What GeoAnalyzer's AI-Visibility Audit Can and Cannot Establish

The published offer, the research paper that introduced GEO, and Google's documentation for its own AI features.

*Image: 2022-11-09 at 16-01-02 Klexikon - 2021 - Web Analytics Reports - Matomo by Ziko, licensed under CC BY-SA 4.0.*

What GeoAnalyzer sells

As of September 11, 2026, the GeoAnalyzer homepage offers a free instant scan with results delivered by email, a full report priced at $19 and discounted to $9.50 with the coupon code GET50OFF, a 4.0 Trustpilot rating labeled Great, and a link to the project's indiehunt.io page. The form takes a website URL and an email address, and the product is marketed as an evidence-based audit of a site's AEO score.

The page lists the scan's coverage as structured data, EEAT (spelled that way on the page), schema, and AI citations, and notes that it is used by SEO consultants, agencies, and SaaS teams. It also states that every score is tied to real content on the site, so users can see what drives it. The scan is described as using custom-trained models that simulate how AI systems read, rank, and cite content.

Where GEO comes from

The term GEO was introduced in a research paper, GEO: Generative Engine Optimization, submitted on November 16, 2023, last revised on June 28, 2024 (version 3), and accepted to KDD 2024. The paper describes generative engines as systems that use generative models to gather and summarize information from multiple sources, and it identifies website and content creators as a third stakeholder. Because those systems are black-box and fast-moving, creators have little to no control over when and how their content appears in responses.

The paper introduces Generative Engine Optimization as a framework for improving content visibility in generative engine responses, and GEO-bench as a benchmark of diverse queries across multiple domains with web sources for answering them. It reports that these methods can boost visibility by up to 40 percent in generative engine responses, and that their efficacy varies across domains, which the paper says points to the need for domain-specific methods. That figure is scoped to the paper's benchmark conditions; it reports what the study measured and is not a guarantee for any particular site or tool.

What Google says its AI features require

Google's documentation on AI features and your website states that there are no additional requirements to appear in AI Overviews or AI Mode, and that no special optimizations are necessary, while the SEO best practices already used for Search remain relevant. To be eligible to show as a supporting link, a page must be indexed and eligible to display with a snippet, the preview text shown under a search result; no further technical requirements apply. Google adds that meeting the requirements and best practices does not guarantee crawling, indexing, or serving.

The documentation also covers files and markup: no new machine-readable files, AI text files, or extra markup are needed to appear in these features. Responses may use query fan-out, in which the system issues multiple related searches across subtopics and data sources, and Google notes that AI Overviews and AI Mode may use different models and techniques, so the responses and links they show can vary.

For control, robots.txt directives for Googlebot govern how sites are crawled for Search, and nosnippet, data-nosnippet, max-snippet, or noindex can limit what Search shows from a page, while Google-Extended covers training and grounding in some of Google's other systems. For measurement, AI-feature traffic is included in Search Console's overall Web search data, within the Performance report. Google also states that clicks from results pages with AI Overviews have been higher quality, which it explains as users being more likely to spend more time on the site; that claim is Google's own, and the inspected material contains no independent measurement of it. The document covers Google's AI features; it does not describe third-party assistants.

What a scan can check, and what it cannot establish

GeoAnalyzer's FAQ defines GEO/AEO as how a business gets recommended by AI tools such as ChatGPT, Perplexity, and Claude, and states the stakes in absolute terms: AI returns one answer, maybe two, and a site absent from that answer does not exist.

A scan of a site can plausibly check signals that live on the page itself: whether crawler access is allowed in robots.txt, whether structured data is present and consistent, and whether key answers appear as short, self-contained statements. That list is an inference about what page-level inspection can see; no scan report from GeoAnalyzer was part of the material inspected.

Structured data is one place the two descriptions meet. Google documents that no special schema.org structured data is needed for its features, and the scan lists schema among the signals it examines. The statements cover different targets: Google's documentation describes Google's own AI surfaces, while the FAQ defines the category through assistants such as ChatGPT, Perplexity, and Claude. No official documentation for those assistants was part of the material inspected. No source in it measures whether the scan's criteria correlate with actual citations, so the connection between a page-level score and an AI recommendation stays open.

Where the evidence runs out

The paper was inspected through its arXiv abstract page. The benchmark's design and result tables were not examined, so the visibility figures remain at the level the abstract reports.

No independent evaluation of the scan's accuracy was part of the material inspected. The only review of the product found in that material is published on GeoAnalyzer's own blog (April 20, 2026), and a self-published review cannot serve as independent confirmation. The Trustpilot profile behind the homepage rating was not inspected, and the badge shows no review count.

The homepage leaves the FAQ answers after the first one collapsed, so positions stated in those answers were not read. Only part of the blog index was available: it shows posts dated January through April 2026, and later entries may exist. The index places labels such as Framework: 10/10 on posts, and the pages inspected name no external rater for them.

The offer's discount is presented as promotional and may be time-limited, and the checkout and report-delivery flow were not inspected. A GeoAnalyzer article comparing GEO and SEO cites outside figures, including a Gartner traffic projection and click-lift percentages; those were not traced to their original sources and are excluded from the analysis.

How to judge any AI-visibility audit

The same tests apply to GeoAnalyzer and to a competitor selling a similar score.

First, require each score to point at page content. GeoAnalyzer's own stated standard supplies the test: each score should trace to real content on the site. A finding that names a page element is checkable, and a number that traces to nothing is not.

Second, separate what is documented from what is claimed. For Google surfaces, the baseline is documented and public: indexing and snippet eligibility, crawl access, textual content, and structured data that matches visible text. For the assistants the category names, treat claims about what those systems reward as hypotheses until tested.

Third, build a baseline before paying. GeoAnalyzer's guide to measuring GEO success recommends tracking citations by manually querying ChatGPT, Perplexity, Claude, and Google AI and counting how often a brand appears per 100 queries. It lists the metrics it treats as worth tracking:

• Citation frequency: how often the brand is cited across a set of queries.
• Share of voice: the brand's percentage of citations in that set, compared with competitors.
• Attribution quality: the percentage of citations that include the brand name, with linked citations counted separately.
• Cross-platform coverage: presence across ChatGPT, Perplexity, Google AI Overviews, Claude, and Copilot.
• Brand search lift: branded search volume in Google Search Console, read for movement after citation activity.
• Sentiment and context: whether answers position the brand positively, neutrally, or negatively, and whether they cast it as a leader, an option, or a cautionary example.

The guidance is the company's own and carries no independent verification in the inspected material; the counting is manual, so the result is a baseline a buyer controls. Google reports AI-feature traffic inside Search Console's existing Web search data, which gives a first-party read on whether anything moves.

The free scan's results arrive before any payment, and the manual queries cost nothing. What a paid report adds beyond those two is the decision that remains.

Disclosure: GeoAnalyzer publishes this blog and is the subject of this article.