How I Use AI in Real Estate
Artificial intelligence has become one of the tools I use throughout my real estate business.
I began integrating AI into both my real estate website and my broader real estate process in late summer 2025. I now use it for research, analysis, organization, writing, individualized work for buyers and sellers, technology development, coding, interactive website widgets and tools, and some pictures and illustrations for this website.
Over time, I have also created around a dozen custom GPTs for real estate-related purposes. These give me specialized workflows for recurring real estate questions and tasks instead of requiring me to start from scratch each time.
AI is, however, a relatively recent addition to a body of work that largely predates my use of it. As of the creation of this page in September 2026, the majority of the pages and articles on my real estate website have had no AI involvement in either their initial creation or subsequent edits to date, and more than 90% (out of over 400 articles and pages) were not initially built using AI.
I believe substantial AI assistance should generally be disclosed rather than presenting work in a way that could leave readers assuming AI played no meaningful role.
At the same time, an AI disclosure does not mean that everything—or necessarily even most of the underlying information—originated with AI.
In many cases, some of the greatest value comes from combining AI with highly detailed original content, unpublished research, professional real estate information, real-world observations, original media and experience that I developed independently of AI.
AI-assisted should not automatically be interpreted as AI-created from scratch.
Why I Started Using AI
One thing that may surprise people is when—and partly why—AI became an important part of my business.
I did not begin using AI during the highest-volume period of my real estate career.
After more than $8 million in sales in 2020, I became increasingly interested in how much more I could offer buyers and sellers if I devoted additional time to research, resources, marketing, technology and individualized client service.
In the years that followed, I intentionally allowed my sales volume to decline while investing substantially more time in buyer and seller research, financing and credit resources, photography, aerial media, 3D tours, neighborhood resources, housing-market analysis, website tools and other improvements to what I could offer clients.
I discuss that progression in more detail on my About Adam page.
A Difficult Buyer Research Project Helped Push Me Toward AI
One of the experiences that helped push me toward AI involved an unusually research-intensive assignment for a buyer.
It was the kind of task that I believe many real estate agents and brokers would never have attempted in the first place because of the amount of research it required.
I was willing to attempt it.
But I was too slow.
That bothered me.
The lesson I took from the experience was not that I should stop doing unusually detailed research for clients. It was that I needed better tools if I wanted to perform that level of research quickly enough for the research to be useful.
AI offered a potential solution.
Instead of reducing the depth of the work, AI could reduce some of the time required to locate, organize, compare, question and synthesize large amounts of information.
That experience continues to shape how I use AI today.
One of its greatest benefits to me is not simply making ordinary tasks faster. It can make more ambitious research practical for an individual buyer or seller when that research might otherwise require more time than the circumstances of an actual transaction reasonably allow.
One reason I adopted AI was not because I wanted to research less for clients, but because I wanted to research more without being too slow to make that research useful.
I began incorporating AI during the lowest-sales year of my career since my first year in real estate sales.
That context matters.
I had already deliberately reduced my transaction volume to devote more time to improving my research, resources, marketing and client service before AI became a significant part of my workflow.
My interest in AI was partly the result of trying to provide more depth, not less.
AI Can Help Me Serve Clients Better—and Grow My Business
I use AI for several overlapping goals.
One is to increase how much useful work I can do for each client.
Another is to make my business more efficient, improve my marketing and resources, reach more prospective clients and, ultimately, increase my sales volume.
Those goals do not have to conflict.
One reason I intentionally reduced sales volume in prior years was to devote more time to developing resources, technology, research and marketing capabilities. AI can now help me make substantially more use of that work while also making it more practical to serve additional clients.
For an individual buyer or seller, AI may help me investigate an unusual question more deeply, research financing possibilities, study a property or location issue, challenge an assumption, improve a seller's marketing, organize information or identify an issue neither my client nor I initially considered.
