Zillow CEO argues AI agents won’t replace home search as AI Mode reaches all signed-in users

Overview

Zillow Group Chief Executive Officer Jeremy Wacksman used the company’s quarterly earnings call to reject the market assumption that autonomous AI shopping agents will inevitably bypass consumer property portals. As reported by GeekWire, Wacksman argued that searching for a home is an experiential, high-stakes decision and not an administrative task that buyers want delegated. “Finding a home isn’t a chore to hand off to a robot; it’s a milestone people want to live themselves,” he said, comparing home search to watching a movie or playing a game, where “the enjoyment is the point.”

The comments came in response to investor questions about personal agents such as Meta’s recently launched Muse and Perplexity’s Comet browser agent. Investors wanted to know whether these tools would conduct property searches on behalf of users and cut portals out of the transaction. Alongside his response, Wacksman reported three operational AI milestones: full rollout of Zillow’s conversational AI Mode, new agentic features in its Follow Up Boss software for real estate agents, and a 60% increase in code changes shipped to production compared with 2024.

This is a United States technology and property story, not an environmental one. It still matters to Australian environmental consultants and their clients because it sets out a clear line between tasks that people will hand to a machine and decisions they want to make themselves, with a trusted adviser. A growing range of AI tools promise to automate desktop searches, data review and reporting, and some are aimed at the property and development sector. Zillow’s approach, and the evidence it presented, offers a useful reference point for deciding where those tools belong in a contaminated land workflow.

Key details

Zillow’s AI Mode, a conversational real estate search assistant, has moved from preview to 100% of signed-in users. Between March and September, the tool logged more than 8.5 million queries from 2.5 million users. According to Zillow’s production data, users who engage through AI Mode spend nearly three times longer on the site, view more homes and submit more tour requests than users of traditional search. Zillow presents this as evidence that an interactive assistant, which keeps the user in control of the search, performs better than an autonomous agent working out of sight.

Zillow reported that approximately 80% of its traffic still arrives directly at its own website and apps. The company maintains integrations with OpenAI and Google Gemini and is onboarding Anthropic, but third-party AI platforms currently account for a negligible share of inbound home shoppers. On Zillow’s own figures, a shift of consumers to off-platform agents has not yet shown up in its traffic, despite the release of capable autonomous agents by several large technology companies.

On the professional side, Zillow is deploying predictive agentic features in Follow Up Boss, its customer relationship software for real estate agents. These features identify high-intent leads and draft recommended actions, but a human agent approves each action before it is sent. Internally, Zillow said AI coding tools have allowed its engineering teams to ship 60% more code changes to production than in 2024, with improved software quality. All of these metrics are company-reported on an earnings call and have not been independently verified.

The remarks were made during a wider dispute over whether autonomous agents should be allowed to transact on consumer platforms at all. Amazon has moved to block Meta’s Muse agent from shopping on its platform without prior agreement and is in litigation over browser-based shopping agents, including Perplexity’s Comet. Some platforms that earn revenue from human attention, advertising and lead generation, Amazon among them, are putting technical and contractual barriers in place against agents that would carry out transactions off-platform.

Australian context

For Australian contaminated land practice, the distinction Wacksman draws between chores and milestones maps closely onto the split between data processing and professional judgement. The chores include tabulating laboratory results against the National Environment Protection (Assessment of Site Contamination) Measure 1999 as amended in 2013 (NEPM 2013) health investigation levels, health screening levels and ecological investigation levels, checking chain of custody records, running QA/QC calculations such as relative percent differences, and formatting reports. These tasks are well suited to automation because the inputs are structured and the outputs can be checked. The judgement calls are different in kind. Whether a site is suitable for a sensitive use such as a childcare centre, or whether a finding triggers a duty to notify under section 60 of the Contaminated Land Management Act 1997 (NSW), the Environment Protection Act 2017 (Vic), the Environmental Protection Act 1994 (Qld) or section 83A of the Environment Protection Act 1993 (SA), carries legal and financial consequences that clients expect a qualified person to own.

The access dispute between Amazon, Meta and Perplexity also has a local equivalent. Environmental due diligence in Australia depends on searches of public and commercial data sources, including the Queensland Environmental Management Register and Contaminated Land Register, the NSW EPA record of notices and list of notified sites, the EPA Victoria Priority Sites Register, the SA EPA public register, historical title and aerial photography services, and laboratory client portals. Each of these sources has its own terms of use, search conventions and limitations. Firms building automated search tools on top of them should expect the same questions about permitted access that are now being argued between large technology platforms, and should not assume that today’s access arrangements will last.

References and related sources

• Primary source: www.geekwire.com
• marketbeat.com
• trendmicro.com
• https://www.geekwire.com/2026/zillow-ceos-reality-check-on-ai-agents-finding-a-h
• NEPM Assessment of Site Contamination

How iEnvi can help

iEnvi integrates technology and data-driven approaches into environmental consulting. We monitor AI and technology developments that affect how environmental professionals deliver services to clients.

• Our capabilities
• About iEnvi
• Talk to the team

This is an iEnvi Machete news summary. Prepared by iEnvi to summarise the source article for environmental professionals tracking AI, data, and technology developments that affect consulting and project delivery.

Published: 05 Nov 2026

Need advice on this topic? Speak to an iEnvi expert at info@ienvi.com.au or 1300 043 684, or contact us online.