Measure for free. Pay for what you need.

What counts as a "question"?

One question = 1 prompt × 1 language. Each question is sent to the AI engines your project selects (up to four on paid plans; the one you picked on the free plan) — the number of engines does not multiply your question count. Six prompts measured in Japanese only cost 6 questions; the same six across 5 languages cost 30. Narrowing languages lets you run more measurements within the same limit (generating question candidates does not consume your quota).

What can I do on the free plan?

One project: pick one language (of five) and one AI (ChatGPT, Gemini, Claude, or Perplexity), and run up to 10 questions a month (about 3 runs), plus the top 3 improvement suggestions. No credit card required. Measuring several languages and several AIs at once, weekly automatic runs, trend history, competitor tracking, and the full set of suggestions are paid-plan features.

Can I subscribe to a paid plan right now?

Not yet. Paid plans are not open — the only thing you can do today is join the waitlist, and we are not taking payments. When we open, waitlist members are contacted first. The reliable way to judge us in the meantime is to run a real measurement on the free plan.

Are the listed prices final?

No. Prices are planned and will be finalized after validation with our first customers.

What happens when I run out of questions?

You cannot start new measurements until the next monthly reset; everything you already measured stays readable. If you burn through it quickly, narrowing the languages you measure is the most effective lever — your question count is prompts × languages.

Do the weekly automatic runs consume questions?

Yes — they draw from the same monthly limit; there is no separate quota for automatic runs. Note that on Solo the automatic run covers only your primary language: running all five every week would exhaust the monthly limit in about three weeks. Build the trend consistently in one language and do cross-language comparisons manually.

AI answers change every time. Is measuring even meaningful?

If you read a single run as "the answer", then not much. We do not treat one run as definitive either. What holds is the pattern: repeat the same question set and stable structure emerges — the competitor who always appears, the brand that never does, the one language where you are weak. That is why weekly automatic runs and trend history exist. Drawing big conclusions from a single run is exactly what we would advise against.

Our brand is written several ways (abbreviations, English, local script). Will you catch them?

You can register variations in the project settings, and a mention counts if any one of them appears (e.g. "Starbucks", "SBUX", the local-script form). This matters most when measuring several languages, because AI refers to you differently in each. On the extraction side we also keep names exactly as written in the answer — so we never transliterate you into what looks like a different brand.

If we already do SEO, will we show up in AI search?

They correlate, but they are not the same thing. Search hands you a list and you pick; AI answers with the picking already done — one or three recommendations. So ranking on page one while being absent from AI answers is entirely normal. What AI needs is a different shape: direct answers to questions, structured data, third-party mentions. Our free site scan checks exactly that.

Why measure Perplexity as well?

ChatGPT, Gemini and Claude are conversation first with search attached; Perplexity is search itself rebuilt as AI. If we claim to measure visibility in AI search, leaving that shape out would skew the sample. Answers genuinely differ per engine, which is why we never call a number from any single engine "your AI search visibility".

Are the numbers real?

Every demo on this site is an illustrative example and labelled as such; real data comes from measurements after you sign up. We also publish our own detection accuracy, position extraction error and report fabrication rate — including where we are still weak.

Which languages and AI engines are supported?

Japanese, English, Korean, Chinese, and Cantonese — five languages (paid plans measure all five together; the free plan measures one of your choice). Four AI engines — ChatGPT, Gemini, Claude, Perplexity (paid plans pick freely among the four; the free plan uses the one you pick).

Why does AI answer differently in Japanese and in English?

AI looks for information in the language you asked in, and leans on what is widely discussed in that language. The sources that carry weight in Japanese and in English are simply different, so the cast of brands changes even for the same question. Being first in one language and absent in your target market's language is not unusual — it is the default assumption. We send the same questions in five languages and put that gap side by side.

We are expanding into Korea, China or Taiwan. Can we see how we look in those languages?

Yes. We run the same questions in Korean, Chinese and Cantonese and show whether you appear when local customers ask AI, and who is recommended instead. Reports come out in Japanese, English or Korean, so you can hand them straight to the local team. To be straight with you though: we have not yet validated detection accuracy on real answers in Korean, Chinese or Cantonese. We are running on the assumption that it matches Japanese and English, and we say so on our methodology page too.

We only serve the domestic market. Do we need the multilingual part?

Correct — in that case, use a single language. Your question count is prompts × languages, so narrowing languages buys you more runs for the same price. Multilingual matters in two cases: people searching for you from abroad, and you moving abroad. If neither applies, measuring one language more deeply is the rational choice.

Are LLMO, AEO and GEO different things?

They point at nearly the same thing. LLMO is the common term in Japanese, while GEO (Generative Engine Optimization) and AEO (Answer Engine Optimization) are more common in English. Whatever it is called, what we do is make sure you show up when AI answers — and that is the single lens to read this pricing table through.