For my broader business, AI can help me develop better website resources, build interactive tools, improve marketing, make existing research more useful and accomplish projects that otherwise would have taken substantially longer or required outside development.
I want AI to help me both improve the service I provide and build a more productive real estate business.
Combining Highly Detailed Original Work With AI
One of the most powerful uses I have found for AI is combining it with the large amount of highly detailed original material and research I created independently of AI.
I was doing that kind of work years before generative AI became a significant part of my workflow.
One example is my Master Spreadsheet for Prospective Home Purchase Considerations, where I brought together an unusually large number of issues a buyer might want to investigate when evaluating a prospective home. I created that underlying resource without AI.
Another example is my research into which MLS systems are actually needed to effectively buy or sell in different parts of Virginia and surrounding areas. That involved original research into MLS coverage, geographic overlap, access and other considerations rather than simply repeating a generic explanation of what an MLS is.
Those kinds of resources create a very different starting point from simply asking AI to write a generic real estate article from scratch.
I can give AI detailed material that I already researched and developed, ask it to compare that material with additional information, identify gaps, integrate relevant material from several of my existing resources, challenge parts of my reasoning, reorganize complicated information and help me turn it into something substantially more useful.
For me, highly detailed original work + AI can be a better combination than either one by itself.
My original work contributes context, local knowledge, professional information, accumulated research and material specific to the way I practice real estate.
AI can help me connect, test, analyze, expand and organize it.
I was already producing unusually detailed material.
AI gave me a way to do considerably more with it.
How I Use AI While Helping Buyers & Sellers
My use of AI is not limited to writing articles.
I sometimes use AI as one of the tools involved in helping an individual buyer or seller with a particular question, property, decision or transaction.
I have created around a dozen custom GPTs focused on real estate-related purposes. These give me more specialized and repeatable workflows around some of the research and questions that arise in my work.
Depending on the situation, AI may help me:
investigate an unusual buyer or seller question
organize and compare large amounts of information
research financing or potential cost-saving opportunities
analyze housing-market information
investigate location-related considerations
identify matters deserving additional investigation
challenge my own reasoning
research an unfamiliar issue arising during a transaction
organize information for a client
examine potential marketing strategies
identify questions worth raising with a lender, inspector, insurer, attorney, contractor or another professional
determine what additional information I should obtain before reaching a conclusion
Sometimes AI saves only a modest amount of time.
In other situations, it can make practical a level of research that would have been substantially more difficult to perform manually within the time constraints of an actual transaction.
That does not mean I simply give a client whatever answer an AI system produces.
It means AI can help me investigate more possibilities before I determine what information is reliable and useful enough to rely upon or pass along.
AI can increase not just how much public content I produce, but how much individualized research and problem-solving I can bring to one client.
Combining Published Content, Unpublished Research, MLS Information & AI
Not all of the original material I combine with AI has ever been published.
I have accumulated substantial unpublished original research, including Google Docs, spreadsheets, notes and analysis developed through my real estate practice.
I also have direct access to four MLS systems and indirect access to two additional MLS systems. That gives me professional real estate information that is different from simply asking a public AI system what is happening in a housing market.
AI does not independently replace that access.
When appropriate, I can combine information or analysis derived from those professional resources with my own interpretation, experience and AI-assisted analysis.
If I have already performed detailed research—whether it is published on this website, stored in an unpublished Google Doc, derived from MLS information or accumulated through professional experience—I do not have to ask AI to start with a blank slate.
I can give it a much richer foundation.
A common part of my current workflow is also asking AI to integrate relevant material from some of my existing articles into a new or expanded resource. Some of those source articles were themselves researched and written entirely without AI.
A Six-Page Unpublished Google Doc Became a Public Article
My article on buying a house near power lines is one example.
The foundation was an approximately six-page unpublished Google Doc that I had already created through my own research.
I later used AI to help transform that existing research into a more organized and developed public article.
That is a good example of why I do not consider “AI-assisted” synonymous with “AI-created.”
