The State of AI Visibility in Real Estate.

A benchmark of what web-grounded AI assistants actually say when consumers ask who the best real estate agents are.

A new discovery layer is deciding who gets considered.

Home sellers and buyers increasingly ask an AI assistant to name a good agent, much as they once opened a directory or asked a friend. The assistant does not return the thousands of licensed agents in a metro. It returns a handful of names, sometimes none, and a different handful depending on which assistant the consumer uses.

For consumers, that means the agents shown are a narrow and inconsistent slice of the real market. For agents and teams, it means a new form of discovery is quietly influencing who enters the interview set. This study measures that recommendation surface directly instead of treating AI visibility as a vague marketing claim.

Nearly one in five answers named no one.

In 24 of the 128 answers captured, or 18.8%, the model returned no nameable agent or team. It offered general advice about choosing an agent instead. An empty recommendation surface is not merely a low ranking; it means there was no shortlist to appear on.

What we measured.

• Markets: Atlanta, Boston, Denver, Houston, Kansas City, Las Vegas, Miami / South Florida, Minneapolis-St. Paul, Nashville, Orlando, Philadelphia, San Diego, San Francisco Bay Area, Seattle, St. Louis, Washington DC / Northern Virginia.
• Models: ChatGPT GPT-4o Mini Search, ChatGPT GPT-4o Search, Perplexity Sonar, Perplexity Sonar Pro. All were web-grounded configurations.
• Questions: a general best-agent query and a high-end seller query, phrased as a consumer would ask them.
• Sample: 128 answers yielding 422 capture rows: 398 named mentions and 24 empty-answer rows.
• Correlation: Spearman rank correlation between AI Visibility and each available public proof field in the 252-row scored corpus. The anonymized paired values and per-metric sample sizes are available in the downloadable validation file.
• Identity: only captured names reconciled to a known candidate identity were included in the scoring input.

Limitations: This is a 16-market snapshot using two question families. Model answers change over time. Citation analysis describes what the model disclosed, not every source it may have used. The study is aggregate and does not publish an expansion-market leaderboard.

Independent measurement, not paid placement.

The Real Estate Agent AI Index measures which agents and teams AI assistants surface by market and model, then compares that visibility with source-backed production and review evidence. Rankings move only through the published methodology. Placement cannot be purchased.

Questions about methodology or a factual correction can be submitted through the public correction workflow. No outreach or commercial relationship changes the figures in this report.