KTRENZ AEO: how AI answers describe Korean brands in English

Why it matters

More US shoppers now ask an AI assistant before they open a search page. The answer they get names a handful of brands and points to the pages the assistant read on the spot. A Korean brand with little written about it in English is easy to leave out, and when it is mentioned, the price or the terms can be wrong. Nobody at the brand sees it happen unless someone asks the same questions and reads the answers.

What one measurement found

On October 28, 2026 we measured Krypton Ø (krypton-0.com), a Korean invite-only encrypted social app that had just launched. We asked 20 questions in US consumer wording, mostly without the brand name, across ChatGPT, Google AI Overviews, Perplexity and Claude, and kept all 80 answers word for word.

• In the 68 answers to questions that did not name the brand, Krypton came up 0 times. Signal came up 33 times, Session and Threema 32 each.
• Asked for a private social app for close friends, ChatGPT listed 15 apps. Krypton was not one of them.
• Asked what the paid plan costs, Perplexity said €15.99 a month (€191.88 a year). The real price is $49.99 or $59.99 a year. 3 of the 12 answers to fact questions were wrong.
• Only 1 of the 80 answers cited the official site. The rest leaned on review sites, app directories and older articles.

How it measures

• Questions are written the way US consumers phrase them, and asked in English from a US location.
• Most questions leave the brand name out. Showing up when someone asks for a recommendation matters more than showing up when they already know the name.
• The same questions go to every engine under the same conditions: ChatGPT, Google AI Overviews, Google AI Mode, Perplexity and Claude. Meta AI is checked by hand, because its API is not open yet.
• Every answer is kept in full and judged sentence by sentence: is the brand named, in what position, are its selling points said, are its facts right, who is recommended instead, and which sources are cited.
• AI answers shift from one ask to the next, so a single diagnosis is a starting point. Brands under management are measured again every other day.

The free site check, in two parts

Before anything is paid, the official site is checked two ways, with a score and a fix list for each. Every item says where the problem is and why it matters, so the brand’s web team can fix it themselves. After a fix, checking again marks the item done.

1. Technical analysis — Seven areas of the site itself: security headers and HTTPS, privacy pages and consent, reliability, speed, accessibility, web standards and search basics. For example: “3 security headers are missing” or “Trackers load before the visitor agrees.”
2. AEO readiness — Whether AI crawlers can read the site: robots.txt for each AI crawler, how much text is there without JavaScript, whether the official facts appear on the page, structured data and llms.txt. For example: “Only 12% of the page is readable without scripts.”

What you get from a diagnosis

1. Answer report — Every question with each engine’s answer in full, and the verdict on each: named or not, position, selling points, wrong facts, competitors recommended instead, sources cited.
2. Technical plan — A list for the web team, most urgent first, with paste-ready code and English copy: structured data, meta descriptions, llms.txt, robots.txt lines and security headers.
3. Content plan — For each question, whether to fix it with a page on the official site, coverage in outside media, or a ChatGPT ad, with an owner and an order.
4. Proposal deck — All of the above as an editable PowerPoint under the agency’s name and logo, plus the full answer set as JSON and ZIP.

Who it is for

• Korean marketing agencies that run US-facing work for Korean brands (K-beauty, K-food, K-pop agencies, travel), and the brands themselves
• Reports, proposals and review links go out under the agency’s own name and logo
• Brands can use it directly too: they sign up the same way and start with a free diagnosis

Pricing (KRW)

What it does not do

• It does not promise rankings or recommendations: there are none to buy in AI answers
• It does not suggest fake reviews, undisclosed paid posts or community seeding
• It does not measure Korean AI answers such as Naver; those are measured by Legit.Show Korea (legit.show/kr)

Questions

Is this the same as SEO?

It overlaps but it is not the same. Search ranks pages; an AI answer reads a few pages and writes its own summary. A site can rank well and still be left out of the answer, or be described with an old price. The technical plan covers the parts both share, and the answer report shows what the AI actually said.

Why only the US and English?

Most Korean brands going abroad start with US shoppers, and that is where the answers are least likely to include them. Korean-language answers, Naver included, are measured by Legit.Show Korea.

Can an agency put its own name on the reports?

Yes. Reports, review links and the proposal deck carry the agency’s name, logo and color, and the KTRENZ mark can be hidden.

Do I need an English website?

No. If there is no English page, or crawlers see it empty, making one is the first item in the plan. The copy and code come in English.

How do I start?

Sign up at aeo.ktrenz.com with your company name. The site check, the question set and three ChatGPT answers are free. The service runs in Korean; for questions in English, write to [email protected].