What does AI cite when people search for a real estate agent or broker?

On 2026-09-26 we sent five real estate prompts, one run each, to gpt-5.1 with its web search tool on, and recorded every URL the model attached to its answer. We sorted each cited host by type with our own rules: an agent's, team's or firm's own site, a directory, listings or review site, a news, reference or government site, Reddit, or social. We chose the buckets; the model chose the hosts.

The prompts covered the main hiring decisions in the category: best real estate agent in Toronto to sell a condo, mortgage broker recommendations in Vancouver, property management company in Calgary for a rental, real estate team in Chicago for first-time buyers, and cash home buyers in Dallas that are legitimate.

Toronto condo agent: five citations, to mytorontocondo.com and yolevski.com, two agent sites, storeyteam.ca, a team site, sothebysrealty.ca, a brokerage, and threebestrated.ca, a directory. The answer named a few agents and teams and told the seller to ask concrete questions and not to choose on commission alone.

Vancouver mortgage broker: six citations, to bcbudget.gov.bc.ca, a provincial budget page, rate-my-agent.com and reviews.birdeye.com. The answer named one broker, referred to others that appear in best-of lists without naming them, and pointed the reader to the BC Financial Services Authority for licensing.

Calgary property management: seven citations, all to rentfaster.ca pages, a rental listings platform. The answer named five property management companies drawn from those listings. Chicago first-time-buyer team: nine citations, all to team or brokerage sites, including ainhoagarcia-co.com, benlalez.com, hartyrealty.com, home-collective.com, kehoegroupchicago.com, revgrouprealestate.com, startingpointrealty.com, theboulevard.group and vestapreferred.com, and the answer named those teams with their brokerages.

Dallas cash home buyers: nine citations to bbb.org, listwithclever.com, rei-ink.com and reviews.birdeye.com. The answer named one local buyer and spent most of its length on how to avoid pressure tactics, check the BBB and state records, and compare net offers.

Where the question was who to hire as an agent or team, the model cited agents' and teams' own sites heavily and named the people whose pages it read. Where the question involved trust or licensing, mortgage brokers and cash buyers, it reached for the regulator, the BBB, rating sites and review aggregators, and hedged its names. Property management was answered from a listings platform, which is where management companies' rental inventory is visible.

By our sorting, 24 of 36 citations were own sites, the highest share of any industry in this study, and the Chicago team prompt alone supplied nine of them.

For agents and teams, the result is direct: the model cited team websites and named the teams it read, so a site that states the neighbourhoods, property types and buyer or seller focus plainly is the page that gets an agent into the answer. For mortgage brokers, the model went to the regulator and rating sites, so a broker's licensing should be easy to verify and the rating and review profiles should be current. For property managers, the rental listings platform was the source, so the company's presence and naming on the platform matter.

For cash buyers and anyone else where trust is the question, the BBB profile and review platforms were what the model read, and its answer was built around warnings; a company in that space should expect to be judged by those pages. SearchPod builds agent and team pages for real estate clients, keeps profiles consistent, and tracks which prompts cite them with a fixed prompt panel.

This is one model on one date, with each of five prompts asked a single time. Change the day, the model or the wording and the citations would change too, and the sorting is our own. The agents, teams and companies the model named are recorded only as what it said; nothing on this page verifies or endorses any of them.