How R.E. Immo Became the #1 Most-Mentioned Property Manager Across 5 AI Engines
A four-month case study on beating competitors 7× larger across ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews. Here is every number, and how each one was measured.
invisible
mentioned
SMARTYPANT has been a creative, innovative and reliable partner. I was genuinely surprised by what they achieved in only four months.
Where it started
R.E. Immo GmbH manages roughly 6,650 residential and commercial units across the Rhein-Neckar region and Baden-Württemberg. Work started on 11 May 2026. The July baseline was 12 buyer questions across four platforms, 48 answers.
Findable if you already knew it existed, close to absent if you did not. That was never a content problem. The company had operating history, regional depth and a working site. Its pages simply were not being selected as evidence when an assistant assembled an answer.
What we actually did
Not one impressive screenshot. A loop, run against more than 40 real buyer questions and the actual regional competitors.
Audit demand
Run the buyer questions across all five AI surfaces. Record citations, mentions, positions, and which sources won instead.
Read the sub-questions
Assistants split a question into their own searches before answering called fan out queries. We compressed between 90 to 120 such queries and used them in the content brief.
Publish and fix
13 new service, location and comparison pages, plus around 80 on-page fixes to titles, descriptions, schema and image metadata on pages that already existed.
Retest
Same questions, same platforms, same competitors. The result decides what gets built next.
The expertise was already on the site. It needed to become retrievable.
The site runs on WordPress with Yoast, which was serving one default pattern across the whole site. Every page technically had a description and a schema block, and every one of them said roughly the same thing, so no page declared what it was actually for. We read each page individually, worked out its subject, and set its title, meta description, schema type and JSON-LD for that page alone. The same for every image on the site. Around 80 changes, and not one of them was a template rule applied everywhere. Our field report CITED. covers why generic schema produces no measurable lift while populated, page-specific schema does.
Two markets, both led
For a fixed set of buyer questions we counted how often each company was presented to the reader as an option, using the same extraction method for every company on both sides.
Size lost to structure
Treubau Verwaltung is the established name in Mannheim. By its own published figures it manages around 45,000 apartments plus 7,000 garages and parking spaces, with over 130 employees.
In July, on the question of alternatives to Treubau, ChatGPT, Claude and Gemini all cited Treubau and R.E. Immo appeared nowhere. The audit logged it and recommended one comparison page.
0 of 4 platforms
Named or linked R.E. Immo on the Treubau-alternatives question. Absent everywhere.
3 of 5 platforms
Cited on the same intent, including position 1 of 5 in Google AI Overviews, the only AI Overview citation in that entire run.
The strongest competitor in that panel was not Treubau either. It was Gräfer Hausverwaltung: roughly twelve employees, based in Leimen, no Mannheim office, beating Treubau three to one. What it has is a network of city-specific pages and consistent directory presence.
Size and age lose to location-specific service pages plus directory presence. We found the pattern from the outside, on a company that is not our client, before we had finished building it from the inside.
It held up when we repeated it
AI answers change from run to run. A single favourable response proves nothing, so the case study is built on repeated panels rather than one good screenshot.
2 Sept
2 Sept
7 Sept
7 Sept
11 Sept
Top-three placement held between 61% and 73% across runs. First place did not: it swung between 20% and 44% on the identical panel hours apart. So we build claims on top three and treat first place as weather.
This is a visibility result, not a revenue claim. No inquiries, calls or revenue have been attributed to this work. Google's AI surfaces report impressions only, with no click or query data, and there is no clean way today to tie an AI-sourced visit to a signed mandate. What it shows is that R.E. Immo became far more likely to enter the sources and shortlists that shape an AI-assisted buying decision.
What search data says
AI panels are one source. Google Search Console is another, covering pages rather than answers. Every page below was published during the project, so its starting point is a true zero.
Why phrasing decides everything
The most useful thing we learned had nothing to do with R.E. Immo specifically. Same topic, same target pages, two phrasings:
This is why content plans built from keyword volume miss. The assistant is not searching your keyword, it is searching the sub-questions it generated. Answering one buyer question about Mannheim, ChatGPT ran twelve searches of its own, one of them checking IHK and VDIV membership. No property owner types that. The trust pages are what satisfy it, and those pages never show a visible link.
