Why 75% of Properties Are Invisible to AI | The Unlock
Zero Click to Home Part 1: Legibility
How can AI be so right about some things and wrong about others? It comes down to a word we’re going to be hearing about more and more: Legibility.
I recently went on a trip with a friend to Italy for a wedding in Sorrento. Neither of us expected a “Mediterranean summer” to entail scorching days and oppressive humidity, so after the wedding we sought somewhere cooler before heading back to New York.
My friend is not a tech guy and far from an early adopter. So I was surprised when he took out his phone, fired up ChatGPT and found an Italian town with cooler air and access to trail running, away from the tourist hordes.
ChatGPT’s recommendations were excellent. We ended up in Scanno, a picturesque town with crisp mountain air and fantastic trails, moderately famous for its photo-friendly steep alleyways but well off the beaten path.
Not every traveler has such a great experience. Earlier this month, three hikers had to be rescued from Mount Shasta after using AI to plan their hike. Apparently, the model underestimated how much food and water the hikers would need. They got lost and, instead of an eight-hour ascent, wound up spending the night on the mountain.
Oof.
This juxtaposition speaks directly to the problem plaguing the multifamily industry: most properties, and almost all of their individual units, simply aren’t legible to the robots.
What AI models can actually read
Early large language models generated answers based solely on training data; remember “I can’t answer this because my training data stopped in October of 2024”? As soon as the model was published, that was it. Now, ChatGPT, Claude, Gemini, etc. all have access to current information: the weather forecast, flight status, what’s open near you.
But they can’t see everything.
The difference between excellent travel recommendations and a poorly-planned hike comes down to what data is readable, or “legible,” to the models. For travelers like me who want to know about which attractions are packed with tourists, what each town is like, and what’s nearby, there’s an abundance of easily accessible content for AI models to choose from.
The same can’t be said for the information required to plan a long hike. If you went through the trouble to input the exact trail data file from Mount Shasta and a history of your hikes or individual fitness, ChatGPT could create a well-planned itinerary. Written content is out there on hiking Mount Shasta, but the trail conditions and exact topography (as well as the hiker’s individual abilities) aren’t.
AI is blind to this nuance… and ends up giving a general (poor) answer.
What happens when a renter asks AI for an apartment
ChatGPT knows a renter based on their other conversations, whether they cook a lot, lead an active lifestyle or care about the proximity to a grocery store and daycare. All of these get folded in by the model when a renter asks, “I’m looking for a 2-bedroom apartment in Buckhead, Atlanta. It needs to be close to work and I have a budget of $3,500.”
But is the information available?
For most communities, it isn’t. Across our research, 75% of communities never surface in AI answers at all, while the top 5% capture more than half of all mentions.
Multifamily properties are getting better about making community-level information accessible to AI: neighborhood detail, amenities, even unit availability. Yet almost nothing is posted about exact units. Is the apartment renovated? How big is the second bedroom? Does it look out into the courtyard? A park? Or a parking lot?
A picture or floor plan won’t do the trick. While LLMs can process image data today, the cost of sifting through pictures or floor plans for every single unit is too high. The models will always prioritize the cheapest, most specific, and most readable information available. For the most part, this means text, words. That’s why structured data, schema, and FAQs matter so much.
Unit-level content will decide the future AI visibility winners
As renters lean more heavily into AI for their apartment searches, unit-level detail will become the key to whether unit 807 shows up or not. That’s where Peek is putting its focus.
We’re excited to share the lessons we learn as we go, and appreciate you coming along on the ride.
See how legible your community is today
Peek Discover reads your community the way an AI crawler does and scores how often you actually surface across ChatGPT, Claude, Gemini and Perplexity. Then it tells you exactly what to fix, with the copy written for you.