AI helped me do more with substantial original research that already existed.
The value is often in the combination: original research + professional information + human judgment + AI-assisted analysis.
Related resource: Buying a House Near Power Lines
I Also Use AI to Build Website Tools, Widgets & Images
My use of AI goes well beyond conventional writing.
I increasingly use AI to help create interactive tools, calculators, data displays and widgets for my real estate website.
This allows some research that might otherwise exist primarily as text or spreadsheets to become something visitors can actually interact with.
One example is my Hampton Roads housing-market tracker, where AI has assisted substantially with developing the interactive widget and its functionality.
AI can help with coding, debugging, interface design, data organization and turning an idea into a functioning tool much faster than I could build it entirely on my own.
I also sometimes use AI to create pictures, illustrations and other visual material for this website.
Those uses are different from photography or 3D capture of an actual property or location. An AI-generated illustration is not the same thing as an actual aerial photograph, listing photograph or Matterport scan.
Related resource: Hampton Roads Housing Market Tracker
How Much of My Real Estate Website Was Created With AI?
As of September 2026, the majority of the pages and articles on my real estate website (over 400) have had no AI involvement in either their initial creation or subsequent edits to date, and more than 90% were not initially built using AI.
That distinction matters because I spent years developing real estate resources before generative AI became part of my workflow.
Even when AI now contributes substantially to a newer article—or to a major expansion of an older one—much of the underlying information may predate my use of AI.
An AI-assisted resource can therefore contain a combination of:
earlier articles I created without AI
unpublished original research
professional data
MLS research
real estate experience
photographs and video
aerial imagery
3D tours
my questions and instructions
new research I perform myself
AI-assisted research and analysis
corrections I make to AI
my final editorial decisions
“AI-assisted” does not mean “AI-created from scratch.”
I Correct AI Frequently
AI is useful partly because it can perform an extraordinary amount of work quickly.
That does not mean it is right every time.
When creating a new article, I frequently correct AI multiple times within the same research or writing process.
Those corrections can involve factual errors, outdated information, misunderstanding a source, geographic mistakes, treating unlike datasets as though they were equivalent, drawing conclusions that are too strong for the evidence, missing exceptions, overlooking information I already know, failing to distinguish something that matters locally, using wording that could mislead, relying too heavily on a weak source or overlooking a practical issue that matters in an actual transaction.
Sometimes the correction changes a sentence.
Other times it changes a substantial part of the article or the direction of the research.
Sometimes one of the most useful things I can tell AI is essentially:
You're wrong. Look again.
I also frequently ask AI to challenge conclusions, look for contrary evidence, identify what may be missing and reconsider an answer when I do not think the available evidence supports it strongly enough.
I do not primarily want AI to agree with me.
I often get more value from it when it helps identify why my initial conclusion—or its own initial conclusion—could be wrong.
Knowing how to use AI well therefore involves more than knowing how to ask it questions.
Knowing when its answer deserves to be challenged can matter just as much.
Sometimes the Common Answer Is Too Simple
Another limitation of relying solely on AI—or solely on a Google search—is that the most commonly repeated answer is not necessarily the best answer for a particular buyer or seller.
Sometimes advice is wrong.
More often, the actual answer is simply much more nuanced than a quick search makes it appear.
One example is evaluating listing agents based on market time.
It may initially seem obvious that the agent whose listings sell fastest must be the better choice.
But market time can be affected by pricing strategy, which listings an agent is willing to accept, seller objectives, offer timing, local MLS practices and other factors. A very short market time does not automatically mean that a seller maximized the outcome.
That illustrates a broader issue.
AI is often very good at identifying the conventional answer.
Google is often very good at identifying the popular answer.
Neither guarantees that an answer is complete, sufficiently nuanced or right for the particular situation in front of me.
Related resource: My Article on Agent Market Time Questions
Why AI Does Not Replace a Good Real Estate Agent
AI can be extremely useful to me.
