Generative AI Use Cases In Real Estate Marketing: Adoption and Constraints in Germany

Scientific Impact Assessment

Core Contribution

This paper offers a qualitative, exploratory account of how German real estate agents (REAs) currently use generative AI in marketing, based on eleven semi-structured interviews conducted mid-2025. Its central deliverable is a "grounded map" of current and potential GenAI use cases, organized via Gioia-style inductive coding into aggregate dimensions (marketing communication, general communication, property analysis, acquisition preparation, general knowledge support, sales) and mapped onto the established real estate transaction flow. The main empirical findings are: (1) marketing communication dominates adoption, with exposé-text writing the only broadly established use case; (2) interaction is predominantly human-in-the-loop (GenAI drafts/structures/retrieves; agents curate/verify/decide); and (3) the binding constraints are integration, data availability, and compliance (GDPR, EU AI Act) rather than model capability. The paper positions itself as extending prior U.S. sentiment work (Kriegbaum et al.) to a more fragmented, heavily regulated German context, shifting the analytic lens from adoption sentiment to task-level workplace practice.

Methodological Rigor

The methodology is appropriate for an exploratory qualitative study and is transparently reported: recruitment (convenience/snowball), sample characteristics table, coding procedure in two cycles, and interview structure are all described. The authors are commendably honest about limitations — they explicitly disclaim theoretical saturation, note that percentages are "descriptive only" and not adoption rates, and acknowledge that the AI-user-only sample cannot explain non-use or resistance. However, the rigor is inherently bounded. Eleven interviews averaging 29.6 minutes is a thin evidentiary base; several use cases rest on a single agent's mention. There is no reported inter-coder reliability, no member-checking, and no triangulation with observational or document data. The sampling bias (only active AI users) structurally inflates the sense of adoption. These are standard weaknesses of small-N interview studies rather than fatal flaws, but they cap the strength of any generalizable claim.

Potential Impact

The impact is likely modest and localized. As a first empirical descriptive account of GenAI in German REM, it fills a genuine gap and could serve as a reference point or motivating citation for subsequent work in real estate informatics, information systems, and professional-services digitalization. The concept maps (Figures 1–5) are a useful taxonomic scaffold that follow-up quantitative surveys or technical pipeline builders could operationalize. The higher-level generalization — that GenAI adoption in professional services hinges on task standardization, data accessibility, output verifiability, and compliance — is plausible but not novel and is asserted rather than demonstrated. The paper does not produce a method, dataset, benchmark, or tool that others would directly reuse; its contribution is a descriptive snapshot with a short shelf-life given the rapid evolution of GenAI capabilities and adoption.

Timeliness & Relevance

Highly timely. GenAI adoption in vertical professional domains is an active question, and empirical (rather than speculative/conceptual) evidence is genuinely scarce, as the related-work section credibly documents. The regulatory framing (EU AI Act, GDPR) is current and relevant. The single-country, single-vertical focus is well-motivated by the institutional specificity of German real estate (land registers, energy certificates, notarial processes). That said, the fast-moving nature of the topic means findings risk dating quickly.

Strengths & Limitations

Strengths: Clear motivation and well-scoped research questions; honest and repeated caveating of claims; a coherent and legible taxonomy; a genuinely under-studied context; sensible practical implications (low-risk experimentation in marketing, caution in screening/personal-data tasks). The human-in-the-loop framing and the distinction between "GenAI-solved" vs. "GenAI-enabled" workflows are analytically sensible.

Limitations: Very small, self-selected sample; no reliability/validity checks beyond author workshops; findings are descriptive with no theory-building or theory-testing beyond confirming prior U.S. observations; the interpretive claims are explicitly labeled "author interpretation" and "tentative." The paper contributes no reusable artifact, no quantitative measurement, and no methodological innovation. Reproducibility is limited by the qualitative design — transcripts and coding schemes are not shared, so independent replication is impossible, though the coding approach is standard and specified. The arXiv date (2026) and reference set suggest recency, but the evidentiary weight remains light.

Overall

This is a competent, honest, exploratory qualitative study that makes a legitimate but incremental contribution: a first descriptive map of GenAI use in a specific national professional niche. It is useful as a scoping reference and motivation for follow-on quantitative and technical work, but it does not advance methods, produce durable artifacts, or generate surprising or field-shifting findings. Its impact will likely be limited to a modest slice of the real estate informatics / IS professional-services literature.