AI Visibility for Property Managers: What Boards Ask and Whether ChatGPT Names You

A homeowners' association board that wants to change its property manager now does something that didn't exist two years ago: it puts its questions to an AI assistant. "Which property management company in Kiel also handles small buildings with eight units?" "Who offers an owner portal with digital access to invoices?" "What does association management cost per unit in Hamburg?" ChatGPT answers, names two or three companies, and the board writes to exactly those three.

Whether your firm is among them is not decided by the quality of your work but by whether your website answers those questions in a machine-readable way. Most don't. This article shows why, and what can be fixed in an hour. How this fits into real estate marketing as a whole is on Immobilienmarketing mit KI (German).

1. The five questions boards ask AI assistants

From conversations with property managers and from what I see in log data and check requests, the questions reduce to five patterns:

1. Territory and size: "property management in [city] for an association with [n] units". The machine needs your territory and your minimum or maximum size.
2. Scope of services: "Does the company also handle individual units / rental management / technical maintenance?" Without a list of services the answer stays "unclear".
3. Digitalisation: "Owner portal, online documents, digital owners' meetings?" In 2026 this is the most common deciding question.
4. Cost: "What does it cost per unit per month?" If you don't state a range you get quoted as "price on request", and that loses against the company that writes "from 22 euros per unit".
5. Trust: "Certified manager under section 26a of the German condominium act? Association member? Reviews?" Proof that exists but rarely appears on the website.

2. Why property managers' websites fail at this

The typical property manager's site was built in 2018, looks tidy and tells people everything they need. For a machine it has three holes:

• No company profile. There is no structured record saying: this is a business of type property management, at this address, covering this territory, with this phone number. An AI system has to guess whether it is looking at a manager, an agent or a portal.
• No questions and answers. The five questions above are answered literally nowhere. The information is scattered through prose, often in a PDF service catalogue that crawlers don't parse.
• No list of services. "We offer comprehensive care for your property" is not a data point. "Association management, rental management, individual unit management, technical building support", each with a paragraph, is.

I see the result in every check: technology good, Google mediocre, AI visibility below 20 points.

3. The four areas of the check, using a property manager as the example

The AI visibility check tests 20 points in four areas. This is what a typical result for a mid-sized management company looks like, anonymised but real:

• Technology & security: 90. HTTPS, mobile rendering, no outdated techniques. Only the homepage is too heavy at 2 MB because the team photo loads at full size.
• Google visibility: 45. Page title "Home", no meta description, no sitemap, no H1. Google finds the site via the company name and hardly otherwise.
• AI visibility (GEO): 15. No structured data, no identifiable business, no question-and-answer content, no llms.txt. On the plus side: AI crawlers are not blocked. The site may be read; it just says nothing.
• Legal & trust: 70. Imprint and privacy policy linked, address and phone found. Missing: reviews on the site and a preview when sharing on WhatsApp, which matters more than you'd think for board members passing links around a group chat.

Total: 52 out of 100. "Room for improvement". And the key insight: every point in the AI visibility area is writing work, not programming.

4. llms.txt, schema, robots.txt in one hour

Three files that lift a property manager's site from "silent" to "able to answer":

1. Check robots.txt (5 minutes). Make sure GPTBot, ClaudeBot, PerplexityBot and OAI-SearchBot are not blocked. Some agency templates block every bot except Google. Then you are invisible to AI systems regardless of what else is on the page.
2. Add schema (30 minutes). A LocalBusiness or, better, ProfessionalService node with name, address, phone, opening hours, areaServed (cities and districts) and sameAs (Google profile, association page). Plus Service for each service and FAQPage for the five questions. In WordPress this works through a schema plugin without code.
3. Create llms.txt (20 minutes). A text file at /llms.txt: who you are, what you manage, where, from what size, what it costs, how to reach you. In full sentences, without marketing. That is the self-description an AI assistant reads first when it finds one. Hardly any management company has it today.

After that: the five questions as a visible FAQ block on the homepage, set title and description, submit the sitemap. That is the rest of the day, and then the site reads to a machine like a fact sheet instead of a brochure.

5. Staff shortage: answering standard questions automatically

Property managers have a second problem that is connected to the first: the phone rings all day with the same questions. When is the annual statement coming, how do I report damage, where do I find the minutes. Each of these costs a case handler three minutes, and case handlers are the scarcest resource in the industry.

The same work that makes the site readable for ChatGPT solves this problem too: once you have answered your standard questions properly, as an FAQ on the website and as structured data, you can hand them to an AI assistant on your own site that answers owners around the clock and only passes through the cases that genuinely need a person. I have built this for clients; the effect is 30 to 40 percent fewer routine calls within a few weeks. The prerequisite is always the same: the answers have to exist somewhere first.

The honest conclusion

No management company loses a board because ChatGPT doesn't know it. It loses the board because it never makes the shortlist. The difference between "not mentioned" and "one of three" is, in this industry right now, an afternoon's work, because almost nobody has done it. That won't stay that way. Run the check today and in 30 seconds you know where you stand and what to change first.