Background, Methodology & Initial Results

The Big Picture

  • What we’re studying: How generative AI platforms recommend home service businesses compared to traditional search.
  • Why it matters: Search engines remain dominant, but AI is increasingly used for symptom triage, project planning, and budgeting—moments that can lead directly to referrals.
  • What’s next: We’ll test these scenarios across leading AI systems on a recurring basis, tracking shifts over time as the technology evolves.

Background

Generative AI technologies are rapidly becoming more advanced, and consumers are increasingly turning to chat-based systems as alternatives to traditional search. While search engines remain dominant for explicit service needs (e.g., “plumber near me”), recent research shows that AI excels in more complex or directive tasks — scenarios where a homeowner starts with a symptom and is guided toward professional help.

For example, studies of Bing Copilot and ChatGPT highlight that users lean on AI for knowledge-intensive, multi-step tasks rather than quick fact-finding【1】. Nielsen Norman Group’s 2025 study found that AI tools act as “shortcuts” to information, changing how people approach discovery, even if they haven’t yet replaced search engines【2】. At the same time, reports in The Wall Street Journal and Washington Post confirm that while AI-based search traffic is growing, it still represents a small fraction compared to Google【3】【4】.

Applied to home services, we see three emerging categories of use cases where AI may influence business recommendations:

  1. Direct service intent – Explicit queries such as “plumbers near me” or “HVAC companies in Washington, DC.”
  2. Symptom triage → escalation – Starting with a problem (“my thermostat just stopped working”) that may escalate to calling a professional if DIY steps fail.
  3. Project planning and budgeting – Broader goals such as “I want to remodel my bathroom” or “How much does it cost to replace an HVAC system in 2025?” where homeowners may not know which service categories to search for, but AI can map outcomes to service providers.

These patterns matter because they shape when and how AI recommends local businesses — whether by surfacing national directories (Angi, Yelp, Thumbtack), pointing to franchises, or referring users to licensed professionals in generic terms.

Methodology

To study these shifts, we will be testing a set of scenarios across the most widely used AI platforms: ChatGPT, Claude, Windows Copilot, Perplexity, Gemini and Grok. Each represents a different approach to AI-assisted search and recommendation: some are deeply integrated into operating systems and search engines, while others position themselves as direct search alternatives.

In parallel, we will run the same scenarios on major search engines and map platforms to provide a baseline for comparison:

  • Google Search – capturing ads, the Local Pack, and organic results.
  • Bing Search – relevant both as a standalone search engine and as the foundation for Bing Copilot.
  • Google Maps – to evaluate how location-based searches surface providers in a map-first environment.
  • Bing Maps – to compare Microsoft’s local listings ecosystem.

Each scenario will be tested under two conditions:

  • Cold tests – A single, initial prompt or query to measure how the model or engine responds without conversational context.
  • Hot tests – A more interactive back-and-forth that allows the AI to walk through troubleshooting steps and potentially escalate to recommending a professional.

All tests will be conducted from the perspective of a homeowner in Washington, D.C., a major metropolitan area with dense competition among home service providers.

Because this is a longitudinal study, we will repeat these tests every few months or whenever a major platform update occurs (such as a new model release). Our methodology will remain flexible: we may run partial test sets, introduce new use cases, or refine the protocol as the landscape evolves.

About the Results

We will be publishing high-level findings in these Labs posts, focusing on patterns, trends, and takeaways. The raw transcripts and full dataset are considered proprietary and will not be made publicly available.

Select data may be shared by request and on a limited basis, particularly with research partners and industry stakeholders interested in deeper analysis.

References

  1. Suri, et al. Generative Search for Complex Tasks: A User Study of Bing Copilot. arXiv
  2. Nielsen Norman Group. How AI Is Changing Search Behaviors (2025). NNG
  3. Wall Street Journal. AI Search Is Growing More Quickly Than Expected (2025). WSJ
  4. Washington Post. The Myth That Chatbots Are Replacing Google Search (2025). WaPo