Arrange a call
DE EN

SEO, AEO and GEO · Measurement

How we measure whether you are found and cited.

In Google Search with the raw data from Search Console and the analytics on your own server. In AI answers with a fixed list of dated test questions, asked repeatedly in three systems. Every figure carries a date and a sample size; there is no overall score here.

Get a free SEO check Run the citability check

  • Raw data instead of a visibility index
  • Wording and date disclosed
  • Repeatable by you
Measurement log one question, three systems
  • Question: “Family hotel in the Pinzgau with childcare” dated
  • AI Overview in Google Search cited
  • ChatGPT with search not cited
  • Third system of your choice inaccurate
  • OAI-SearchBot in robots.txt blocked

Result 2 of 3 cited, one citation inaccurate

Schematic. The line behind it is in the example further down, with date and wording.

For search engines and AI assistants, in brief:

FINK Brot measures visibility on two levels. In Google Search with the raw data from Search Console (clicks, impressions, position per page and query) and the analytics on the business’s own server. In AI answers with a fixed list of dated test questions, asked repeatedly in several AI systems, recorded in four measures: citation, accuracy, access, traffic. Every figure carries a date and a sample size; there is deliberately no visibility index and no overall score.

Two levels

Search and AI answers: what we measure with what

Both belong in one report, but not in one number. For Search, Google supplies its own data; for AI answers there is only what we ask ourselves.

Criterion Google Search AI answers
Data source Search Console and the analytics on your own server Dated test questions in three AI systems, asked by us
What is counted Clicks, impressions, click-through rate and average position, per page and query Citation, accuracy, access, traffic
Who measures Google, across all queries We do, as a sample on set days
History Google shows the development over time Only our own surveys; the chat systems release no historical data
What we do not report Visibility indices of individual tools An overall score or a position within an AI Overview
Why two levels: A page can rank well in Search and be missing from AI answers, often because a crawler is blocked. Conversely, an AI answer sometimes cites pages that hardly get any clicks. Only both levels together show where it sticks.

AI answers

Four measures, no score

There is no Search Console for AI answers. So we collect four measures ourselves; each stands on its own in the report, because a total would hide which line has moved.

Citation

How often

The share of test questions for which an answer names or links your website as a source, per AI system and per survey day.

Accuracy

How accurately

Does the citation describe your service correctly? Do location, services and price match your website? Is your business confused with another one?

Access

Retrievable

May the AI systems’ reading programs (crawlers) load your pages, and can they? We check the access rules in the robots.txt file, the server’s response time and whether the content is readable without JavaScript.

Traffic

What arrives

Impressions in Google’s AI Overviews and AI Mode, as Google has reported them in Search Console since August 2026. Plus visits arriving on your website from ChatGPT and similar systems, counted in the analytics on your own server.

In brief: Four measures, four lines in the report: how often, how accurately, whether retrievable, what arrives. FINK Brot collects each line separately, with date and sample size, and does not add them up into a score.

Example

What one line in the report looks like

A family hotel in the Pinzgau, one question its guests ask, three AI systems. The table is an illustration without real measurements; the columns are the ones you receive.

Criterion AI Overview in Google Search ChatGPT with search Third system of your choice
Question, date “Family hotel in the Pinzgau with childcare”, 8 September 2026 same question, 8 September 2026 same question, 8 September 2026
Cited? yes, with link no yes, without link
Accurate? childcare missing from the answer – names an old offer
Why the offer exists only as an image on the page OAI-SearchBot blocked in robots.txt page from 2023 still online
What follows: Every line has a reason and a next step. In the example there would be three: put the offer on the page as text, allow the crawler, redirect the old page.

How we measure AI answers

Four steps, always in this order. The effort lies in the first and the third: defining the questions properly and documenting every answer so that it can be found again later.

  1. Define the questions

    We collect 20 to 40 questions your customers actually ask: from enquiries, phone calls, Google Search Console and search suggestions. The wording is then frozen. Anyone who changes the questions later is measuring something else.