I do not believe that means it replaces a good real estate agent.
One reason is access to information.
Another is real-world observation.
Another is professional judgment.
And another is that some services or ideas a good agent provides may be uncommon enough that a generic AI system or even a real estate-specific one would never think to recommend them in the first place.
Related:
Should You Hire a Buyer's Agent? What do They Do?
The Importance of Choosing a Great Listing Agent
The Importance of Choosing a Great Buyer's Agent
Historical First Street Research Is One Example
For years, one of the resources I used when researching properties was First Street's property-specific flood-risk mapping.
First Street's flood model incorporates multiple forms of flooding, including heavy rainfall, rivers and streams, high tides and coastal storm surge, and property-level risk can vary with factors such as elevation and proximity to water.
I no longer have the First Street access I previously used.
However, I retained more than 100—and likely hundreds—of screenshots from First Street maps that I captured while I had access.
Those screenshots now form part of my own historical research archive. They are not a substitute for current risk information, but they can provide historical context that an AI system would not independently have unless I supply the relevant material.
The larger point goes well beyond First Street.
An AI system may not have access to a particular MLS screen, historical screenshot, private spreadsheet, unpublished Google Doc, interactive map or other professional information I have accumulated.
And even when it has excellent online information, it does not independently have eyes on the ground.
Related resource: First Street vs. FEMA Flood Likelihood
Eyes on the Ground + AI
I can stand at a property and notice things that do not necessarily appear online.
I can hear nearby traffic or an unusual sound.
I can notice an odor.
I can look at drainage and topography.
I can see deferred maintenance.
I can notice that something looks different in person than it did in photographs.
I can photograph a model or serial number when permitted.
I can look into an accessible dark area with a powerful flashlight.
I can recognize that something deserves additional research.
That physical context can also make AI more useful.
During a showing, I may notice something specific, have a buyer ask a question, photograph a model or serial number when permitted, or identify a particular issue I want to research.
I can then use my custom buyer GPT or another digital tool to research that specific question or observation while I am still at the property.
I do not stand in a house and simply ask AI an open-ended question about everything else I should consider.
The value is more targeted.
For example, I might use it to research a particular model number, investigate something unusual that I or the buyer noticed, identify an object or material, or better understand a specific question that arose during the showing.
That can provide something close to the best of both worlds:
eyes on the ground + professional information + specialized physical tools + AI-assisted research.
The AI may have access to an extraordinary amount of general information.
I have the actual property in front of me, the buyer beside me and the context of what matters to that particular client.
The AI does not replace the eyes on the ground. The eyes on the ground can make the AI considerably more useful.
Related resource: My Buyer Feedback Process
Pictures, Physical Tools & Permission During Showings
One of the ways my buyer process differs from a more typical showing is the number of physical tools and digital resources I have available to help investigate a property.
That does not mean every tool or activity can simply be used without permission.
Pictures and video are a good example.
I commonly seek permission in advance from the seller or listing agent to take pictures or video during a showing. Those images can help a buyer remember a property, document something that deserves additional investigation, identify model or serial-number information, or provide useful input to digital research tools.
Many other tools or types of access also depend on the circumstances and permission.
Powerful Flashlights Are a Major Exception
A flashlight is one of the simplest but most useful tools I bring.
My lighting equipment includes an IMALENT MS18, an Olight Marauder 2, two Marauder Minis and additional powerful flashlights and headlamps, with lights available for clients to use as well.
These are substantially more capable than an ordinary phone flashlight.
Powerful lighting can be particularly useful in dark garages, sheds, closets, mechanical spaces, attics or crawlspace areas where access is appropriate, houses with limited electricity, nighttime showings and other poorly illuminated locations.
Sometimes being able to see better means being able to notice something that otherwise would have been missed.
I also own other specialized equipment that can be useful in buyer work, although many tools require permission or have circumstances where use would not be appropriate.
I am not a home inspector, and these tools do not replace a professional home inspection.
