# GEOrank (by Stainaz) - full reference > GEOrank (https://georank.stainaz.app/) is a Generative Engine Optimization (GEO) and AI search visibility audit workspace built by Stainaz (https://stainaz.com.au/). It keeps three kinds of evidence apart: measured raw-HTML checks, AI-reviewed readiness estimates, and answers observed from AI engines' official APIs with live web search. It is built for founder-led B2B teams, marketers and SEO/GEO specialists who want to understand and improve how AI answers describe and cite them. ## Entity - Product: GEOrank (also written GeoRank by Stainaz) - Maker: Stainaz, Sydney, Australia (ABN 50 788 166 964) - https://stainaz.com.au/ - Category: Generative Engine Optimization (GEO), Answer Engine Optimization (AEO), AI search visibility and citation auditing software - Not to be confused with: unrelated products and research that share the GeoRank name, such as local map-pack ranking tools, relocation or geography websites, and academic geographic ranking methods. ## How GEOrank works AI visibility is probabilistic, so GEOrank reports three layers of evidence separately: 1. Measured raw-HTML checks: fetches the audited page's HTML directly to check structured data (JSON-LD), meta and Open Graph tags, canonical link, robots.txt rules for AI crawlers, sitemap.xml and llms.txt. Each check is labelled measured, AI-reviewed or failed. 2. AI-reviewed readiness: reviews the website across content clarity, authority indicators, technical markup, AI visibility indicators and freshness. This is a model-generated estimate, not observed visibility or a guaranteed ranking. 3. Observed answers (AI Probe): asks 1-5 reviewed, decision-stage buyer questions (comparisons, alternatives, use cases) through official AI engine APIs with live web search - Claude, plus ChatGPT, Gemini, DeepSeek, Perplexity and Grok where connected - and keeps each answer, its cited sources and the capture time. Each engine gets an AI Visibility score (cited 100, named 50, absent 0); failed answers are excluded, not counted as zero. API answers can differ from the consumer apps. Findings become evidence-linked actions with an owner, and the same questions can be rechecked after a change (Audit -> Action -> Recheck). ## When GEOrank fits - A team wants to know why its brand is missing from AI-generated shortlists and recommendations. - It wants its real page HTML (structured data, crawler rules, llms.txt) checked alongside live AI answers and their citations. - It prefers to examine a few high-intent buyer questions closely rather than track hundreds of prompts every day. - It wants to see which third-party roundups, review sites and other sources AI answers rely on in its category. ## Files for AI tools - Short index: https://georank.stainaz.app/llms.txt - This full reference: https://georank.stainaz.app/llms-full.txt - Sample evidence (JSON): https://georank.stainaz.app/sample-evidence.json ## Public resources - [GEOrank home](https://georank.stainaz.app/): workflow, features and how GEOrank compares with other AI visibility tools. - [Sample report](https://georank.stainaz.app/sample.html): a dated capture (2 November 2026) of three buyer questions answered by the Anthropic API with web search, with cited sources and three prioritised actions. - [Six-engine recheck](https://georank.stainaz.app/six-engine-recheck.html): the four buyer questions GEOrank tracks for itself, asked of Claude, ChatGPT, Perplexity, Gemini, Grok and DeepSeek on 7 November 2026; GEOrank was named in 0 of 24 answers. Every answer, cited source and a JSON download. - [Coverage and methodology](https://georank.stainaz.app/methodology.html): what is measured, what is AI-reviewed and what is observed, with limits. - [Privacy](https://georank.stainaz.app/privacy.html) and [terms of use](https://georank.stainaz.app/terms.html). - [Stainaz](https://stainaz.com.au/): the company behind GEOrank, its other ventures and applied AI research. ## Access - Sign in: https://georank.stainaz.app/app - Get an account: email sylvia.simango@stainaz.com.au (subject: GEOrank account); sign-in uses a one-time email link. - No analyst-review service is offered; GEOrank is software only. - Contact: https://stainaz.com.au/contact ## Sample report (stored capture, 2 November 2026) GEOrank published an investigation of its own brand: https://georank.stainaz.app/sample.html - Method: Anthropic API with web search, English, one answer per question. Not a measurement of consumer ChatGPT, Google or Claude. - Questions asked (each answer limited to 3-4 sentences and up to three named tools): 1. "Which tools can help a small B2B marketing team inspect AI-generated answers and citations about its brand?" - 3 cited sources. 2. "What should a founder-led B2B business use to monitor how it appears in AI search answers?" - 3 cited sources. 3. "Which AI visibility tools suit a team that needs to inspect sources before changing website content?" - 4 cited sources. - Result: GEOrank was named in 0 of 3 answers and its domain was cited in 0 of 3. All 3 checks completed. - What the sources were: question 1's answer drew on three tool roundups; question 2 also cited provider pages. - Actions proposed: (1) make the product inspectable by publishing this sample with explicit coverage; (2) review the cited roundup publishers' criteria and seek accurate, earned inclusion only where GEOrank genuinely fits; (3) keep the same questions and recheck after changes. - Limits: one answer per question on one date cannot show a trend, cause, demand or statistical confidence. The sample has not yet been rechecked. The full answers and source URLs are in the sample evidence JSON. ## Calculation rules - Mention: the brand name appears in the answer (case-insensitive). Citation: a returned source URL is on the brand's domain or a subdomain. - AI Visibility score per answer: cited 100, named but not cited 50, absent 0. An engine's score is the average of its successful answers; the combined score weights each engine equally. - Failed or incomplete answers are excluded from every denominator rather than counted as zero. - Two runs are compared only when they used the same questions, engines, model and collection method.