GeoAnalyzer Stopped Selling Its 0-100 Scan. The $25 Course That Replaced It Leaves the Citation Test With You

The five categories graded your own pages, and the site now states that a score cannot answer whether ChatGPT recommends your brand

*Image: Blocked robots, according to website's robots.txt as of June 2026, sample size 1000 by ChippyTimeCom, licensed under CC BY-SA 4.0.*

GeoAnalyzer, the GEO/AEO audit tool that scored sites from 0 to 100 on readiness to be cited by AI assistants, no longer sells that scan on geo-analyzer.com. The homepage checked on October 26, 2026 reads "GeoAnalyzer is now AEO Mastery" and prices a course at a $25 launch rate against a $59 list, paid once for lifetime access. Its FAQ answers "What happened to GeoAnalyzer?" by saying the scanner showed a score, the course teaches the process behind it, and links to existing paid reports keep working.

The change moves the buyer's decision. Someone who wanted an outside number now assembles one from two parts: whether the crawlers each engine runs can read the pages, and whether the engine names the brand for prompts buyers type. The first part is a checklist with documented rules from each operator. The second is a sample you design and repeat, and although the course teaches the process behind the score, the citation measurement stays with the reader.

The five categories the score used

The about page still describes GeoAnalyzer in the present tense and lists the five categories behind the score: Entity Clarity, Direct Answers, Trust Signals, Competitive Positioning and Technical Accessibility. Each question it attaches concerns the reader's own site.

The questions run: can AI understand who you are and why you are authoritative; is the content structured for extraction and citation; do you carry the credibility indicators engines use to verify sources; how do you compare with competitors currently being cited; can engines access the content without barriers. Those questions are answerable by inspection. A schema validator, a robots.txt read, an about page and a set of author bylines supply most of the inputs. The categories survived the product change as a syllabus: the homepage's Module 5 covers entities, structured data and trust signals, and Module 3 covers crawler access.

The about page has not been updated to match. On October 26 it still ends with "Get your free GEO score in under 60 seconds" and an "Analyze My Site" button, while the homepage no longer carries the URL-and-email scan form. A reader who lands there first gets a product description that the homepage contradicts, so read the homepage FAQ for the current offer and treat the rest of the site as older copy.

Where crawler access splits

Technical Accessibility was one of the five categories, and each engine operator documents its own agents with different jobs and different consequences. Google's list of common crawlers states that crawl preferences addressed to Googlebot affect Google Search, including Discover and all Google Search features, while Google-Extended is a separate token that "does not impact a site's inclusion in Google Search nor is it used as a ranking signal" and instead governs whether crawled content may train Gemini models or ground Gemini Apps and Vertex AI. Allowing one and disallowing the other is a coherent policy, and the two rules do different things.

OpenAI splits the same way, and its crawler page gives three agents different jobs. Its crawler documentation describes OAI-SearchBot as the agent that surfaces sites in ChatGPT's search results and states that sites opted out of it "will not be shown in ChatGPT search answers, though can still appear as navigational links". GPTBot crawls content that may be used to train foundation models, so disallowing it signals exclusion from training data. ChatGPT-User handles fetches a person triggers inside ChatGPT, and because those requests are user-initiated, the documentation says robots.txt rules may not apply to them.

Anthropic documents a matching trio in its article on its crawlers: ClaudeBot for training data, Claude-SearchBot for search indexing, with the warning that disabling it "may reduce your site's visibility and accuracy in user search results", and Claude-User for user-initiated fetches. Perplexity's crawler page separates PerplexityBot, which it says surfaces and links sites in search results and is not used to train foundation models, from Perplexity-User, which it says "generally ignores robots.txt rules" because a user requested the fetch.

AEO Mastery's own robots.txt lesson puts the same rule in its public summary: give each AI crawler a named user-agent group, allow the search agents if you want citations, disallow the training agents if you object to training use, then confirm with a live fetch and server logs that no CDN rule overrides the file. That last step covers a failure the file itself cannot show, because a WAF or bot-management rule in front of the site can refuse a request that robots.txt allows.