The purpose is to have more useful tools available when they can help a buyer make a better-informed decision.
This is one of many areas where I have deliberately tried to build a buyer process that is more client-friendly than simply doing what is typical.
Related resource: My Showing Process and Permission Requests
Pictures Can Become Research Inputs
Pictures taken during showings, when permitted, can be useful for much more than helping a buyer remember what a room looked like.
I may photograph an HVAC label, water heater, appliance, unusual material, plant, hard-to-see area or something else that deserves further research.
That image can then become an input for tools such as ChatGPT Business or other specialized services.
This is another example of why I see AI as more useful when it is combined with information collected in the real world.
AI may be able to interpret a photograph.
But someone first had to be at the property, notice what was worth photographing and provide the relevant context.
Related resource: Why I Ask About Pictures & Video at Showings
AI Can Miss Ideas That Are Unusually Rare
AI has access to enormous amounts of information about things people are already doing and writing about.
That can create another limitation.
An idea that is highly unusual may not reliably appear in a generic AI recommendation at all.
My Matterport neighborhood tours are a good example.
In addition to creating a 3D tour of a seller's house, I have created Matterport experiences of neighborhood and community amenities that can be incorporated into the marketing of an individual home.
Because virtually no agents offer that type of marketing, I would not expect a generic question such as “How should I market my house?” to reliably produce “create a Matterport 3D experience of neighborhood amenities and integrate it with the property's marketing” as one of the recommendations.
That does not mean AI cannot understand or help develop the idea once I give it the information.
It means AI can be influenced by what is common enough to be well represented in the information available to it.
The same principle applies to some of the tools I bring into my buyer process.
If few agents integrate unusually powerful flashlights, thermal equipment, moisture-detection tools and other specialized equipment into showings, an AI system describing what a typical buyer's agent does may never tell a buyer that those additional services are possible.
Sometimes differentiation comes from doing something that isn't typical enough for AI to treat it as typical advice.
Related resources:
Adam Garrett's Video Visuals for Out of Town Real Estate Buyers/Sellers
Adam's Area Photos For Listings (every listing even if not in an HOA or condo association)
Photography, Aerials, Zillow 3D, Matterport, Video & Virtual Enhancements
A significant amount of the media used in my real estate work comes from real-world photography and capture rather than AI generation.
However, I do not personally create every piece of media used to market my listings.
For the on-the-ground still photography of homes I list, I hire professional real estate photographers.
For every listing, I combine those professional still photographs with additional media and marketing capabilities I provide myself.
I am an FAA Part 107 remote pilot and personally produce aerial photography and video. I have also built a substantial library of my own area, amenity and off-site aerial imagery that can be used to show more than just the house itself.
Matterport Pro3 + Ricoh Theta Z1
I also personally create 3D tours.
When I create a Matterport tour of a home, I use my Matterport Pro3.
Before acquiring the Pro3, I had already reached the point where I often preferred the Matterport work I produced myself with my Ricoh Theta Z1 to Matterport tours I had hired outside professionals to create.
The Pro3 subsequently made the technological difference between what I can produce and many typical real estate 3D-tour setups even greater. Its LiDAR-based capture also makes high-quality exterior Matterport work substantially more practical than it was with my earlier 360-camera setup.
I still use the Ricoh Theta Z1, but one of its major roles now is creating Zillow 3D Home tours.
Rather than treating Zillow 3D and Matterport as an either/or decision, I use both in some of my listing packages depending on commission and package level.
All of my 3% listing-side commission house listings receive both a Zillow 3D Home tour captured with my Ricoh Theta Z1 and a Matterport tour captured with my Matterport Pro3.
This is representative of how I think about technology generally.
The best answer is often not choosing one useful tool and rejecting another. It is understanding where each has strengths and combining them intelligently.
Related resource: Matterport vs. Zillow 3D Tours
Digital Staging & Other Virtual Photo Enhancements
Not everything in my listing marketing is untouched photography.
