Resources

We don’t hide how we measure.

A visibility number is only useful when the method behind it is known. What we measure in each engine, from which surface and how often, is written here.

Measurement method

01
In every engine we measure the surface the user actually sees
There is no single method; the surface is chosen per engine and the choice is verified by measurement. For ChatGPT and Gemini we measure the consumer interface, because API and interface do not return the same answer: in one the API cannot see the interface, in the other it inflates visibility. Perplexity is read through its own API, Google AI Overviews from the search result.
02
A scan every two days, no repeats within a round
Questions run once per scan round. Asking the same question many times a day does not improve the measurement; published work shows there is no information after the fifth repeat. Model non-determinism is handled in the analysis layer, not by repetition.
03
Non-determinism is measured, not hidden
The same question can produce different answers on the same day. That is why we keep a window metric alongside instant coverage: in what share of the questions measured in the selected period you appeared at least once. The gap between the two numbers is the non-determinism itself.
04
The score is read over a whole period
A single day says nothing. The score is the share of all answers in the selected period that mention your brand, so one wobbly answer does not move it while a real change shows in the curve as measurements accumulate.

Glossary

GEO: Generative Engine Optimization. Not your rank in the results, but whether you appear inside the answer the AI writes.
Fan-out: The sub-queries an engine generates for itself while building an answer. The real search chain behind the answer.
Citation source: The sites and pages feeding the answer. Which source describes you is often more decisive than what you wrote.
Window metric: In what share of the questions measured in the selected period you appeared at least once. Its gap to instant coverage shows non-determinism.
llms.txt: A file describing what your site offers AI engines. The panel generates it from your own content.
Schema: Structured data explaining to machines what your page is about. The technical side of getting into the answer.

Help

Write to us with questions about setup, measurement or billing. We have no phone line; everything runs in writing so what was said stays on record.

Write to us · support@kleur.ai

Start Your First Analysis

Guides and case studies are in the works. You can start your first measurement now and see results without waiting.

Start free

Measurement method

In every engine we measure the surface the user actually sees

There is no single method; the surface is chosen per engine and the choice is verified by measurement. For ChatGPT and Gemini we measure the consumer interface, because API and interface do not return the same answer: in one the API cannot see the interface, in the other it inflates visibility. Perplexity is read through its own API, Google AI Overviews from the search result.

A scan every two days, no repeats within a round

Questions run once per scan round. Asking the same question many times a day does not improve the measurement; published work shows there is no information after the fifth repeat. Model non-determinism is handled in the analysis layer, not by repetition.

Non-determinism is measured, not hidden

The same question can produce different answers on the same day. That is why we keep a window metric alongside instant coverage: in what share of the questions measured in the selected period you appeared at least once. The gap between the two numbers is the non-determinism itself.

The score is read over a whole period

A single day says nothing. The score is the share of all answers in the selected period that mention your brand, so one wobbly answer does not move it while a real change shows in the curve as measurements accumulate.

Glossary

GEO

Generative Engine Optimization. Not your rank in the results, but whether you appear inside the answer the AI writes.

Fan-out

The sub-queries an engine generates for itself while building an answer. The real search chain behind the answer.

Citation source

The sites and pages feeding the answer. Which source describes you is often more decisive than what you wrote.

Window metric

In what share of the questions measured in the selected period you appeared at least once. Its gap to instant coverage shows non-determinism.

llms.txt

A file describing what your site offers AI engines. The panel generates it from your own content.

Schema

Structured data explaining to machines what your page is about. The technical side of getting into the answer.