Can AI Select Good Investment Properties?

AI’s capabilities are powerful these days and are advancing at an astonishing pace. I asked myself, can I, or any investor, soon use AI to select good investment properties?

AI has no magic powers, nor can it produce useful answers from nothing. AI will always give you an answer, but whether the answer makes any sense depends upon its knowledge base and its training. It’s the old trash-in and trash-out situation.

Let’s try a simple property selection.

Suppose your goal is high initial cash flow and you do not care about anything else. In this case, AI could work today with the right prompts. Below are two prompts and the responses from GPT5.

Prompt: “Provide a list of the top 10 US cities with a metro population in excess of 800,000 and with significant population declines over the last 10 years.”

Two of the cities it provided:

• New York Metro Area: Lost nearly 550,000 residents since 2020
• Philadelphia Metro Area: Population fell by over .4% last year; continuing a longer-term population decline.

Prompt: “What are the median single-family home prices in the New York metro area and the Philadelphia metro area?”

What it provided:

• New York Metro area – median single family home price: $818,000
• Philadelphia metro area – median single family home price: $250,000

So, if your only goal is initial cash flow and you do not care about anything else, buying a low cost single-family home in Philadelphia is possibly a good option. AI is excellent for looking up data, but the human has to ask the right questions and interpret the results.

What about using AI to select investment properties that attract our target tenant segment?

It all falls down because it is not just looking up data. It requires interpretation of data. To do what we need, AI will must make selections based on the historical behaviors of the historical tenant segment we target. What do I mean by historical behaviors? Some examples:

• The lot size must be greater than 3000 ft.²
• The property must be a one story or two story. No three story or split-levels.
• The property must be greater than 1000 ft.² but less than 2500 ft.²
• It cannot be near high-voltage power lines.
• There cannot be nearby nuisances like noise from highways or busy streets.
• The driveway must be long enough so that an average length car can fit on it and still have about 1 m to the sidewalk.
• Can not have bad floor plans or be in subdivisions that do not rent well.

Our data-mining engine evaluates each property against about 40 criteria based on our research into our target tenant segment behaviors. Unless it meets all 40 requirements, we will not consider the property.

Even if I provided the AI with all 40 or so “rules,” it wouldn’t be enough. Take paint and flooring choices, for example. A home with black carpet, black walls, and black ceilings (yes, I’ve actually seen this once) would not rent for as much as a similar property with neutral colors (if it would rent at all).

The real problem is that we’re asking AI to make subjective decisions based on a specific tenant segments’ preferences. At this time, AI can’t make subjective evaluations because there’s no formula or data set that predicts what our tenant segment will find desirable, or how the subject property is likely to perform. What we know of our target demographic preferences comes from years of research and is based on (evolving) behavior patterns, not rules.

Even if we found a way to assign a numerical value to desirability, we’d face another issue: Property photos don’t always accurately depict a property’s current condition. Sometimes, the home is in much worse condition than the photos show, or there are questionable neighbors. This is why we have to go on site before we have a good assessment of a property. Even if the MLS photos are current, there’s still no way to know from pictures if the home has strong pet urine or smoke odors.

As for automated rent and price estimates? Rentometer predicted a rent range of $1,900 to $2,200 per month for this property.

This is just one example where inadequate data resulted in erroneous results. An AI would do no better.

Another problem is the quality of the available data. Recently, a ~1600 ft² property listed 62 bathrooms. That is obviously false information.

What AI Is Good For

Research

I was writing a paper that needed a chart showing personal income growth in Clark County, Nevada (Las Vegas). I wrote a prompt, and ChatGPT provided multiple options. I chose a chart from the St Louis Federal Reserve. In the past, it would have taken me at least a few hours to (maybe) find the data. With AI, it took me minutes.

Software Development

Our business relies heavily on custom software. Custom software requires a lot of time to write, test, and maintain. AI has completely changed how I write software. Instead of starting with a blank screen, I now give GPT a prompt and it does the first draft. An example prompt: “Write a Python function to find all images in “pathToDirectory” that do not have an aspect ratio of approximately 16×9”

Below is the code that GPT5 produced:

def find_non_16x9_images(pathToDirectory):
    bad_images = []
    for root, _, files in os.walk(pathToDirectory):
        for file in files:
            if file.lower().endswith((".jpg", ".jpeg", ".png", ".gif", ".bmp", ".tiff", ".webp")):
                filepath = os.path.join(root, file)
                try:
                    with Image.open(filepath) as img:
                        width, height = img.size
                        # Check ratio: width/height vs 16/9
                        if abs((width / height) - (16/9)) > 1e-3:
                            bad_images.append(filepath)
                except Exception as e:
                    print(f"Skipping {filepath}: {e}")
    return bad_images

What used to take me days to accomplish now can be done within a few hours. This is how I write most code today.

Improved Writing Efficiency

I write several articles and or posts every day. Before AI, finishing a single article could take me an entire day, mostly because I struggled with flow, grammar, spelling, verb-tense match, run on sentences and more. Now, I draft a paragraph, then use a prompt I’ve developed (and keep refining) and have (Usually GPT5) AI rewrite it. More often than not, the rewritten version reveals that I didn’t explain my ideas as clearly as I thought, or it sounds like a robot wrote it. It is rare that I can use the AI rewrite, but it speeds up my writing tremendously.

Conclusion

AI is an exceptional tool for increased productivity. It excels at tasks that can be accurately and reliably defined by algorithms. However, since there are no algorithms that can reliably predict human behaviors (such as tenant preferences), I do not see how AI can do a good job of selecting good investment properties anytime soon.

Important Legislative Update

This summary reflects my understanding of the legislation and is provided for general information. It is not legal advice and not a legal interpretation. Laws change, and situations differ, so please consult a qualified attorney if you have questions or concerns. Here is a link to the actual legislation.

Nevada’s AB 121 legislation takes effect October 1, 2025. It changes how rent is displayed on the MLS, Zillow, etc. and in your owner portal, and it slightly increase the management fee (for now.)

What changed

AB 121 requires all mandatory tenant fees to be included in the advertised rent.

Before AB 121

If the rent was $2,000, listings showed $2,000. Sewer and trash, $25 each, appeared as separate line items. On the owner portal, you saw three line items: rent, sewer, and trash. The 8% management fee applied only to the $2,000 rent, not to sewer or trash.

After AB 121

All mandatory fees must now be included in the advertised rent. In this example, the listing will show $2,050/Mo. AppFolio, the software most property managers use, cannot currently separate mandatory fees from rent, so the 8% management fee is applied to the full $2,050. This will add $4/Mo to your management fee. This limitation affects all Nevada users of AppFolio, and I expect they will enable the fee calculation to exclude mandatory fees in the future. I will keep you posted.

AB 121 Will Affect All Listings

Including mandatory fees will increase the advertised rent for all listings. During the transition, a property may look slightly less competitive, but that will fade as all managers comply with AB 121.