I sometimes hire professional third parties to digitally stage photographs or perform other virtual enhancements when I believe those services add value.
I do not necessarily perform those edits myself, and the exact technology used by a provider can vary.
Digital staging can help prospective buyers visualize how an empty room might look when furnished. Other virtual enhancements can improve presentation or illustrate possibilities that would be more difficult or expensive to create physically.
When an MLS listing image has been digitally staged, I disclose the digital staging in the MLS.
My position is therefore not that digital tools, virtual enhancements or AI should never be used.
I use them when I believe they add value.
The important distinction to me is between improving presentation and using technology in a way that could materially mislead someone about what physically exists at the property.
The more an alteration changes what a consumer could reasonably believe is physically present, the more important appropriate disclosure becomes.
AI Use & Disclosure in the Real Estate Industry: Some of the Most Prolific Real Estate "Authors" Don't Touch Their Own Written Content: They Outsource it to a Company that Uses AI to Write it Without Their Involvement
AI is no longer unusual in real estate.
The National Association of REALTORS®' 2025 Technology Survey found that 46% of surveyed REALTORS® were using AI-generated content, with 20% reporting daily AI-tool use and another 22% weekly use.
A separate 2026 Realtors Property Resource survey of 225 NAR-member agents found that 92% were already using AI or planning to use it, while 63% identified accuracy of AI outputs as their leading concern.
What is harder to quantify is how often all of those uses are disclosed to consumers.
I do not think the available evidence supports claiming that a particular percentage of agents fail to disclose AI use in articles, emails, listing descriptions and every other category.
There is, however, much stronger evidence in one visible part of real estate marketing: digitally altered listing imagery.
A 2026 Coraly analysis of just under 40,000 primary listing images from Zillow, Redfin, Realtor.com and Homes.com found that 10.8% showed evidence of digital alteration, while more than 90% of those altered images carried no visible disclosure.
I use digital staging and other virtual enhancements myself.
The difference is that when I digitally stage an MLS listing image, I disclose that fact in the MLS.
My concern therefore is not that the technology should never be used.
It is that consumers should receive appropriate information when technology materially changes what they are seeing.
California has already adopted disclosure requirements for certain digitally altered real estate advertising images, including disclosure of the alteration and availability of the original image.
My Experience With a Major Real Estate Website Provider
There can also be practical obstacles to AI disclosure.
When a major company that builds websites for real estate agents and brokerages approached me about moving my real estate website to its platform, I specifically discussed AI disclosure with them.
I was told that their system limited the ability to include the type of AI-assistance disclosure I wanted even if I, as the agent, wanted that disclosure included.
I did not ultimately move my website to that company.
I am intentionally not identifying the company here.
My purpose is not to criticize a particular business, and its technology or policies could change.
The conversation nevertheless reinforced something important to me: when an agent or broker gives substantial control of a website or its content system to a third party, they may not always have complete control over how AI involvement is communicated to consumers.
Because I maintain substantially more control over this website, I can make my own decisions about when and how substantial AI assistance should be disclosed.
My goal is not to label every sentence AI has ever touched.
There is a meaningful difference between using AI for a minor wording correction and relying on it substantially for research, analysis, coding, imagery, organization or drafting.
But when AI's contribution is substantial, I generally favor meaningful disclosure—even when similar AI use elsewhere may not be disclosed.
I recognize that this can result in more visible AI disclosure on my real estate website than on some competing websites that also use AI.
I consider that transparency a positive, not a disadvantage.
AI, Privacy & Private Client Information
Some of my AI use involves individualized work for actual buyers and sellers.
That means there are circumstances in which I provide private or client-specific information to AI when I believe doing so is useful to the work I am performing for that client.
I do not approach that the same way I would approach putting private information into an ordinary public consumer AI workflow.
For my professional use, I use ChatGPT Business, including custom GPTs within my Business workspace.
