LLMOcheki launches two analyses that work backwards from how AI perceives your brand

Itera, Inc. (head office: Shinjuku-ku, Tokyo; Representative Director: Takayuki Muto) has begun offering two features in LLMOcheki, its tool for measuring and improving LLMO, that work backwards from how generative AI perceives a brand: the Conditional Exposure Probe and Citation Structure Breakdown.

Appearance rates are compared condition by condition — “if you want to keep the budget down”, “if it is for a large enterprise” — to identify the conditions under which you drop out of the recommendation. Why competitors get cited is broken into four families, showing where to invest.

Background: in AI search, you are chosen on a conditional question

As information seeking moves through generative AI, the shape of users’ questions is changing. Instead of entering keywords, people ask conditional questions — “what can I use on a small budget?”, “which has a track record with large enterprises?” — and AI narrows to the candidates that fit the condition.

Which means there are brands that drop out of the running the moment a condition is attached. Ask about them by name and they are described correctly; ask a conditional question and their name never comes up once. Conventional search rankings cannot show this.

The other change is that the basis on which AI assembles an answer has spread beyond your own site. Third-party media articles, news coverage and user posts are cited, and if you do not appear in them you are not reflected in the answer. These two features address those two structural shifts.

Feature 1: the Conditional Exposure Probe

Starting from a base question — “recommend something in [category]” — a set of questions is put to AI with one of eleven conditions attached each time: price, company size, a preference for domestic providers, ease of adoption, track record and so on. Your appearance rate is then compared condition by condition.

Seeing which conditions keep you in the recommendation and which make you disappear lets you work backwards to the position AI assigns you. Where there is a gap between the positioning a company intends and AI’s perception of it, that gap is itself the starting point for the work.

Judgements are based on statistical confidence intervals, and conditions with too few measurements are marked explicitly as “pending”. By design, the state of a brand is not declared from a single result.

Feature 2: Citation Structure Breakdown

The reasons a competitor brand gets cited in AI answers are broken down automatically into four families: appearing in third-party media, primary information on the official site, news coverage, and user-generated content.

A competitor’s visibility often leans on one particular family, and you can identify the family where the gap with you is largest. That informs where to invest among negotiating placement in comparison media, putting your official information in order, strengthening public relations, and gathering reviews.

Because it re-aggregates measurement data you already have, it is available at no additional cost.

About LLMOcheki

LLMOcheki measures brand mentions and citations automatically every day across eight major engines including ChatGPT, Gemini, Perplexity, Claude and Google AI Overview, supporting everything from content production to verifying the effect. It provides LLMO work built around measurement: appearance probability based on statistical confidence intervals, structural analysis of citation sources, and measurement of traffic arriving via AI.

White-label delivery for agencies and partners, and measurement in overseas markets, are both supported. Multilingual display in the admin screen is being extended, starting from the main screens.

Comment from our Representative Director

“Ask AI about us by name and we are described correctly. Attach a single condition, though, and we can vanish from the candidates. You cannot know how you are seen without changing the conditions and comparing the answers. This feature does that comparison mechanically.”

LLMOcheki https://llmocheki.com/