The chart at the top of this article, self-published on Wikimedia Commons, ranks the user agents that 1,000 domains block in robots.txt. The wildcard group leads by a wide margin, and the list carries Googlebot, OAI-SearchBot and PerplexityBot alongside GPTBot, ClaudeBot and Google-Extended. A blanketed rule copied from a list like that one blocks the agents that put pages into answers along with the agents that collect training data.

What Google's documentation says about the payoff

Access is a precondition, and Google's AI features documentation states what eligibility adds beyond it. It states that there are "no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary", and that a page is eligible to be shown as a supporting link when it is indexed and eligible to be shown in Google Search with a snippet. The same page states that meeting every requirement does not mean Google will crawl, index or serve the content.

The gap matters for any structural audit. A site can pass all five categories and still be absent from an answer, because the audit describes the page and the engine decides whether to retrieve and quote it.

What the citation side requires

Citation is a rate over a set of prompts, and the set decides the number. GEO: Generative Engine Optimization, accepted to KDD 2024, built a benchmark called GEO-bench from queries paired with relevant web sources and measured how rewriting source content changed its visibility in generated answers. The reported gain reached 40 percent, and the authors state that the effectiveness of the methods varies across domains.

Scale moves the expectation further. A June 2026 preprint by Pratyush Kumar analyzed more than 100,000 prompt responses across more than 100 brands tracked between March and May 2026. It reports first-run visibility of 73 percent for global household names, 44 percent for established mid-market and regional brands, and 11 percent for niche and small brands. About 78 percent of the cited sources in that sample were corporate websites, and the analysis reports that whether a brand is framed positively or negatively flips about 6.7 times more often than whether it is mentioned at all. Kumar is affiliated with Ranqo, and the sample came from brands tracked on that platform, so those figures are a vendor measurement of one sample. The paper's own next step is a set of protocols for testing whether recommendations can causally improve visibility.

The two results describe different work. The benchmark's gains come from how a passage is written. The visibility ladder comes from brand stature and the surrounding web, which no page edit reaches this quarter. The citation sample therefore belongs in its own record: 20 to 50 prompts a buyer would actually type, unchanged between runs, each recorded for mention, position, the competitors that appear and how the brand is described, with every prompt run several times because answers vary run to run.

What the $25 buys now

The homepage lists AEO Mastery at $25 against a $59 list price, one payment for lifetime access, with 32 lessons across eight modules, 8 interactive tools, quizzes, a 90-day plan, a certificate and future updates included. Checkout runs through Stripe, and sign-in uses a one-click link sent to the checkout email. The free path is Module 1 without a sign-up, plus a six-question self-check that the page says needs no email.

The eight tools turn the lessons into work on the reader's own site. The same homepage describes a Prompt Basket for turning a brand and its competitors into a test set, an Appearance Tracker for logging runs to get "the rate that survives randomness", an AI Crawler Checker, an Answer Grader for scoring a passage's citability, a Chunk Checker for sections an engine cannot lift alone, a JSON-LD builder and a source gap map. A reader operates those instruments and keeps the output, and no outside 0-100 number comes back.

The homepage also states the course's limits. Asked whether it will guarantee that ChatGPT recommends a brand, the FAQ answers "No, and be wary of anyone who promises that", and adds that AI answers vary from run to run. The site states that limit on the same page where it sells the replacement.

The decision

Three conditions cover most buyers. If the open question is whether the pages are readable by crawlers and answerable in a passage, start with the free module and run the access check yourself against each operator's documentation, because that check costs nothing and the rules come from the operators. If a client needs a repeatable number, the prompt basket and appearance tracker build one from your own runs, and you keep the raw output that backs it. If a paid GeoAnalyzer report was bought before the change, the FAQ states that its links keep working, so the per-page health audit inside it remains a usable baseline and its findings can be re-checked against the current site.

Two items stay open. The pages inspected for this article do not say whether the scanner will return, and they do not say how long report links will keep working, so both statements are current as of October 26, 2026 and no stronger.

Run the access check and start the prompt list in the same week, so the first citation rate and any access fix share a baseline.