4
posts
on this topic
27
posts
in English
Nov 14
latest post,
2026
Scraping or API: two ways to measure
Scraping reads the answer from the public version of the service, under fixed conditions. It is the closest to what a person sees, though without their chat history and settings.
API is a separate connection to the model. Convenient for automation, but the answer does not have to match what ChatGPT or Gemini shows in the browser.
How do they differ? Not in freshness: the API can use web search too. The difference is what you study: the service a person uses or the model itself. That is why I never merge numbers collected in two ways into one series. AI Overviews and AI Mode are a separate case: Google does not offer them through an API, so every tool reads them from the results page.
AI visibility is measured differently from classic search results. Instead of one position you get several traces at once: the Generative AI report in Google Search Console, the AI Performance report in Bing Webmaster Tools, presence in AI Overviews and brand mentions. Each source shows a different part of the picture and each has its own traps.
The posts collected here check that data in practice: what exactly the reports count, where impressions without clicks come from and what grounding queries in Bing reveal.
Want to know how AI sees your brand? AI search visibilityPersonal brand SEO
Start with
