Get found in the language
your next market speaks.

Because customers now ask AI before they ever open a comparison site — and AI answers with three candidates, not ten links. The shortlist is decided before anyone compares you. If your name is not in that answer, your price and quality are never evaluated at all, because you never make it onto the table. Traditional SEO was a race to be findable. AI search is a race to stay on the shortlist. Everything starts with knowing which side of that line you are on today.

Monthly for most brands, weekly in fast-moving categories. AI answers shift not only when a model is updated but whenever a competitor publishes a new article or case study — your position can fall while you do nothing at all. A single measurement is just a snapshot of today; the value starts with the second one. Weekly automatic measurement is planned for paid plans.

Not necessarily. Translating a page and being cited by AI in that language are two different things. AI does not translate your site and read it — it assembles answers from sources written in that language: local media, reviews, comparison articles. So a fully translated website can still leave you as a brand nobody in that market talks about. A translation tool can only tell you whether you translated. We tell you whether your name comes back in the answer when someone actually asks in that language.

Both before and after. Before you enter, it works as a map of who that market's AI currently recommends: which local competitors come up, at what rank, and which media are cited as the evidence behind them — that citation list is your first target list. After you enter, it becomes a fixed-point check on whether your mention rate in that language is climbing. The direction does not matter. A Japanese brand going abroad and a foreign brand entering Japan are asking the same question: when someone asks in that country's language, does my name come back?

Structurally, yes — they can. AI leans on sources written in the language of the question, so asking in Japanese surfaces Japanese media and asking in Korean surfaces Korean media. When the sources change, the brands in the answer change with them. How large the gap is depends heavily on your industry, and we have not measured every industry. So we do not guess: we send the questions in each of the five languages and put the answers side by side. Measuring instead of assuming is the entire point of this feature.

They measure different languages. Japanese tools measure Japanese; English-market tools measure English. Both work well inside their own language, but a gap like "top-ranked in Japanese, invisible in Korean" only becomes visible if the same question goes out in several languages. miruAI runs Japanese, English, Korean, Chinese and Cantonese on the same questions against the same yardstick. Every paid plan is designed to include all five — we do not sell languages as a paid add-on, because the comparison across them is the product itself.

Sending them is the easy half. The hard half is counting the answers correctly. Three things we got wrong and then fixed: (1) Brand notation — normalising "토스" into "Toss" or "支付宝" into "Alipay" turns a real mention into zero, so we extract names exactly as the answer wrote them. (2) Source domains — if you cannot parse two-level domains like co.kr or com.cn, entire local outlets vanish from the tally. (3) Subdomains of one outlet — counting biz.example.com separately from example.com hides which media are actually moving your position. Measuring multilingually is mostly getting this unglamorous plumbing right in every language.

The key is asking without your brand name. If you ask AI "tell me about [your company]," it will of course describe you — that proves nothing about visibility. Real customers do not know your name yet; they ask "what do you recommend for X?" or "compare the options for X," naming only the category and the need. Whether your name comes up naturally in that answer is the only number worth tracking. miruAI auto-generates questions in exactly that form for each buying stage, sends them to four AI engines, and records whether you were mentioned, where you ranked, and which sources the answer was built on.

That is the actual product. Stopping at "0% mention rate" makes a thermometer, not a doctor. miruAI breaks down why you are absent: you are not present in the media the AI cites as evidence, your site has no page that answers the question the way it was actually asked, or you are being described in a different context entirely. Then it hands you an ordered list of what to fix first. The outlets citing your competitors are your next target list. Once you act, run the same questions again and see whether it moved.

That is part of what we measure. AI will state, in a confident tone, that you still offer a service you discontinued, quote a pricing tier that never existed, or confuse you with a different company. miruAI compares the answer text against your brand profile and surfaces the statements that do not match. To stop us from inventing errors of our own, any finding we cannot ground in evidence is downgraded from "wrong" to "unverified" — a false alarm about a false statement is not something you can act on.

Sign up free with no credit card, pick one language and one AI engine, and measure up to 10 questions a month — a run uses about three, so roughly three runs. Even on the free plan we send real questions to real AI engines, so what comes back is an actual answer, not a simulation. You get your AI visibility score, whether and where you were mentioned, a comparison against top competitors, exposure by buying stage, and the top three improvement suggestions. Paid plans are planned to open all five languages and all four AI engines, cross-language comparison, and weekly automatic measurement.