ChatGPT vs Perplexity vs Gemini: which AI recommends realtors

The example, and nothing beyond it

On one live Does AI Pick Me scan, a real estate agent in Redlands, CA was checked across ChatGPT, Perplexity, and Gemini. The free check asks three buyer questions and sends each question to each assistant. This run returned 9 answers, so every question came back from every assistant. The agent was named in 0 of those 9. The score was 0 out of 100. The most-named other agent appeared 2 times.

• Names stay off this page. Publishing the scanned agent would turn a private check into a public ranking. Publishing the other names would do the same to people who did not ask for the article. The counts are the point.
• Two appearances is not a market share. We are not saying that person appeared in 2 of 9 answers as a percentage you should quote, beyond the plain fact that they were named 2 times. We are not saying they were recommended, only that they were the most-named.
• Redlands is the market on the scan, not a sample of California. One agent, one run, one city. A different agent in the same city could look different the same day. We are not publishing a city average because we do not have one.

What was recorded

What 0 of 9 means in the formula

An answer that does not name the business scores 0. Nine answers at 0 average to 0, which is a score of 0 after multiplying by 100. The rank bonuses never apply, because there is no place in line when you were not named. A recommendation score of 0.85 never applies either. The zero is not a penalty for a failed call. The calls succeeded. The name was not in the text. How the score is calculated

• Per-engine scores are not in this write-up. We know all three engines returned answers and that the combined score is 0. We are not splitting the 0 into three engine trophies. A 0 on a complete set already says the name was absent in the answers we scored.
• The other agent’s two mentions do not change the 0. Competitor names are recorded so you can see who was quotable. They are not subtracted from you, and they are not added. Your score is only about whether you were named.
• A later run can move. This article does not claim the next scan of the same agent will also be 0. It claims this one was. Treat it as a log line with a date.

Why this cannot crown an assistant

A comparison that picks “the AI that recommends realtors” needs many agents, many cities, and a rule for ties. We have one agent and a zero. Declaring ChatGPT, Perplexity, or Gemini the winner would be a story pasted on top of an empty shortlist.

• The questions are the same on purpose. The free check uses “best real estate agent in [city],” “who should I hire to sell my home in [region or city],” and “top buyer’s agent in [city].” Your name is not in them. That is discovery. It is the harsh version of the test, and it is the version a stranger runs.
• The report asks more, including your name. The $49 report adds questions such as “[Your name] realtor reviews” and whether you are a good agent in the city. A zero on discovery does not tell you what those branded questions will say. It tells you the unbranded ones did not name you.
• Citations are not a personality contest. Perplexity often shows sources under an answer. ChatGPT and Gemini can too, depending on the product mode. In this article we are not grading whose citations look nicer. We are counting names.
• Consumer price is a different comparison. What a buyer pays for ChatGPT Plus, a Perplexity plan, or a Google AI plan changes their own account. It does not decide which agent is named. Those prices are omitted here so they are not mistaken for results.

How to read a zero without panicking

• Open the questions and read them out loud. If you would not ask them that way, note that, and still take the absence seriously. Buyers do ask “best agent in [city].” The wording is ordinary.
• Look at what was quotable about the people who were named. You do not need their names in a blog post to do this on your own report. You need the pages the answer leaned on: a profile with the city, a brokerage, and a sentence about the work.
• Fix your page before you buy a bigger monitor. A zero from nine live answers is enough to justify a first paragraph that states your name, your city, and your brokerage in text. It is not enough to justify a theory about which lab “prefers” agents.
• Repeat the check later. One scan is a snapshot. The product does not promise the next snapshot. When you change the page, give it time, then run the questions again and compare the log.

Which fits you

• You want to know if you are this kind of zero. Use the free check with your name and your city. Do not borrow the Redlands result. The report, if you want the written answers and the fixes, is $49 once.
• You already have a zero and you want the branded questions too. Those sit in the $49 report, not in the free check. They still do not guarantee a mention. They show what the assistants say when the question includes you.
• You wanted a winner among the three assistants. This scan does not provide one. If a future set of scans does, it will have to be labeled as a study, with a count of agents and a method. This page is not that.
• You are not an agent. The local-business scan uses trade questions, not agent questions. The point rule is the same. The prompts are not. Do not paste this 0 onto a plumber or a dentist.