OpenAI states that ChatGPT Business inputs and outputs are not used to train its models by default, that Business data is encrypted both at rest and in transit, and that each Business user has a separate chat history that other workspace members do not automatically see. OpenAI says its business-data commitments also apply to GPTs used within Business workspaces.
That gives me a private Business workspace with meaningfully different data handling from an ordinary consumer AI account.
I would not describe any cloud AI service as literally sealed off from its provider or equivalent to keeping information only on a computer in my office. That said, I have many other workflows that integrate cloud data & have done so for years prior to integrating AI into my workflow.
ChatGPT Business remains a cloud service.
But there is an important difference between using a business environment where workspace data is excluded from model training by default and encrypted, versus treating every AI environment as though it handles private client information the same way.
I still use judgment about what information is relevant to a task.
At the same time, I do not want to create the false impression that all of my AI use consists of sanitized hypothetical questions.
Sometimes AI can be more useful to a buyer or seller precisely because I can give my private real estate workflows enough relevant context to address the client's actual situation.
For me, responsible AI use with private information means considering what information is needed, which system I am using, how that provider handles business data and whether using AI provides enough value in the situation to justify providing that information.
Research, Sources & Fact Checking
AI can be exceptionally useful for research.
It can also be confidently wrong.
I therefore do not regard an AI answer itself as proof that something is true.
Depending on the subject and its importance, I may return to MLS information, government sources, statutes and regulations, lender documentation, program guidelines, municipal information, original studies, maps, manufacturer documentation, tax information, original reporting, my own previous research or other primary and specialized sources.
This becomes particularly important with information that changes over time, such as mortgage programs, grants, interest rates, insurance, taxes, housing-market statistics, MLS rules, local development, laws, forms and property-specific information.
Even real estate professionals actively using AI identify accuracy as a major concern: 63% of agents in RPR's 2026 survey cited accuracy of AI outputs as their leading concern.
AI can make finding, comparing and understanding information considerably faster.
It does not eliminate the need to determine whether that information is correct.
What AI Can Help With—and What Still Requires Me
AI can be part of my work for an individual client, but that does not make the AI system the client's real estate agent.
I remain responsible for understanding the client's objectives, determining what deserves investigation, evaluating results and applying useful information to the actual circumstances.
AI can help research a property.
It cannot independently walk through the property and observe what I see.
AI can help analyze negotiating considerations.
It does not conduct the negotiation or understand every circumstance and relationship surrounding it.
AI can help research financing possibilities.
It is not the lender determining what a buyer actually qualifies for.
AI can identify something that may deserve inspection, legal, tax, engineering or insurance attention.
It is not a full substitute for an appropriate inspector, attorney, CPA, engineer, contractor, lender, insurance professional, surveyor or other specialist.
And AI may fail to recommend a particular service, tool or marketing approach simply because that approach is too uncommon to appear in a generic answer.
AI can substantially assist with the work. It does not assume professional responsibility for that work. I do.
How I Disclose AI Assistance
Not every use of technology warrants the same level of disclosure.
Using spellcheck or receiving an isolated wording suggestion is different from having AI substantially research, analyze, organize, code or help draft a resource.
When AI materially contributes to the research, analysis, organization, drafting or substantial revision of something I publish, I generally believe disclosure is appropriate.
A typical disclosure may read:
AI Assistance Disclosure: This resource was developed with AI assistance through an iterative process involving human direction, original and/or external research, correction, revision and final editorial review. AI may have assisted with research workflows, organization, analysis, drafting, coding, critique, clarity, editing and/or visual development. Some material may also originate from my earlier articles, unpublished research, professional analysis or other original work that predates my use of AI. I remain responsible for the final content, sources, conclusions and publication decisions. See How I Use AI in Real Estate for additional information.
Some resources warrant additional explanation.
Others may need very little.
The purpose is not to make AI sound either better or worse than it is.
It is to give readers a more accurate understanding of how the resource was produced.
