How to check whether ChatGPT, Gemini and DeepSeek mention your brand — in English and Chinese

Why test in two languages

Your customers may ask the same question in English one day and in Chinese the next. AI engines often answer the two very differently: a brand that appears for “best food packaging supplier in Malaysia” can be missing for “马来西亚有哪些食品包装供应商”, or the other way round. If you only test one language, you only see half of the picture.

The method below takes about an hour and a spreadsheet. GeoLens automates part of it (DeepSeek and Gemini today; other engines are being connected), but everything here can be done by hand.

Step 1: write 10 prompts, then translate them

Cover three kinds of question. Category: “Which suppliers in Singapore make X?”. Use case: “I need X for Y with a budget of Z — who should I talk to?”. Brand: “What is <your brand>?” and “<your brand> vs <competitor>”.

Then write the same 10 questions in Chinese. Use the script and wording your customers actually use — Simplified Chinese for Singapore, Malaysia and mainland China buyers, Traditional Chinese for Taiwan and Hong Kong. Do not machine-translate word for word; rewrite them the way a customer would type them.

Step 2: fix the test conditions

Start a new chat for every prompt and note whether web search was on. Answers with web search usually cite live pages; answers without it rely on what the model already knows. Both matter, but never mix them in the same column.

Use the same engines every time — for example ChatGPT, Gemini and DeepSeek — and write down the date.

Step 3: ask each prompt at least twice

AI answers vary. A brand named once and missing the next time is normal, so a single answer proves nothing. Ask every prompt two or three times per engine and language.

Step 4: log five fields per answer

Named (yes/no, and which name was used); position in the list; whether the description is accurate (location, products, prices); which sources were cited; and which competitors were named instead.

Mention rate = answers that named you ÷ answers you collected. Keep English and Chinese results in separate columns so you can compare them. A direct question such as “What is <brand>?” only counts as a mention if the answer actually describes you, not if it repeats the name and says it found nothing.

Step 5: read the gaps

Missing in both languages usually means there is too little consistent public information about you. Missing in Chinese only often means your Chinese pages are thin, machine-translated, or use the wrong script for that market. Missing in English only often means your English site is too short or your company name differs across directories and registries.

Wrong facts — an old address, a discontinued product — are a signal too: AI is reading something outdated about you.

Step 6: fix the source, then re-test weekly

Publish one consistent set of facts in both languages: a one-line definition of what you do, who you serve, where you operate, your main products and what makes you different. Put it on crawlable pages that need no login, and keep your company name identical across your website, LinkedIn, business directories and marketplaces.

Re-run the same prompts every week under the same conditions and look at the trend, not a single result. This is a method, not a guarantee: no one can promise that an AI engine will recommend a brand.