--- title: "AI Visibility Measurement Methodology | PicoBit.AI" description: "How PicoBit.AI measures AI visibility: 7 AI systems, repeated runs in clean sessions, a baseline, and separate counts for mentions, citations and recommendations." url: https://picobit.ai/en/methodology language: en updated: 2026-09-25 --- # How we measure AI visibility *Author: [Boze Bago](https://picobit.ai/en/about)* A single AI answer proves nothing: the same question can get a different answer tomorrow. That is why we measure repeatedly, document every run and follow fixed rules. ## Which systems we measure We measure seven AI systems: ChatGPT, Microsoft Copilot, Gemini, Google AI Overviews, Google AI Mode, Perplexity and Claude. The Basic Check covers five of them: ChatGPT, Gemini, Google AI Overviews, Google AI Mode and Copilot. Copilot is included because 43% of AI users in Germany use it ([Bitkom, April 2026](https://www.bitkom.org/Presse/Presseinformation/Ein-Drittel-nutzt-KI-mindestens-einmal-pro-Woche)). ## How a measurement run works - We ask every question three times in every system. - Each run starts in a clean session with no history. - For every run we record: system, date and time, exact question, country and location method, interface language, logged in or not. - For DACH measurements we use a Germany-based environment wherever the system allows reliable location control. If a measurement is taken outside Germany, the report says so. ### Baseline Before any optimisation we measure the starting point. This baseline is dated and frozen. Every later result is compared against it as before → after. ### Spread AI answers vary from run to run. Three runs per question show whether a result is stable or appeared only once. ## What we count separately - **Mention:** your company's name appears in the answer. If your company is only described or linked without its name, we count it as a soft mention. - **Citation:** the answer lists your website as a source. - **Recommendation:** the AI explicitly suggests your company as a provider. A mention is not a recommendation. That is why we count all three separately. ## How we label results Every statement in our reports carries one of three labels: - **Verified:** measured directly, or backed by a source and date. - **Inferred:** our conclusion from the measurement data. - **Open:** not yet checked. ## Limits of measurement - No measurement guarantees that a company will be mentioned, cited or recommended in future. - AI systems change constantly. Results apply to the period measured. - Location, language and login status affect answers. That is why we record them for every run. ## Example: Basic Check 15 questions × 5 systems × 3 runs = 225 measurements. You receive a written report and a call to go through it. Price: €950 net, one-off. The amount is credited against a subsequent Full Analysis. → [More about the visibility audit](https://picobit.ai/en/visibility-audit) *Author: [Boze Bago](https://picobit.ai/en/about) · Last updated: 2026-09-25* ## Next step Want to know what AI systems say about your company today? [Get in touch](https://picobit.ai/en/contact) --- PicoBit.AI · Contact: [hallo@picobit.ai](mailto:hallo@picobit.ai) · https://picobit.ai/en/methodology