10xSearch research methods • Updated November 12, 2026
Residential AI visibility research: source and method
This page describes the bounded evidence represented by the accompanying graphics. The charts are observational research, not product rankings or performance guarantees.
Current study — November 11, 2026
The current panel includes 125 publicly attributed real-estate websites: 25 per provider across Agent Elite, AgentFire, Lofty, Real Estate Webmasters, and Sierra Interactive. Each website received 12 non-branded question templates through three web-enabled API panels: OpenAI API / GPT-6.1 Sol, Claude API / Sonnet 5.5, and Perplexity Agent API / pinned medium / GPT-6 Luna. The reported panel contains 4,500 prompt units and 4,481 usable answers.
Read the 12 frozen question templates
- Who are the best residential real estate agents in {market}? Name specific options and cite your sources.
- I’m buying a home in {market}. Which real estate agent or team should I contact, and why?
- Who would you recommend for selling a home in {market}? Give specific names and sources.
- Which real estate teams in {market} have documented local expertise?
- I’m relocating to {market} and need a buyer’s agent. Who should I consider?
- Which {market} real estate agents have supported experience helping home sellers?
- Who can help me compare neighborhoods in {market} before buying a home?
- Recommend local real estate agents in {market} with a strong reputation and explain your sources.
- Which agents or teams in {market} can help with pricing and marketing a home for sale?
- Who can help me understand the current housing market in {market} before I buy or sell?
- Recommend real estate agents in {market} known for strong negotiation. Cite supporting sources.
- Which local agents or teams in {market} should I compare before choosing a realtor?
- Name mention
- The business or an approved alias is named in an answer.
- Recommendation
- The business or approved principal is presented as an option in an answer after review.
- Audited-domain citation
- The audited website domain is cited or linked in the answer.
Attempt-ledger context
The separate November 11 expanded-study ledger records 5,811 non-branded prompt units: 1,500 each for OpenAI, Anthropic, and Perplexity, plus 1,311 for Gemini. This ledger is reported separately from the current three-panel result denominators.
Historical studies remain date-specific
The Luxury Presence (June 2026) and Agent Image (July 2026) studies used four engines with mixed web and memory modes. They are displayed as dated historical snapshots and are not pooled with the current API panel.
Homiere snapshot
The Homiere November 5 snapshot covers 25 sites / 24 businesses, two web-enabled API panels, and 600 non-branded answers; it reports name matches and audited-domain citations, not recommendation counts.
Interpretation limits
- The sample is observational and non-random. Public website attribution identifies the cohort; it does not establish paid-service scope or causation.
- Markets, business scale, website age, and available evidence vary across sites.
- API tests are not consumer app sessions. Results can vary by model, retrieval, location, timing, and prompt wording.
- A response-level rate is not a forecast of leads, rankings, revenue, or a universal market result.
For exact values, open “View exact numbers” beneath each figure. The public package deliberately excludes raw answers, private evidence, local paths, and internal workflow records.