LLMOcheki launches CDR, its own metric scoring how much a domain is cited in AI search

Itera, Inc. (head office: Shinjuku-ku, Tokyo; Representative Director: Takayuki Muto) has begun offering CDR (Citation Domain Rating) in LLMOcheki, its tool for measuring and improving LLMO. The metric scores how much a domain is cited by AI search engines.

Citations of sources across eight AI search engines are observed continuously and indexed from 0 to 100 at the domain level. Domains with insufficient observation are shown no score, and every score carries a confidence grade.

Background: AI search lacks a common yardstick

In the era of search engines, a metric for the strength of a domain worked as a shared language for the industry, used to decide on activity and to choose where to be listed. More than the accuracy of the number itself, what mattered was that everyone could speak to the same yardstick.

Now that generative AI has become the point of contact with information, there is no well-established metric for comparing, across the industry, how much a site is cited by AI. Each company produces a citation rate within its own measurement scope — but that is within the questions it set itself, and cannot be lined up against anyone else’s.

CDR is designed as a common, domain-level metric that does not depend on any one company’s measurement design.

How CDR is designed

Based on measured data from eight AI search engines

Citations of sources in the answers of ChatGPT, Gemini, Claude, Perplexity, Copilot, Grok, and Google’s AI Overview and AI Mode are observed continuously and scored at the domain level.

Composed of three sub-scores

The breadth of citation, the diversity of engines, and stability are each calculated and can be inspected as the breakdown of the overall score. A temporary rise in citation on one engine can be distinguished from sustained citation across several.

Numbers that cannot be produced are not produced

Domains with insufficient observation are shown “insufficient data” rather than a score. Every score carries a confidence grade (A, B or C) and the volume of observed data. A figure calculated from few observations is never presented in the same form as one calculated from many.

Your score does not move because other domains did

The scale is fixed, so your own score is unaffected when the population being compared changes. A change in the number can be read as a change in your own state.

Main features

Comparison against competitor domains, a 90-day trend, and lookup of any domain are supported. Checking the CDR of a publication you are considering being listed in tells you which outlets are more likely to lead to citations.

It is also available via the REST API, for building into your own dashboard or BI tools.

On our internal validation

In our internal validation, across roughly 690,000 citation observations over eight weeks, a high rank correlation was found between a domain’s CDR and the citations it actually won the following month.

This is a correlation within our own observation data, and an improved score does not guarantee that citations will follow.

Availability

Available on every LLMOcheki plan at no additional cost.

What comes next

We will extend it as a basis for evaluating domains in the age of AI search, with CDR rankings by industry and a visualisation of the citation network among other things.

A note on this metric

CDR is a relative assessment within our own database and does not guarantee the absolute quality or authority of a domain. It is our own metric, unrelated to similarly named metrics provided by third parties.

Comment from our Representative Director

“One reason the discussion about AI search work does not progress is, I think, the absence of a shared number. Each company produces a citation rate within the questions it set itself, and you cannot compare that with anyone. There is value in building a number everyone can argue from.”

LLMOcheki https://llmocheki.com/