What gets checked
What does AgentRank Checker look at?
Every check AgentRank Checker runs, grouped by category with the points each carries: crawler access, search foundations, structure, AI signals, content shape, real estate.
The checker runs 40+ checks on every site: exactly 43 in rubric v1.0, 41 of them scored on every real estate site and 2 that apply only when a Texas address or license appears. The five core categories carry 80 points, the real estate layer carries 20, and the Texas notices add 3 when they apply. Your AIQ is the share of applicable points you earned, and the whole run takes about 15 seconds.
Every row below is a question the report answers with evidence. Partial credit exists on purpose: a sitemap that exists but is not referenced in robots.txt earns two thirds of its points, and a robots.txt that allows 6 of 8 AI crawlers earns three quarters.
A. Can AI crawlers get in the door?
AI Crawler Access, 30 points. Whether GPTBot, ClaudeBot, PerplexityBot and friends can reach your page at all: robots.txt rules, firewall behaviour, HTTPS, page weight, and whether your words are in the HTML or hidden behind JavaScript.
| # | Check | Pts | Applies |
|---|---|---|---|
| A1 | Reachable by crawlers | 9 | every site |
| A2 | AI crawlers not firewalled | 5 | every site |
| A3 | robots.txt allows AI crawlers | 7 | every site |
| A4 | Content is server-rendered | 6 | every site |
| A5 | HTTPS | 2 | every site |
| A6 | Page size under 2.5 MB | 1 | every site |
B. Are the search basics in place?
Search Foundations, 19 points. The foundations Google relies on, which AI leans on too: a title, a description, a sitemap, a canonical, social cards, and the noindex switch that can hide a whole site.
| # | Check | Pts | Applies |
|---|---|---|---|
| B1 | Title tag (25–65 characters) | 3 | every site |
| B2 | Meta description (80–175 characters) | 3 | every site |
| B3 | Not blocked by noindex | 3 | every site |
| B4 | XML sitemap exists and is referenced | 3 | every site |
| B5 | Canonical tag is self-referencing | 2 | every site |
| B6 | OpenGraph title, description, image | 2 | every site |
| B7 | Twitter card | 1 | every site |
| B8 | html lang attribute | 1 | every site |
| B9 | Favicon | 1 | every site |
C. Can a machine tell what is what?
Structure & Semantics, 10 points. One h1, landmarks, a heading order that makes sense, alt text on images, a viewport tag, and nothing loading over plain http.
| # | Check | Pts | Applies |
|---|---|---|---|
| C1 | Exactly one h1 | 3 | every site |
| C2 | Landmarks (header, nav, main, footer) | 2 | every site |
| C3 | Heading hierarchy has no skipped levels | 2 | every site |
| C4 | Image alt-text coverage | 1 | every site |
| C5 | Viewport meta tag | 1 | every site |
| C6 | No mixed content | 1 | every site |
D. Are the AI answer signals there?
AI Answer Signals, 15 points. Structured data, an llms.txt file, an author with profiles, a real modification date, enough pages to matter, and prose that extracts cleanly.
| # | Check | Pts | Applies |
|---|---|---|---|
| D1 | JSON-LD structured data | 4 | every site |
| D2 | llms.txt at root, well-formed | 3 | every site |
| D3 | llms-full.txt | 1 | every site |
| D4 | Authority signals (author + sameAs) | 2 | every site |
| D5 | Freshness (dateModified within 12 months) | 1 | every site |
| D6 | Content depth (sitemap size + blog) | 2 | every site |
| D7 | Clean prose extraction | 2 | every site |
E. Is the writing quotable?
Content Shape, 6 points. A direct answer up front, real numbers, question-shaped headings, quotations and cited sources: the Princeton GEO findings, checked line by line.
| # | Check | Pts | Applies |
|---|---|---|---|
| E1 | Front-loaded direct answer | 1 | every site |
| E2 | Concrete stats and numbers | 2 | every site |
| E3 | Question-shaped headings | 1 | every site |
| E4 | Quotations | 1 | every site |
| E5 | Outbound citations | 1 | every site |
R. Who are you, and where do you work?
Real Estate Signals, 20 points. Agent identity schema with a license number, local business data, reviews that live in the HTML rather than a widget, contact details that match, community pages, market names and your brokerage.
| # | Check | Pts | Applies |
|---|---|---|---|
| R1 | Agent identity schema (name, title, license) | 4 | real estate sites |
| R2 | Local business schema (address, geo, area) | 3 | real estate sites |
| R3 | Review schema in server-rendered HTML | 2 | real estate sites |
| R4 | NAP consistency (visible vs schema) | 3 | real estate sites |
| R5 | Community / neighborhood pages | 2 | real estate sites |
| R6 | Market signals (place names in content) | 2 | real estate sites |
| R7 | Contact completeness | 3 | real estate sites |
| R8 | Brokerage identified | 1 | real estate sites |
T. Are the Texas notices linked?
Texas Compliance, 3 points. Only when a Texas address or license shows up on the page: the IABS and Consumer Protection Notice links in the footer.
| # | Check | Pts | Applies |
|---|---|---|---|
| T1 | TREC IABS link in footer | 2 | Texas only |
| T2 | TREC Consumer Protection Notice link | 1 | Texas only |
Where do the checks come from?
The crawler names and rules come from the published documentation of OpenAI, Anthropic and Perplexity, and from RFC 9309. The Content Shape checks are the three edits Princeton’s GEO study found most effective, which moved pages up to 40% in AI answers. The freshness check counts anything within 12 months as current, and the depth check wants a sitemap of at least 12 pages and a blog or guides section.
What does the report show without scoring it?
Three things are noted at the bottom of every report with no points attached: a video embed, an IDX or home-search tool, and an FAQ page in the navigation. Video is deliberately unscored because there is no solid evidence yet that it moves AI citations; the other two are just useful to know about. The grade ladder explains how the points become a letter, and the how it works page explains what gets fetched.