AI-based search on homepage & result list

As part of a broader innovation initiative, ImmoScout24 explored the integration of AI into the core search experience across web and mobile.

The goal was to understand how emerging AI technologies could support discovery, express innovation and create future-facing product capabilities, even though there was no explicit user demand for AI-driven search at the time.

Search is the central entry point of the platform and already highly optimized through a mature, filter-based logic.

AI Search · Homepage

Challenge

From a UX perspective, the challenge was not whether AI could be introduced, but how it could be integrated responsibly and without disrupting a well-established and trusted search experience.

Key constraints included:

No strong user pull for AI-based search
A complex, highly tuned filter-driven search already in use
High expectations around innovation and experimentation

The main risk was introducing novelty without delivering meaningful user value.

AI Search · Result List

My Role & Responsibility

UX strategy and interaction design for the AI search initiative
Translating an exploratory innovation goal into concrete, testable UX solutions
Designing coexistence models between classical and AI-driven search
Planning and evaluating usability tests and A/B experiments
Aligning UX insights with Product, Engineering and Data teams

A key part of my role was balancing innovation goals with user trust and usability.

Initial Approach

The first implementation focused on integrating AI into the existing search logic.

User prompts (text or voice) were translated into the established filter system
AI primarily acted as an alternative input method rather than a fundamentally new search paradigm
Voice input was introduced as an optional interaction layer

This approach allowed fast experimentation while keeping the underlying search behavior consistent.

AI Search · Filter Cockpit

User Research & Insights

Usability testing revealed clear patterns:

Most users preferred the classical search with explicit filters
AI search was perceived as interesting, but not essential for task completion
Voice input had very low perceived usefulness for real estate search
Users valued clarity, predictability and control

These insights highlighted that AI should support, not replace, user decision-making.

Key UX Decision

Usability testing revealed clear patterns:

AI search would remain optional and be introduced as a beta experience.

User should:

Choose between classical and AI-driven search
Switch freely between both modes at any time
Continue to rely on familiar filters and controls

This approach minimized risk, preserved trust and enabled learning without forcing behavioral change.

Iteration & Validation

To improve the experience and deepen understanding, we:

Introduced optional follow-up questions to validate whether AI correctly interpreted user intent
Used this feedback to refine prompt-to-filter translation
Measured adoption and engagement through A/B testing

The beta setup provided valuable qualitative and quantitative insights.

Outcome & Impact

Adoption of AI search remained limited compared to classical search
Core search KPIs were not negatively affected
The initiative generated valuable learnings about AI UX in complex decision-making domains
Engineering teams gained hands-on experience with generative AI-based systems

While not a mass-adoption feature, the initiative successfully served as an experimentation and learning platform.

Key Learnings

Innovation initiatives do not always start from user demand
UX responsibility includes protecting established, high-performing experiences
Optional and reversible integrations are essential when value is uncertain
AI UX requires transparency, control and trust to succeed