Measurement & Reporting · 9 min read · July 15, 2026

Measuring mentions: which AI prompts surface your real estate agency — and where competitors are ahead

When a seller asks ChatGPT "Which agent in Regensburg sells older apartments reputably?", visibility today gets decided inside the AI's answer — not on page one of Google. Measuring mentions means systematically checking which prompts surface your agency, in what context, and where competitors get named ahead of you.

Why mentions are the new currency for real estate agents

You used to measure your visibility with a single number: your Google ranking for "agent + city." That world is fading fast. More and more owners and buyers now ask their first question not in a search box, but of ChatGPT, Gemini, Perplexity, or Google's AI Overview. The answer comes back as a finished piece of prose naming two or three agencies — and everything else effectively doesn't exist for that reader.

For you as an agent this means your visibility is no longer a ranking, it's a mention. Either the AI writes something like "An established contact for terraced houses in Freiburg-Vauban is agency XY" — or it doesn't. There's no middle-ground result on position seven you can still catch with a bit of scrolling. You're mentioned, or you're invisible.

That's why measuring mentions isn't a gimmick — it's the foundation of any honest GEO strategy. If you don't know which prompts you already appear in, you're optimizing blind. Before you change a single sentence on your website, get a baseline: where do you stand in the answers today, and where do your competitors stand?

The right questions: how sellers and buyers actually ask

The most common mistake in measuring is testing the wrong prompts. Almost nobody types "real estate agent Munich" into ChatGPT. People talk to AI in full sentences, describing their actual situation: "I need to sell my inherited house in Nuremberg-Zabo, but I need someone who understands communities of heirs." Those are the questions you need to collect — not abstract keywords.

Build your list from four question types. Recommendation questions: "Which agent do you recommend in Leipzig for apartment buildings?" Comparison questions: "Is Engel & Völkers or a local agent better for my condo?" Problem questions: "Who do I turn to for a forced sale in Dortmund?" Process questions: "Who can help with the sale and notarization of a semi-detached house in Potsdam?"

Add variants for your location, neighboring towns, your specializations, and your property types. A solid starting list runs 30 to 60 questions. This list is your measuring instrument — keep it stable for months so you can actually compare change over time. Don't keep rewriting the questions, or you'll measure noise instead of progress.

How to run a mention check, step by step

Take your list and run each question through at least three systems: ChatGPT, Google Gemini, and Perplexity. Log out or use a fresh, non-personalized session, or your own search history will skew the results. You want to see what an unfamiliar prospect sees, not what the AI shows a known, logged-in agent.

For every answer, log four things in a simple table: were you mentioned (yes/no)? In what position (first, middle, last)? In what tone (recommended, mentioned neutrally, cited only as an example)? And which competitors showed up? Those four columns are your entire dashboard. Perplexity also shows its sources — look closely at which pages it cites, because those are your levers for improvement.

Repeat the run monthly, with the same list, ideally on the same weekday. AI answers fluctuate, so a single snapshot tells you little. Only repetition reveals patterns: which questions you consistently appear in, which you never do, and where something shifted after you made changes. Log the date and the model version too, since updates can move results noticeably.

Where competitors are ahead — and why

If a neighboring agency gets mentioned more often, it's rarely luck. Usually that competitor has three things AI systems favor: a large volume of consistent reviews on Google and ImmoScout, editorial mentions in local press, and a website that states clearly, in text, what they actually do. AI pulls its knowledge together from exactly these kinds of sources.

Analyze your competition by evidence, not gut feeling. Ask the AI directly: "Why do you recommend agency XY for apartment sales in Cologne?" It will often name the reasons — long market presence, many reviews, a specialization in one district. That's honest, free feedback about your gaps. Note which selling point competitors get credited with in the answers.

Some gaps you won't close in four weeks. An agency with hundreds of reviews and two decades of press coverage has a real head start. But you'll almost always find niches where nobody is strongly positioned — "agent for barrier-free apartments in Wuppertal" or "farm sales in the Allgäu." There, you can become the named name while the big players stay generic.

From measurement to action: what to do next

The measurement is worthless if you draw no conclusions from it. Take the questions where you don't appear and check: does the matching answer even exist anywhere in text form on your website? If the AI doesn't know your agency for "sell inherited house community of heirs Kassel," you probably have no clear page on that topic. That's exactly the page to write next — concrete, with location, problem, and your approach.

