AI Search Readiness Report: How 52 US Real Estate Brokerages Score (2026)

Buyers and sellers have started asking ChatGPT, Perplexity and Google's AI answers which brokerage — and which agent — to call. Those answers are assembled from whatever the AI crawlers can read. So on 2026/09/04 we pointed our free AI-visibility checker at the homepages of 52 of the largest and best-known US brokerages and franchisors, from Compass and Redfin to regional independents, and scored what the crawlers actually see.

The headline numbers#

• •52 sites tested · 30 scored · 22 blocked the checker (HTTP 403) · 0 unreachable
• •Grades among the 30 scored: 1 A, 3 B, 11 C, 7 D, 8 F — median 55, mean 52
• •Top scorer: Michael Saunders & Company (85/A). Bottom: Sotheby's International Realty (18/F), Lamacchia Realty (18/F)
• •AI crawlers allowed in robots.txt: 97% · llms.txt present: 23% · any JSON-LD: 60% · Organization schema: 50% · RealEstateAgent schema: 3% · FAQ schema: 0% · answer-first opener: 10%

The full ranking#

Sorted by score. "Blocked" means the site returned HTTP 403 to our checker, so no score could be produced — see Finding 1 for why that matters. Click any report link to re-run the checks live.

Finding 1: The bouncer problem — 22 sites block unfamiliar crawlers#

The single biggest result isn't a score. 22 of 52 sites (42%) — including Keller Williams, Berkshire Hathaway HomeServices, Howard Hanna, Long & Foster — refused our checker with a 403 before it read a single line of HTML. Our crawler identifies itself honestly and asks for one page; it was turned away by bot-management rules (the Cloudflare / Akamai / Imperva layer most large sites run) that challenge any user agent they don't recognize.

That is exactly how the AI crawlers arrive: as unfamiliar user agents named GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot and Google-Extended. A robots.txt that says "welcome" doesn't help if the firewall in front of it says "no". We can't prove from a 403 that every AI crawler is blocked on these sites — bot rules can allowlist specific agents — but we couldn't verify anything else either, and neither can a crawler that's been bounced.

• •Ask your web vendor or IT team to allowlist GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot and Google-Extended at the WAF / bot-management layer (not just robots.txt).
• •Re-run the checker afterwards. If it can read your homepage, so can they.

Finding 2: The front door is open — the house is empty#

Among the 30 sites we could read, robots.txt is almost never the obstacle: 97% allow every major AI crawler (the exception: Engel & Völkers Americas (3/13 blocked)). What's missing is the content AI systems actually lean on.

• •llms.txt — the plain-text "who we are and what matters here" file AI models look for — exists on just 7 sites: Michael Saunders & Company, Compass, Latter & Blum, @properties, Realty ONE Group, JB Goodwin, Fathom Realty.
• •12 sites (40%) ship no JSON-LD structured data at all: JB Goodwin, NextHome, Douglas Elliman, Weichert, Century 21, Better Homes and Gardens Real Estate, ERA Real Estate, John L. Scott, Engel & Völkers Americas, LPT Realty, Sotheby's International Realty, Lamacchia Realty.
• •RealEstateAgent schema — the one that lets an AI answer "who's a good brokerage in X" with your own data — appears on 1: Michael Saunders & Company. FAQ schema: none.
• •Only 10% open the homepage with a sentence that says who they are and who they serve — the kind of line an AI can quote.
• •10 sites (33%) look client-side rendered or nearly empty to a crawler that doesn't run JavaScript: Fathom Realty, Christie's International Real Estate, Century 21, Better Homes and Gardens Real Estate, ERA Real Estate, John L. Scott, Engel & Völkers Americas, LPT Realty, Sotheby's International Realty, Lamacchia Realty. To GPTBot that page is close to blank.

Finding 3: National franchise brand sites score worst#

The pattern in the bottom of the table is franchisor brand sites. 5 of the 6 brand sites owned by Anywhere Real Estate scored D or F — Century 21 (26/F), ERA Real Estate (26/F), Better Homes and Gardens Real Estate (26/F), Sotheby's International Realty (18/F), Coldwell Banker (51/D) — with Corcoran (67/C) the exception. Meanwhile the highest scores came from independents and regional firms: Michael Saunders & Company (85/A), Compass (74/B), Latter & Blum (74/B), @properties (72/B). One reading: a brand-portal homepage built to route visitors to franchisees carries little citable substance of its own, while a firm that sells its own market expertise puts that expertise on the page.

A second pattern worth noting: Latter & Blum, which runs on Compass's platform, scored 74/B — identical to Compass itself (74/B). Your platform sets your floor. Two of the three llms.txt-bearing leaders, Michael Saunders & Company, Compass, are on custom builds.

What the top scorers do differently#

Michael Saunders & Company (85/A) is the only A. It passes every check except faq schema, open graph, headings. Compass (74/B) and @properties (72/B) both publish an llms.txt, ship JSON-LD and render server-side — and both lose the same points: no RealEstateAgent schema, no FAQ schema, no answer-first opener. In other words, even the leaders leave 15–25 points on the table, and those points are the cheapest ones to earn.

What to fix first, in order of points#

• •Get past your own firewall. If the checker gets a 403, nothing else matters — allowlist the AI crawlers at the bot-management layer, then re-test.
• •Add structured data: Organization + RealEstateAgent (name, areaServed, address, ratings) + FAQPage. Together these are the single largest block of points in the score.
• •Publish an llms.txt at your root: who you are, the markets you serve, links to your key pages.
• •Open the homepage with one plain sentence an AI can quote: who you are, where you work, who you help.
• •Make sure core content is in the initial HTML, not rendered later by JavaScript — most AI crawlers don't execute it.

Methodology and limits#

• •Tool: our free AI-visibility checker — 17 weighted checks across discovery (llms.txt, robots.txt, sitemap), structured data, content and rendering. Every check and its weight is visible in each live report.
• •Run: 2026-09-04, from a US cloud region, homepage only, one request per site with an honest, self-identifying user agent.
• •Sample: 52 large and well-known US brokerages and franchisors chosen by us — not a random or complete sample of the industry.
• •A 403 means the site refused our crawler; we report it as blocked rather than guessing a score. It does not prove every AI crawler is blocked.
• •Scores measure readiness to be discovered and cited, not whether AI assistants currently mention a brand.
• •Sites change. If your score has moved, re-run the live report — and if we've got something wrong, email us and we'll correct it.