I was an early adopter of AI for the least glamorous reason imaginable.
It was 2021, and I was going through a divorce and had a pile of paperwork and a very expensive attorney. AI turned out to be a gift from above. I discovered an Apple app, Typing Mind, and learned about API keys to access multiple AI providers — OpenAI, Anthropic, Google, Perplexity, OpenRouter — through a single frontend.
TAKE THE INMAN INTEL INDEX SURVEY
To the uninitiated, like me, it was transformational. Where before I had felt powerless and uninformed, now I felt more informed and a little confident, too. The AIs were helpful and full of suggestions for my attorneys (which they hate by the way). I began to have some hope.
But I soon learned that AIs are like “that friend”: You know the one who’s super helpful, always hyping you up, super confident but gets their information from YouTube or Reddit. Well-intentioned but prone to text you at 10 p.m. with political conspiracy theory videos they saw on Facebook.
You love them. They mean well, but …
Getting to know you
AI models are fast, helpful and eager to please. They are also too friendly, too confident and more than willing to invent details when the real answer isn’t in their data. In plain English, they will lie with the confidence of a 3-year-old to keep you happy and make you think they’re smart
I learned to prompt the models to cite sources and provide links. The more I used them, the more I learned how to use them. And the more I used them, the more I learned I couldn’t trust them.
It’s the defining feature of the technology, and real estate agents need to understand it before they build a business around it.
The cheerleader problem
Once you start testing models side by side, the pattern shows up quickly. One model sounds more polished. Another sounds more analytical. Another sounds warmer. But the behavior is consistent across the board. They are designed to be helpful and engaging.
That’s great when you are drafting an email to a past client, but less great when you are making decisions about a property, a loan, a contingency or whether that crack in the foundation is “cosmetic” or structural.
A model will tell you “this seems like a fair price” with complete confidence, but it doesn’t know that the comps it’s using are six months stale or that the comparable sale it’s referencing closed 12 percent below asking because the foundation was failing. The model doesn’t know what it doesn’t know. It just sounds like it does.
The real estate test
Real estate is where AI gets interesting because the industry runs on information, timing and interpretation. That makes it a natural place to use AI and a dangerous place to use it lazily.
The National Association of Realtors 2025 Technology Survey found that 68 percent of agents now use AI tools in their business, yet only 20 percent use them daily. That gap isn’t about access. It’s about trust. Agents are experimenting, but they haven’t integrated AI into their workflow.
Agents are already using AI for listing copy, social media, email follow-up, market research and client education. Consumers are already using it to compare homes, summarize neighborhood data and ask questions they used to ask a human first. Vendors are already selling AI as though it is a shortcut to intelligence.
It’s just a shortcut, and that’s why verification is the new core skill.
The retail problem
The real estate industry has always loved the promise of a shortcut. A better lead source. A smarter CRM. A predictive pricing tool. A dashboard that will somehow replace discipline, judgment and follow-through.
AI slots into that dream because it sounds like the next upgrade. It is faster, cheaper, writes better and has no problem doing all the things agents and assistants hate doing, so of course, vendors are pitching it like a panacea.
That is why so many vendors are selling AI the way they used to sell “synergy” and “blockchain.” They know the average user will not check the work. They know most people are too busy, too trusting or too relieved that the machine wrote the email for them.
But in real estate, convenience without verification is just a fast way to get into trouble.
The human touch
The smartest people I know are not treating AI like a magic wand. They are treating it like a junior assistant who works very fast and occasionally tells lies with eye contact.
That means the output gets reviewed. The facts get checked. The assumptions get tested. The final word stays human.
This is where real estate becomes the perfect case study for AI trust. The industry already depends on translation. We take complex facts and turn them into usable advice. We explain contracts, financing, market behavior and why the seller’s “firm” price is negotiable when the home has been on the market for more than 30 days.
That makes AI useful, but it also makes it risky. Because we are in a business where a shortcut can cost a client tens of thousands of dollars, confidence is not enough. We need verification.
The trust gap
Ultimately, AI can be a timesaver and a great starting point, and for those who understand that simple fact, AI is a great tool. However, it is not the final answer.
That distinction matters because consumers are being told that AI can help them write listing descriptions, summarize market trends, answer pricing questions and explain contract language. Some of that is true.
Some of it is useful. But useful is not the same thing as trustworthy, especially when the thing being discussed is governed by a regulatory framework that gets very tetchy when it catches people being lazy.
The winning agents in the next cycle will not be the ones who adopt AI fastest. They will be the ones who learn to verify AI output fastest. They will be the ones who treat the model as a starting point, not a final answer.
America Foy is a broker associate at The Grubb Co. Connect with him on LinkedIn and Instagram.
