. Agents must understand how those companies make money before judging the risks. He said there is no free lunch.
He told of two pigs who mistook a barn’s shelter and food for generosity. The ending, he warned, makes the customer the product.
Free language models can ingest uploaded documents, learn from them and surface what they have learned to the next person who queries the service.
He described language models as data-hungry systems. He advised using paid services for real or confidential information. Agents should also select data-retention opt-outs.
Those settings stop models from retaining or training on inputs. On free tiers, use dummy data. Never enter confidential material.
His cautionary case involved a fictional former Air Force employee’s performance review. Such sensitive text should never enter a free or overseas-based model.
How reliable are AI-generated property answers?
Woodward called overconfidence the second trap. He rejected the idea that language models resemble the Oracle of Delphi.
His tests used the kinds of queries his clients run.
Copilot searched for supermarkets and schools near a Massachusetts town but answered too literally, omitting stores and schools in neighboring towns.
Claude supplied records for his home. Five of seven facts came back wrong.
His conclusion was not that one model dominates. Agents should choose a tool, learn how it works and adapt it to their business.
Woodward also rejected elaborate prompt engineering. His advice: use short sentences, simple words and clear instructions.
The lesson reached beyond errors. Buyers now arrive with AI-generated impressions of towns and properties. Agents need to understand what those buyers think they know.
Which uses carry the lowest risk?
Woodward separated helpful, low-risk tasks from work that can mislead buyers or expose confidential information.
- Summarize a soil-analysis report in plain English and produce a step-by-step repair plan with Claude.
- Draft a listing description from a property fact sheet, a task that might take an agent 15 minutes and take seconds instead.
- Research companies, build charts and generate presentations as starting points for further work.
Image editing follows a similar line.
Exposure, white balance and horizon correction are familiar professional fixes, now available without proficiency in Photoshop.
Removing temporary blemishes can cross into misrepresentation. Adding a fireplace, moving a window or inserting a swimming pool can also breach the Realtor Code of Ethics.
AI may also change images that agents did not request to alter. The published listing still carries the agent’s name.
He rejected the excuse that the software made the change.
What does this change for Realtors?
Woodward said the backlash is already here.
He cited a recent front-page New York Times article in which real estate professionals appeared in a poor light. He also pointed to pages of angry posts on Reddit.
State legislatures and New York City are weighing regulation. Woodward believes AI-detection tools work well.
Woodward said: “If you use AI and you don’t confess to it, you’re going to get caught.” He expects litigation and describes consumers as highly angry.
Woodward’s lawyers identified a legitimate use: showing buyers how a property could look. Virtual staging can make an empty rental’s bare walls feel welcoming.
Disclosure is the condition. Show buyers the original beside the edited version. Label the edit as AI-generated.
His final advice was practical. Use paid tools for confidential information. Reread the code of ethics. Check the laws applying in your state. Consult counsel when uncertain.
With legislatures weighing rules and buyers already using AI, agents will need to verify facts, disclose edits and use paid tools with retention opt-outs for confidential work.
Source: housingwire.com
More from Priya Raman
Show full bio
Market editor covering people, launches and day-to-day practice in Job market data and labor economics at Path & Pay.
561 articles