One more measurement worth stating plainly: across all three September runs there was not a single case of a re-immo.de page being retrieved and then not used. When the page was reached, the model used it. The constraint was reach, never content quality.
How it was measured
Six measurements between July and September 2026, by two independent methods.
No platform is consistently strongest. Claude led one Rhein-Neckar test at 55% and was the weakest in Stuttgart at 27%. Gemini did the reverse. That is why single-platform screenshots are worthless and only the pooled figure is worth quoting.
AIO is the weakest and least measurable surface. It failed to return an overview at all on roughly half the questions in every test, which is why its denominators are so small. Its 7% direct-link rate also understates it: AIO featured R.E. Immo in both 7 September tests on the Treubau-alternatives question, once as a direct citation at position 1 of 5, and once by naming the company first of five providers without a link. Against twelve July questions that produced zero citations and zero mentions, either is a change.
Definitions and exclusions
- Cited means a clickable re-immo.de URL appeared in the answer's source list. Named means the company appeared in the answer text. A page can ground an answer without the name showing, and a name can appear with the link pointing to a directory instead. We counted both, separately.
- Valid responses exclude answers where Google returned no AI Overview and answers where the model replied from training data without searching. Neither gives any company a chance to be cited, and the exclusion was applied identically to every competitor.
- Position figures use answers that actually listed providers as the denominator, never total responses. Roughly half of all answers explain a topic instead of naming companies.
- We do not compare the two audits' composite scores. The weighting changed between them, so the comparison would be meaningless in either direction.
- Individual questions swing hard. One city question went from three citations to one in a single afternoon on the identical panel. Anyone showing you a single AI answer as proof is showing you a coin flip.
- The September measurements fall inside the Baden-Württemberg summer holidays, which ran from 30 July to 12 September 2026.
What this means
R.E. Immo did not become a bigger company in four months. It became easier for AI systems to retrieve: clearer about who it serves, where it operates, what it is certified to do, and which page answers which buyer question.
You do not need to be the biggest company in your market. You need to be the clearest, most useful and most verifiable answer to the question your buyer actually asked.
The work is not finished. R.E. Immo runs several services across six locations, and full coverage is a matter of quarters rather than weeks. Directory presence is the other open channel: a handful of directories feed AI answers far more heavily than the rest, and the brand is still missing from one of the biggest. Every new page gets measured before the next ones are chosen.
Frequently asked questions
How do you measure whether AI assistants mention your business?
Build a fixed panel of the questions your buyers actually ask, run it across ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews, and record two things separately for every answer: whether a clickable link to your site appeared in the sources, and whether your name appeared in the text. Then repeat the identical panel. Single answers fluctuate heavily, so only the repeated aggregate means anything.
How long does it take to become visible in AI answers?
Four months here, from 8% to roughly 35% direct citations. Individual pages took four to eight weeks to move from publication to appearing as a cited source, and several pages published in the final weeks had not been picked up yet when we measured.
Does company size decide who AI assistants recommend?
No. The region's established competitor, managing around 45,000 units against R.E. Immo's 6,650, received 3 mentions to R.E. Immo's 18. The strongest competitor in that same panel has roughly twelve employees. What those companies had was location-specific service pages and directory presence, not scale.
Why do good guide pages sometimes never get cited?
Because the question does not trigger retrieval. Broad informational questions are often answered from training data with no search run at all, so no source is ever selected. The same content can win when the question carries commercial and regional intent. One question in our panel scored zero on every platform at two separate timepoints; rephrased, it produced citations on three of five.
Is AI visibility the same as SEO?
They overlap but are not the same. An AI answer is assembled from sources retrieved for sub-questions the assistant generates itself, not for the query the buyer typed. A page can ground an answer without your name appearing, and your name can appear with a link pointing to a directory instead of you. Both have to be measured separately.
Can a case study like this prove revenue impact?
Not honestly, not yet. Google's AI surfaces report impressions only, with no clicks or queries, and there is no clean way to tie an AI-sourced visit to a signed mandate today. This case study claims visibility and competitive position, both measured directly. It does not claim inquiries or revenue, and you should be sceptical of anyone in this market who does.
Do AI assistants name your business, or your competitors?
We run the questions your customers actually ask, across all five platforms, and show you who gets mentioned, which sources win, and what to fix first.
Check your AI visibilityNot ready yet? Read CITED., our free field report on how AI picks which companies to name.