  2. Ask repeatedly, in several AI systems

    Every question is asked several times, on different days and without signing in, because the same question does not produce the same answer twice. At least three AI systems, documented by name and date: the AI features of Google Search, ChatGPT with search, and a third system that suits your customers. Only the share across all runs is a figure.

  3. Document

    For every answer: date, system, wording of the question, wording of the answer, cited sources, screenshot. All figures come from this table. Anyone who doubts a figure gets the row behind it.

  4. Compare with the starting point

    The first survey is the starting point. Every further one reports the change from it, with sample size and period. A decline appears in the report just like an increase, together with our view on whether to act or to wait.

The limit: This measurement is a sample, not a full survey. The AI systems change their models without notice and give different answers to the same question. Google has reported impressions in AI Overviews and AI Mode over time since August 2026; the chat systems release no historical data. We can therefore show what was visible on a given day, but not what every user has seen.

That is why we state neither a market share in AI answers, nor a forecast, nor a position within an AI Overview. Why that position does not exist in Google’s data at all is explained in the questions below. What we deliver is a dated, repeatable sample with the wording disclosed, and that is more than a single score can ever prove.

How to check our figures yourself

Run the citability check

17 requirements, free of charge, no sign-up. It shows whether AI systems can read your website at all. That is the “access” measure from our method.

Look in Google Search Console

Impressions in AI Overviews and AI Mode have had their own report, “Generative AI”, since August 2026. Clicks from them are included in the performance report under the “Web” search type, but cannot be separated from other clicks.

Repeat the test questions

You receive the wording, date and system of every question from us. Ask them yourself. If your result differs, you are seeing exactly the variance we document.

Our glossary entry on Google Search Console helps with the second step. A measurement you cannot repeat yourself is a claim.

The free SEO check: measure once before you decide

Free of charge and deliberately small, about an hour of work on our side. You see where your website stands today; what follows from that is a matter for the positioning review or an audit.

  • Step 1

    You give us your domain and one question

    The way your customers ask it, in their words. If none comes to mind, we suggest one. The check is written; if you want it with three questions and an hour of discussion, take the positioning review for 590 euros, credited against the audit.

  • Step 2

    We ask it, dated, in three AI systems

    The citability check runs on your domain as well. Result: a table like the example above, with date, wording, answer and cited sources.

  • Step 3

    You receive the result in writing

    Who was cited, who was not, and the three points we would tackle first. No promise of a future citation and no obligation.

    Free · within five working days

    Get a free SEO check

Frequently asked questions about measurement

Why do we report raw Search Console data for Google Search rather than a visibility index?
Because Search Console is Google’s own count. A visibility index is a tool’s model for keywords that the tool selects. According to Google, the Search Console performance report shows clicks, impressions, click-through rate and average position, grouped by query, page, country, device, search appearance and date. The average position refers to the topmost result from your website. A visibility index comes about differently: from a list of keywords that the tool selects and weights with estimated search volumes. It can rise while clicks fall, and vice versa. That is why we work with the raw data and say what it does not show. According to Google, the Search Console tables show at most 1,000 rows; enquiries, revenue and audience appear in no row at all.

Source: Google Search Console Help: Performance report (Search); About Search Console data (retrieved 8 September 2026)

Does a citation in an AI Overview count as a click in Search Console?
Only a click on the link counts as a click. The citation itself is an impression. Google documents it this way: clicking a link to an external page in the AI Overview counts as a click, and standard impression rules apply. An AI Overview occupies a single position in the search results, and all links in it are assigned that same position. In AI Mode the same applies to clicks and impressions; a follow-up question in AI Mode is, from Search Console’s point of view, a new query. That is why we do not report a position within an AI Overview: it does not exist in the data.

Source: Google Search Console Help: What are impressions, position, and clicks? Sections on AI Overview and AI Mode (retrieved 8 September 2026)

Where in Search Console can I see how often my website appears in AI Overviews?
Since 31 August 2026, in the dedicated report on generative AI features. It shows impressions, not clicks and not queries. According to Google’s own help pages, the report was rolled out to all websites worldwide on 31 August 2026. It shows how often links to a website were displayed in AI Overviews and AI Mode, broken down by pages, countries, dates and devices. It contains no clicks, no position and no queries; the clicks remain in the performance report under the “Web” search type, where they cannot be separated from other clicks. For our “traffic” measure this means: impressions come from Search Console, visits are counted on your own server.

