She Saved $24,000 Using AI Instead of a Real Estate Agent. He Used ChatGPT to List His House and His Agent Says He Gave Away $40,000. Both Stories Are True. Here Is How to Be the First One.

The Florida buyer's name is not public, but the transaction is. She used Homa's AI-powered home buying platform — built by a former Zillow senior director, trained on state-specific real estate law, property records, and licensing exam coursework — to navigate her purchase from search to closing without a buyer's agent. The Homa platform handled property search, negotiation guidance, automated contract generation, and closing coordination. In Florida, where the average buyer's agent commission is 2.78%, the savings on her transaction exceeded $24,000. This was the first fully AI-guided home purchase in the United States. It worked.

The Miami seller's story is different in a specific way that matters. He listed his home using ChatGPT for the listing description and pricing research — no agent. Within 72 hours he had five offers and told a reporter the process 'exceeded expectations.' The buyer's agent on the opposite side of that transaction — Elinor Solomonoviz, with Avanti Way Realty — and her broker Ines Hegedus-Garcia, told the National Association of Realtors what the seller did not know: a home that attracts five offers in 72 hours is almost certainly underpriced. A skilled listing agent, with direct experience of buyer demand in that specific neighborhood, would have known the property would generate multiple competing offers — and would have structured the listing and bidding process to extract maximum value. The AI research that priced the home did not have access to that ground-level signal. The seller likely left more money on the table than a full commission would have cost.

Both stories are true. The difference between them is not the quality of the AI — it is the type of transaction and what AI can and cannot know. This guide explains the exact line.

What Changed in 2024-2026: The NAR Settlement Shifts the Calculation

In August 2024, the landmark NAR settlement went into effect requiring that buyer's agent compensation be negotiated, disclosed, and agreed to before a buyer tours a home. For the first time in decades, buyers are having explicit conversations about what they pay their agent and what they get for it. Two years into the settlement, two-thirds of agents report their commissions have not meaningfully changed, according to a Cotality and ResiClub survey of 213 agents. But the framework has changed: the permission to question, negotiate, and potentially opt out of the traditional buyer's agent structure now exists in a way it never did before. That permission — combined with AI tools that are now genuinely capable of handling significant portions of the research and document work — is what made transactions like the Florida buyer's possible.

What AI Does Well in Real Estate: The Specific Tasks

• Neighborhood research — AI's strongest real estate capability: Perplexity Pro with live web search can synthesize recent crime data, school ratings, walkability scores, commute times to specific employers, flood zone risk, historical home price appreciation by zip code, and planned zoning changes in a single 30-minute research session. This is work that previously required hours across multiple disconnected sources. Fundrise launched RealAI in January 2026 — trained on 3.5 trillion real estate data points, every property in America — making institutional-grade neighborhood and property analysis available to individual buyers for the first time at $69/month.
• Contract and disclosure review: Claude's 200K token context window can review an entire purchase agreement and disclosure package — which in California can exceed 100 pages — flag unusual clauses, identify missing standard protections, and explain every section in plain English. This does not replace a real estate attorney for complex transactions. It does allow a buyer to enter attorney review fully informed, saving legal fees and reducing the risk of missing something important.
• Mortgage and financing comparison: AI can compare mortgage products, calculate the total cost of ownership across loan types, explain the break-even analysis for paying discount points, model how different down payment amounts affect monthly payments and long-term cost, and help you understand what your pre-approval letter actually means. The caveat: AI provides educational analysis — lock actual rate quotes through a licensed mortgage broker.
• Comparative Market Analysis research: Perplexity can pull recent comparable sales, average days on market, list-to-sale price ratios, and current offer competition levels in specific zip codes. This data informs your offer strategy even if the final pricing judgment requires human experience of local market dynamics.

What AI Cannot Do Well — Where the Miami Seller Went Wrong

The Miami seller's pricing mistake was not a failure of AI capability — it was a failure to understand what AI can and cannot access. ChatGPT can pull recent comparable sales. It cannot know what a buyer's agent who shows homes in that specific neighborhood every week knows from direct experience: which streets have parking issues that suppress offers, which school feeder zones create buyer urgency, how this specific floor plan competes with three others that hit the market in the same week. A comparative market analysis from AI is a reasonable starting point. It is not a substitute for the ground-level market intelligence that comes from hundreds of real transactions in that specific area. For a property that will attract multiple offers — which is precisely when pricing strategy matters most — that missing intelligence is expensive.

California added a specific legal complexity in 2026 that sellers using AI tools need to understand. AB 723, effective January 1, 2026, made undisclosed AI-altered listing photos a misdemeanor — not a fine, a misdemeanor. A real estate agent in Kelowna, British Columbia was already fined for failing to disclose AI virtual staging. NAR's 2026 Code of Ethics requires disclosure of all AI-generated or AI-enhanced content in listings. HAR MLS in Houston requires a watermark on every digitally altered photo. If you use AI to generate or enhance listing images, descriptions, or other marketing materials, you must comply with these disclosure requirements — which vary by state and MLS system. Non-compliance carries legal risk that the AI tool itself will not flag.

The AI Real Estate Toolkit: Matched to Your Situation

The Decision Rule: Which Story to Be In

The Florida buyer and the Miami seller each made the right decision for their specific situation — and only one of them knew it. The Florida buyer was purchasing a property at a fair market price in a transaction that did not require dynamic pricing strategy. The AI-guided platform gave her all the support she needed at a fraction of the traditional cost. The Miami seller was in a market with active buyer demand that was driving above-list offers. He needed someone who could read that demand and monetize it — and he chose not to hire one. The $24,000 saved in Florida was real money. The amount left on the table in Miami was likely larger.