What agents ask, answered plainly.
The questions sellers, buyers, and agents ask about AI assistants and real estate, answered as plainly as I can. If yours is not here, email me.
Do sellers really ask ChatGPT which agent to hire?
Yes, and the assistants answer with names. On October 15, 2026, asked who should list a $1.5 million home in one Pennsylvania town, ChatGPT named a specific agent, printed her Google listing card with a badge reading “My first call,” and cited two of her sales as evidence. Perplexity agreed and cited her Zillow profile. Both answered in under thirty seconds. How many sellers ask is a number nobody publishes. What the assistants say when they do is something I can show you.
Where do the assistants get their information about me?
Mostly from records, not from your website. In thirty answers for one agent, the sources were Zillow, Homes.com, Realtor.com, her brokerage's site, a regional newspaper's business directory, FastExpert, her own site, and one competitor's site. LinkedIn was cited once. Facebook and Instagram were cited zero times.
Why does the assistant name someone else when I sell more than they do?
Because it cannot verify your number and it can verify theirs. When one agent's homepage said she had the highest volume in town with no source, year, or measure, Perplexity rejected the claim in writing and picked a competitor. When the same claim was restated with a dollar figure, a year, and the business journal that published it, both assistants repeated it back as fact. The claim did not change. Its packaging did.
Do reviews matter to AI assistants?
More than anything else I have measured. On the best-agent question, ChatGPT ranked one agent third and gave its reason for first place in one line, a competitor's 370 transactions and 100 five-star reviews on Zillow. The agent it ranked third had 16 Zillow reviews, the newest from December 2021, against 26 team sales in the last twelve months. The ranking got her on the list. The review count decided the order.
Which review site counts most?
Zillow, in every answer I have recorded where an assistant explained its ranking by review count. Google reviews show up in the Google listing card the assistant prints beside its answer.
My brokerage handles my marketing. Does this still apply to me?
Yes. Your broker's roster page is one of the records the assistants read, and it is often the one with the oldest bio. The diagnosis tells you what that page says about you today and what to ask your broker to change. The questions about you are about you, not the brokerage.
I am on a team. Does the diagnosis cover the team or me?
Whichever name a seller would type. If sellers know the team name, I test the team and the lead agent both. One retest found Zillow's team layout assigning the lead agent's biography to a junior teammate, and an assistant printed it that way.
Can an assistant get facts about me wrong?
Yes, and it will tell the seller with full confidence. In one retest Perplexity described a junior team member as a fourth-generation native of the town, licensed since 2014. That was the lead agent's biography, read off the wrong card on Zillow. The fix was moving one paragraph so it opened with the right name.
Why does the diagnosis cost $600 when I can find an audit for less?
Because the cheaper reports are tool output, a crawl of your site scored against a checklist. The diagnosis is hand-run. I type the questions, read the answers, read the records, and write the fix list myself. You get what the assistant said in its own words, not a score.
What do I get that I can act on?
A fix list where every item names the record, what it says now, and what it should say. For one agent that was the Homes.com contact phone, the Google listing hours, the Zillow bio order, the broker roster merge, and the one sourced sentence to put on three pages. Most of it costs nothing but hours and her own logins.
Do you make the fixes?
If you want me to. The repairs are $1,800 over thirty days, and I rerun the original questions at day sixty to show what changed. Anything that touches your broker's roster or a claim about volume goes to your broker for sign-off first.
How do I know if the fixes worked?
The same questions, the same wording, sixty days later. One agent went from named in 12 of 22 questions to named in 16, and from first choice in 8 to first choice in 13, after record fixes alone. That is the only honest measurement I know of.
Does this work outside Pennsylvania?
Yes. The records are the same in every state. Zillow, Homes.com, Realtor.com, Google, your broker, and your state's license lookup. The questions change with your market and your specialties, which is why I write them from your own site.
What should I do before I buy anything?
Ask ChatGPT who the best listing agent in your town is and read the reasons it gives. One question is a useful start. The diagnosis checks a wider set and reads the records behind the answers.