Who Is Responsible for the Final Product?
I am.
Using an AI system does not mean the company that developed that system has independently reviewed, verified or endorsed something I publish or recommend.
I choose what questions to ask.
I determine what information deserves additional investigation.
I decide which sources deserve weight.
I reject suggestions.
I correct AI.
I decide what ultimately gets published or communicated to a client.
And if something I publish is wrong, the fact that AI contributed to the process does not eliminate my responsibility to correct it.
Building on Work That Started Before AI
My use of AI is best understood as another stage in a process that began long before generative AI.
Before I used it extensively, I already had years of published real estate articles, private buyer and seller resources, unpublished Google Docs, spreadsheets, financing research, MLS research, housing-market analysis, professional experience, photography, aerial media, 3D tours, video, physical tools and other material.
I did not use AI to replace those years of work.
I increasingly use AI to help me do more with those years of accumulated research, professional information, writing, experience and original material—and to expand upon them.
Sometimes that means researching with AI before or during a transaction.
Sometimes it means standing in a house, noticing something specific and using my custom
buyer GPT to research that particular question.
Sometimes it means combining MLS information with AI-assisted analysis.
Sometimes it means using historical research an AI system would never see unless I supplied it.
Sometimes it means integrating an article I wrote entirely without AI.
Sometimes it means turning a six-page unpublished Google Doc into a public resource.
Sometimes it means creating an interactive market widget that would have been impractical for me to build on my own.
And sometimes the most valuable thing I contribute is recognizing that the generic AI answer is incomplete, wrong or missing an option because the way I approach the problem is itself unusual.
The strongest combination I see is not simply human or AI.
It is:
original research + professional information + real-world observation + specialized physical tools + human judgment + AI.
That is ultimately what interests me most about the technology.
Not replacing the work, but making considerably more of it possible.
— Adam Garrett
Sources & Further Reading
1. First Street — Flood Model Methodology. First Street describes its property-level flood model and the flooding types and property-specific factors incorporated into its risk estimates. First Street Flood Model Methodology
2. National Association of REALTORS® — 2025 Technology Survey. NAR reported that 46% of respondents used AI-generated content, with 20% using AI daily and another 22% weekly. NAR 2025 Technology Survey Findings
3. NAR / Realtors Property Resource — 2026 AI Survey. The survey of 225 NAR-member agents found that 92% were already using or planning to use AI and 63% identified accuracy as their leading concern. You've Tried AI, But Can You Trust It?
4. Coraly study reported by HousingWire — Digitally Altered Listing Imagery. The 2026 analysis examined just under 40,000 primary listing images on Zillow, Redfin, Realtor.com and Homes.com. It found indicators of digital alteration in 10.8%, with more than 90% of the altered images lacking visible disclosure. This finding concerns listing imagery specifically and should not be extrapolated to every other form of AI use. HousingWire: Most AI-Altered Listings Go Undisclosed
5. California Department of Real Estate — Artificial Intelligence in California Real Estate. The advisory covers responsibility for AI-assisted work, privacy and requirements applying to certain digitally altered real estate advertising images. California DRE AI Advisory
6. OpenAI — ChatGPT Business Privacy & Security. OpenAI states that Business data is excluded from model training by default and encrypted at rest and in transit; users within a Business workspace have their own chat histories rather than automatically seeing one another's chats. OpenAI Business Data Privacy, Security and Compliance
Related Examples From My Own Work
AI Assistance Disclosure: This article was developed with substantial AI assistance through an iterative process involving my direction, original and external research, integration of my existing published and unpublished work, corrections, revisions and final editorial review. I repeatedly corrected and refined AI-assisted drafts, including descriptions of my workflow, priorities, tools, media practices, privacy practices and use of AI with clients. The hero image and workflow infographic were AI-generated, and the sales-history chart was created with AI from data I supplied. I remain responsible for the final content, sources, conclusions and publication decisions.
Comments