> ## Documentation Index > Fetch the complete documentation index at: https://docs.pokpok.ai/llms.txt > Use this file to discover all available pages before exploring further. # AI Visibility Truth: GEO at PokPok > Generative Engine Optimization at PokPok: how AI answers form an opinion of a brand (read, cite, name), the four P's of AI visibility, and which PokPok product measures which part. When a buyer asks ChatGPT or Gemini a question, the engine **reads** pages, **cites** a few of them and **names** some brands. GEO is the work of being read, cited and named. PokPok measures all three, and shows where a brand's AI opinion comes from. 01 // How an AI opinion is formed ## Read, cite, name FOUND ### Pages the engine read While answering, the engine searches the web and opens pages. Most of them never appear in the answer. ChatGPT cites about one page in four of the pages it reads (PokPok, 10 days, 4 brands: 25–35%). CITED ### Pages the engine shows as its sources A link tied to a sentence of the answer, or a source card. This is the evidence the engine chose to stand behind. MENTIONS ### Brands the answer names What the buyer actually reads. A brand can be named from a cited page, from a page the engine read and did not cite, or from nothing at all: the engine's memory. FOUND BUT NOT CITED ### Read, then left out Your page was among the pages the engine read for this question, and the answer neither cites it nor names you from it. The engine saw you and chose someone else. This is the list an agency works from. The vocabulary follows the industry's: Ahrefs' "found but not cited", Trackerly's "in-line citation vs retrieved source", and the IAB's Mention Rate and Citation Rate (Measuring Visibility in the AI Era, August 2026). 02 // The four P's ## The IAB's hierarchy, and where PokPok measures each | IAB level | Question | Metric | PokPok | | - | - | - | - | | **Presence** | Does the brand appear? | Mention rate, citation rate, share of voice, momentum | AI Visibility Truth: visibility, share of voice, change versus the period before. Report: named in N of 16 answers | | **Prominence** | Where, and how prominently? | Position | Not yet available | | **Portrayal** | In what context, with what accuracy? | Framing, factual accuracy | Report: every answer word for word, the brands it names, how it describes yours | | **Persuasion** | Does visibility drive action? | Recommendation strength, click-through | Report: recommended in N of 16 answers | Definitions and formulas: [GEO metrics](https://docs.pokpok.ai/geo/metrics.md). The three lists in detail: [Mentions, Cited, Found but not cited](https://docs.pokpok.ai/geo/mentions-cited-found.md). 03 // Two instruments ## A snapshot, and a line over time REPORT Brand Perception Truth Eight buyer questions written from your own website, asked once to ChatGPT and Gemini in your market: sixteen answers, every brand named, every page cited, and your page's read by people and by AI on 17 markers. €150, delivered next day. [The report →](https://docs.pokpok.ai/diagnose/brand-perception.md) MONITORING AI Visibility Truth From 50 questions per brand, with no upper limit (custom sets on request), asked every day on ChatGPT, Gemini, Google AI Mode, Google AI Overview, Grok and Claude. Visibility, share of voice, cited pages and sources over time; competitors side by side. From €50 / month per brand. [How it measures →](https://docs.pokpok.ai/analyse/ai-visibility-tracker-methodology.md) > [!NOTE] > > Engines do not give the same answer twice. The report is one day's answers, shown as given. AI Visibility Truth asks every day, so a change can be read against the days before it. 04 // What only PokPok adds ## From "not cited" to "why not" Every PokPok report already reads your page the way people and AI read it, on 17 markers and five AI metrics. The same reading, applied to a page the engine read but did not cite and to the page it cited instead, shows what the cited page has that yours lacks. That is the fix list. [What to show a client →](https://docs.pokpok.ai/geo/for-agencies.md) This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.