Work on the sources AI draws from. Systematically ask satisfied sellers for Google reviews with concrete context ("sold our apartment in Bogenhausen in six weeks"). Detailed reviews like that carry far more weight with AI than a bare star rating. Keep your listings on Google Business Profile, ImmoScout, and industry directories completely consistent — same name, same address, same specializations, everywhere.

After each change, wait and measure again the following month. Is the mention moving? Sometimes it takes two or three cycles before models pick up your new content. That lag is normal. What matters is the trend over six months, not the jump from one week to the next. That's how guessing turns into a controllable process.

Tools or manual work? An honest look

Specialized GEO monitoring services now exist that query your questions automatically and track mentions daily. For a single agency, these tools are often overkill and expensive. A manual monthly check with about 40 questions and a spreadsheet costs you maybe two hours — and in the process you'll understand far more about how AI talks about your market than any dashboard could show you.

If you want to automate, start small. A simple spreadsheet with a monthly history column is enough to get going. Only once you're managing several locations or a larger team, and the manual effort becomes a real burden, does a paid tool start to pay off. Don't buy monitoring software before you've measured manually for three months and learned what actually matters for your market.

Tool or manual work, the value isn't in the measurement itself — it's in the decisions that follow. An expensive dashboard nobody acts on is money wasted. An honest spreadsheet that leads to one concrete new page or review campaign a month is worth far more.

The most common measurement mistakes

Mistake one: checking only ChatGPT. Your customers use different systems, and the answers vary a lot between them. Perplexity cites sources, Gemini draws on Google's index, ChatGPT leans more on training knowledge. If you only measure one system, you get a distorted picture and risk optimizing for the wrong audience.

Mistake two: measuring while logged into your own account and history. The AI then recognizes you and mentions your agency more favorably — a completely false signal. Always test in a neutral, logged-out session. Mistake three: measuring once and drawing conclusions from it. A single run is a snapshot with a lot of randomness baked in. Only a repeated rhythm makes the data reliable.

Mistake four: getting annoyed at a bad mention instead of using it. If the AI describes your agency incorrectly or cites an old specialization, that's a signal your public sources are outdated. Fix it at the root — on your website and in your profiles — instead of blaming the model. The AI is only mirroring what it finds about you.

Your roadmap for the next 90 days

Week one: build your list of around 40 realistic prompts drawn from your area and your specializations. Run the first mention check across ChatGPT, Gemini, and Perplexity, and log everything in your table. That's your baseline, and everything else gets measured against it. Take your time here — this foundation carries all the work that follows.

Weeks two through eight: work through your three biggest gaps. Usually that means a missing topic page, too few detailed reviews, and inconsistent directory listings. Write the pages, collect reviews with real context, and clean up your profiles. Focus on questions where you can realistically move the needle, not the single most contested search term in your city.

Day 90: run the full mention check again with exactly the same list. Compare it line by line against your baseline. Where do you appear now that you didn't before, where are you still missing, and how has the tone changed? This comparison shows you in black and white whether your work is paying off — and turns a vague sense of AI visibility into a real, controllable metric for your agency.

Common questions

How often should a real estate agent measure their AI mentions?

Once a month is the right rhythm. Measure more often and you'll mostly be tracking noise, since AI answers shift slightly from day to day. Use the same list of questions, the same weekday, and a logged-out session every time, so results stay comparable across months and you can see real trends instead of chance.

What do I do if a competitor keeps getting mentioned ahead of me?

Ask the AI directly why it recommends them. It usually cites reasons like a large volume of reviews, long market presence, or a clear specialization. That tells you where your gaps are. Instead of attacking the big players head-on, claim a niche — a district, a property type, a selling situation — where you can become the clearly named agency.

Is it enough to check only ChatGPT, or do I need several systems?

You need several. Your customers use ChatGPT, Gemini, and Perplexity in a mix, and the answers differ noticeably between them. Perplexity even shows you its cited sources, which is genuinely useful for your optimization work. If you only measure one system, you'll get a distorted picture and end up optimizing for the wrong behavior.

Read on

Local & Industries