Free AI Check

Methodology: how we measure AI visibility: Measure instead of guessing.

How GEOHACKER measures your business’s visibility in AI answers: with a measurement concept that fits your business, continuously over months and with an open account of measurement accuracy.

Measured inChatGPTGeminiClaude

From customer question to report in five steps

  1. 1

    Develop the measurement concept

    It all starts with your business. Together we decide which product lines or service areas are examined, which questions your customers ask an AI, which competitors are relevant and at what rhythm measurements take place. The question set is based on an analysis of your website and your market and then remains stable – this is the only way to compare months.

  2. 2

    Ask the AI systems

    Each question is put to ChatGPT, Gemini and Claude under identical conditions, each with web search as in real use. Every question exists in several phrasings with the same meaning, rotating from one measurement day to the next, so the result does not depend on a random choice of words.

  3. 3

    Measure continuously

    Measurements take place on fixed days over weeks and months. Depending on the measurement concept, this produces several hundred to over a thousand answers per month. This reveals patterns instead of chance hits – and makes changes over time visible.

  4. 4

    Evaluate the answers

    All answers are stored in their original form. The evaluation covers whether and how your business is mentioned, how it is presented and which sources the AI relies on – overall and per product line, AI system and question type.

  5. 5

    Calculate and report the score

    Each dimension is calculated individually and then weighted. All rates are reported with a 95% range – calculated across questions because answers to the same question are related – and explained in the monthly online report.

How accurate is the measurement?

AI answers are not deterministic: the same question produces a different answer every time it is asked. In a study by SparkToro and Gumshoe with almost 3,000 runs, the same list of recommendations reappeared in fewer than one in a hundred cases. What is reliable, therefore, is only how often a business is mentioned across many runs.

Analyses of AI visibility measurement recommend at least seven to eight repetitions per question and system within two to four weeks for a reliable mention rate. GEOHACKER sets the scope and measurement rhythm for each business so that this value is reached in every evaluation period – with measurements three times a week, that is around 13 repetitions per month.

The measurement design builds on the founder’s master’s thesis on Generative Engine Optimization and on current industry studies.

Sources: SparkToro/Gumshoe via Search Engine Land, ThriveStack Research. Both are industry studies, not peer-reviewed publications.

AnswersUseAccuracy
30Free AI Check± 18–30 percentage points
≈ 350Analysis per AI system and month*± 8–12 percentage points
≈ 1,000Analysis overall per month*± 5–7 percentage points

* Example with 30 questions, three AI systems and measurements three times a week; the actual scope depends on your measurement concept. 95% range in the least favorable case of a 50% mention rate; at lower rates it is narrower. The report shows the range for every single value.

Why not count every answer individually? The repetitions of a question are not independent: if an AI mentions a business for a question, it usually does so in the repetitions too. Counting every answer as an independent sample would make the result look more precise than it is. GEOHACKER therefore calculates the intervals using a cluster bootstrap: whole questions, together with their repetitions, are resampled many times. This is more honest. The ranges are wider, but they are reliable.

Six dimensions, one score

Mention rate

35%

Share of answers in which your business is mentioned at least once. Multiple mentions in the same answer count once.

Recommendation strength

20%

Is your business only mentioned in passing, listed as an option or explicitly recommended?

Position

15%

How early and how prominently your business is mentioned on average across all runs. Individual placements vary a lot and are therefore not assessed in isolation.

Tone

10%

How positively, neutrally or negatively your business is presented, and with which attributes.

Source presence

10%

Is your business considered in answers that cite sources? Answers without sources are not counted against you.

Source diversity

10%

How broadly your presence is spread across different sources. Many references to the same website do not count as diversity.

Visibility Score = weighted sum of the six dimensions, each from 0 to 100.

How to read the score

0
Not visible

The business does not appear in the answers examined.

1–24
Barely visible

Occasional mentions, mostly in passing.

25–59
Partly visible

Present for some questions and systems, absent for others.

60–100
Clearly visible

Mentioned regularly and often explicitly recommended.

What the score is not

  • The Visibility Score is GEOHACKER’s own methodology and not a scientifically validated industry benchmark. The weighting is reviewed as the data base grows.
  • AI answers change with models, sources and time. Results apply to the measurement period and are therefore repeated regularly.
  • The platforms are evaluated separately. A single overall value would hide the fact that one system knows a business and another does not.
  • Competitors are measured with the same questions under the same conditions. Only then is the comparison fair.

AI Check and analysis

The free AI Check is a simplified snapshot: ten questions per system, asked once, giving 30 answers with an accuracy of around ±20 percentage points or more. It shows a tendency. The analysis works with a tailored question set per product line, measures continuously over months, calculates all six dimensions and shows the development in a monthly online report.