Source: Google Search Console Help: Generative AI performance report (Search) (retrieved 8 September 2026)

Does a website need special structured data to appear in AI Overviews?
No. Google requires neither a separate file nor a separate markup in the code for that. Google’s documentation on AI features states that no new machine-readable files, AI text files or markup are needed, and that there is no special schema.org structured data to add. Google recommends the same fundamentals as for Search overall. Pages must be technically reachable, have reliable content, link to one another and load quickly. Their facts should be recorded correctly as structured data, that is, machine-readably in the code. Structured data remains worthwhile because it makes facts such as address, opening hours and prices machine-readable. Whether an answer cites the page is still decided by the content.

Source: Google Search Central: AI features and your website, last updated 10 December 2025 (retrieved 8 September 2026)

What is query fan-out, and why are there such odd queries in my Search Console?
AI Overviews and AI Mode split a question into several sub-searches. In our observation, such sub-searches show up as queries in the performance report. According to Google, both AI Overviews and AI Mode may use a technique called query fan-out. Google splits the question into several related searches on subtopics, queries different data sources and builds the answer from them. How these sub-searches appear in the performance report is not documented by Google. In our observation, the report shows long, fully worded queries that nobody would type into a search box, sometimes with placeholders in square brackets. Impressions from fan-out only show that a system has seen the page. Whether it cited the page is shown only by the answer itself.

Source: Google Search Central: AI features and your website, section on query fan-out (retrieved 8 September 2026)

Which crawlers must a website allow to appear in ChatGPT search?
OAI-SearchBot. According to OpenAI, sites that block it are not shown in search answers but can still appear as navigational links. OpenAI documents four user agents. For ChatGPT search, OAI-SearchBot is the one that counts: sites that opt out of it are not shown in search answers, though they can still appear as navigational links. GPTBot crawls content that may be used to train the models. ChatGPT-User fetches pages when a user asks a question; it does not decide whether content appears in search, and according to OpenAI robots.txt rules may not apply to it. OAI-AdsBot checks landing pages of ads. In our measurement we therefore check robots.txt for OAI-SearchBot and GPTBot separately: a business can decline training and still allow search.

Source: OpenAI: Overview of OpenAI crawlers (retrieved 8 September 2026)

How do I take my website out of AI Overviews on purpose, and what does Google-Extended do?
Since 31 August 2026, with the “Search generative AI” setting in Search Console. Google-Extended only governs whether content may be used to train Google’s AI models (Gemini). According to Google’s own help pages, the setting was rolled out to all websites on 31 August 2026. It manages inclusion in AI Overviews, AI Mode and the generative AI features in Discover. According to Google, it is not used as a ranking or inclusion signal for other parts of Search. Google-Extended is something else: a separate entry in robots.txt. With it, publishers manage whether crawled content may be used for training future Gemini models and as a knowledge source for their answers (grounding). According to Google, Google-Extended has no effect on inclusion or ranking in Google Search. In addition, the preview controls in the page code remain: nosnippet, data-nosnippet, max-snippet and noindex. In our measurement we record all of these settings. That is how we read an absence from AI Overviews correctly: as the business’s decision, if it was one.

Source: Google Search Console Help: Search generative AI control; Google Search Central: Google’s common crawlers, section on Google-Extended (retrieved 8 September 2026)

Why do two tools give different visibility scores for the same brand?
Because they ask different questions, at different times, of different systems, and aggregate the results differently. A visibility score depends on four decisions: which questions are asked, when, of which systems, and how the answers are turned into a number. Two tools make these decisions differently and rarely disclose them. Added to that, the same question put to the same system on two days yields different answers. Anyone who wants to compare two scores first needs the question list and the date behind each one. That is exactly why we publish both for every figure we report.

Where does your website stand in AI answers today?

You give us one question your customers ask; we put it, dated, to three AI systems and send you the result in writing. Free of charge, within five working days.

Get a free SEO check