AI Doesn’t Kill the ILS. It Makes It More Valuable.

The Front Door Is Moving, But the House Remains the Same

We’re entering an era where LLMs are functioning less like websites and more like new operating systems. Think of them as the iOS of the 2020s. In this landscape, the ILS isn’t being replaced. It’s becoming an app inside a new platform.

We don’t say Apple Pay killed credit cards just because we stopped swiping physical cards. The transaction still runs through Visa. The interface changed, but the infrastructure didn’t.

That’s exactly what’s happening in rental search. Instead of fighting for traffic against chat interfaces, major players are integrating directly into them. By launching the Zillow app inside ChatGPT, the marketplace ensures that when a renter starts a conversation about apartments, the trusted experience remains Zillow’s. The front door starts in a chat window, but the marketplace keeps the relationship, the interface, the data, and most importantly, the conversion.

Think about what Apple did with the iPhone. They didn’t just build a phone. They built the operating system, then opened the App Store so an entire ecosystem could build on top of it. That’s the playbook unfolding right now. LLMs are becoming the new operating system for how consumers find information, and ILS platforms are becoming the apps that power apartment search inside of them.

Why AI Needs a “Truth Engine” to Actually Work

Here’s where the “AI replaces everything” argument falls apart. LLMs are prediction engines, not truth engines. They’re designed to predict the next word in a sentence, not to verify whether a unit is still available at 2PM on a Tuesday.

They hallucinate and they can’t verify. They don’t have real-time access to structured rental inventory data. And in the high-stakes world of housing, a hallucinated price or a phantom listing loses consumer trust and can be a legal liability.

Think about what a chatbot can actually get from scraping a property website versus what an ILS gets through direct PMS integration.

Here’s what that gap actually looks like

LLMs guess. ILS platforms know.

A chatbot can read a property website. It can’t book a tour at 3PM on Thursday without calling an PMS API, which an ILS can do. It can summarize marketing copy, but it can’t see into the direct API integrations with Yardi, RealPage, Appfolio, and Entrata that power real transactions.

The ILS acts as the infrastructure layer underneath the chat interface. Direct PMS integrations provide real-time plumbing. Normalized data schemas organize messy, unstructured data from tens of thousands of properties into a consistent format the AI can actually use. And data refreshes measured in minutes ensure the concierge never recommends a unit that was leased an hour ago.

The concierge is only as helpful as the accuracy of their information.

From Search to Match. The End of the Directory Era.

The directory era was defined by passive browsing. The AI era will be defined by high-intent matching. This is the transition from filtering (removing what you don’t want) to matching (finding exactly what you do).

The old model had a renter searching 500 listings and filtering down to five. The AI model has a renter describing what they need and the system surfacing the best three. That doesn’t kill the ILS. It redefines what the ILS actually is. When you stop thinking of it as a directory and start thinking of it as a structured data repository with up-to-the-minute inventory, pricing, and availability, the value becomes clear.

What AI kills is the directory experience. Endless scrolling. Generic filters. What it strengthens is the engine underneath. Curation. Recommendation systems. Concierge-level matching. The ILS stops being the place renters visit and starts being the reason the AI can give a good answer in the first place.

The Behavioral Moat

There’s a critical distinction between general world knowledge and renter knowledge. AI has broad understanding across many topics, but it lacks the decades of behavioral data required to understand how people actually make housing decisions.

An ILS doesn’t just see the listing. It sees the lease. For 20+ years, these platforms have tracked not just what renters say they want, but what they actually sign for. That gap between stated preference and actual behavior is where the real intelligence lives. This proprietary data turns the system into a compromise engine, predicting tradeoffs that no amount of floorplan analysis can replicate.

Historical transaction data shows that a family will almost always sacrifice square footage if it means access to a higher-rated school district. The ILS knows which features actually drive a signed lease in a specific submarket. It knows that for a certain renter profile, a two-bedroom with a particular den layout outperforms the three-bedroom every time, based on thousands of prior leases.

An LLM can read a floorplan, but it can’t predict human compromise. That requires historical transaction data, not scraped text.

The Trust and Compliance Moat

What happens when an AI chatbot recommends an apartment that violates Fair Housing guidelines? Or hallucinates pricing and a renter shows up expecting a different rate? Or books a tour at a property that’s fully leased?

The answer is legal exposure, regulatory scrutiny, and consumer backlash.

AI platforms like OpenAI and Google don’t want to become real estate compliance engines. They don’t want to be Fair Housing auditors or legal guarantors. They want to plug into companies that already are.

Zillow has already extended its Fair Housing safeguards directly into the ChatGPT experience. That includes Fair Housing classifiers built into the AI layer, ongoing testing with OpenAI, and verification flows that route back to Zillow before any action occurs.

The ILS becomes the verified source of rental search. In a world of AI-generated noise and fake listings, the verified source gets more valuable, not less.

Lead Volume Trends

If AI were truly disintermediating the marketplace, we would see a collapse in ILS performance. We’re seeing the opposite.

Despite the meteoric rise of generative AI, Zillow has reported that multifamily direct leads increased significantly in 2025 vs 2024. Average leads per property grew significantly too.

This is because AI is expanding the top of the funnel. By creating more starting points for discovery, whether through a search bar, a chatbot, or a voice assistant, AI is driving more traffic toward the essential execution layer.

Chat is the new browse mode. But action still needs a marketplace. Renters start in ChatGPT, powered by Zillow’s real-time listings and insights. When they’re ready to tour, contact, or apply, they flow into Zillow.

AI isn’t cutting off the marketplace. It’s expanding the top of the funnel.

The Infrastructure Era

The ILS is undergoing a metamorphosis from destination to infrastructure. In the coming years, the ILS will function much like Visa does for payments or Twilio does for communications. A critical, often invisible layer powering the rental market.

In this new API economy, even if a renter never visits an ILS website directly, the ILS infrastructure powers the transaction every time an AI assistant needs access to the truth. By 2028, I predict the traditional destination experience will continue to decline, but the ILS infrastructure will power 80%+ of all AI-driven rental searches invisibly.

The ILS will move from being a website you visit to the invisible operating system that lives inside every AI assistant.

My thesis isn’t that ILS survives despite AI, it’s that ILS thrives as the data and execution layer AI depends on.

AI changes the front door. It doesn’t burn down the building.

Credit to my friends at Apartment List and Zillow, who both provided helpful information in my research on this topic.