Keyword Research in Malaysia: When Customers Search in Two Languages

Most keyword lists for Malaysian businesses are built the same wrong way. Someone brainstorms in English, runs the list through a tool, then hands the Malay version to a translator.

The translated list looks tidy. It also contains phrases no Malaysian has ever typed. “Perkhidmatan pemasaran digital” is correct Malay and close to dead as a query, because the people looking for it type “digital marketing agency” or “agency marketing murah”.

This article covers how bilingual search behaves here, how to harvest the real phrasing free, and how to decide which language a page belongs in.

How Malaysians actually type a search query

Watch anyone here use Google on their phone and three patterns appear that a monolingual keyword list cannot capture.

The first is code-switching inside one query. “Kos aircond service”, “harga smartphone terbaik”, “cara claim insurance kereta”. Malay grammar, English nouns, and the query is neither a Malay keyword nor an English one. Tools that sort keywords by language file these badly or miss them.

The second is that technical vocabulary stays in English no matter how Malay the sentence is. Nobody types “pengoptimuman enjin carian”. They type SEO. Same for booking, delivery, warranty, quotation and refund. If your Malay page uses the formal translation and the searcher uses the English word, the page will not match.

The third is that query language follows the size of the decision. Local and casual searches lean Malay, high-value comparison research leans English. A restaurant owner might search “kedai printer murah shah alam” and, an hour later, “commercial printer supplier malaysia” for the same purchase.

Why translating an English keyword list into Malay produces dead keywords

Translation optimises for correctness. Search optimises for what people type, and those two things separate fast in Malaysia.

Take “air conditioner service cost”. A translator returns “kos perkhidmatan penyaman udara”. Proper Malay, and formal register, the kind used in a government circular rather than in a phone search at 11pm when the unit stops blowing cold. The real query is “kos service aircond”.

The gap is register, not vocabulary. Written Malay here has a formal layer for official documents and a spoken layer for everything else. Queries come from the spoken layer, English loan words already baked in, and a translator hands you the formal layer every time.

Translation is also one to one, and the mapping between languages is not. One English keyword can split into three Malay phrasings with different intent, and three English keywords can collapse into one Malay phrase. You see that in the results, never in a spreadsheet.

Harvesting the real phrasing from autocomplete and People Also Ask

Autocomplete is built from what people actually searched, so Google gives you the real phrasing free. Be systematic instead of typing two or three guesses.

1. Set search settings to Malaysia. Autocomplete is location sensitive, so suggestions from a Singapore VPN are the wrong data.
2. Seed in the mixed form. Type “kos service aircond” and read every suggestion before pressing enter.
3. Run the alphabet. Add a space then each letter from a to z. 26 passes surface long-tail variants the first list never shows.
4. Prefix with question words. Try “macam mana”, “berapa”, “kenapa”, “bila”, then the English equivalents. Question phrasing holds most of the informational traffic.
5. Read People Also Ask. Each expansion loads more. Record the questions word for word, mixed-language ones included.
6. Take the bottom of the page. Related searches are a second harvest, usually more commercial than autocomplete.

Do this once in Malay-leaning phrasing and once in English. The two lists overlap less than you expect, and the overlap tells you which topics need only one page.

Record queries exactly as Google shows them, spelling quirks included. “Aircond”, “air cond” and “aircon” are three different strings, and the one Google suggests is the one your customers use.

Your Search Console query data beats any third-party estimate

Every tool that sells search volume is modelling it. Search Console reports what actually happened on your own site, and Google keeps 16 months of that history.

Open Performance, switch to Queries, set the range to the full 16 months, export, then sort by impressions rather than clicks. High impressions with almost no clicks means Google already shows you for a phrase you never targeted, usually on page two.

That export is the best bilingual keyword source a Malaysian business owns, because it holds the mixed-language queries no tool categorises properly. Filter the query column for Malay markers such as “kos”, “harga”, “cara”, “murah” and “berdekatan”. Most owners find Malay queries they never wrote a word for.

The caveat is honest. Little traffic means little data, so this method starts weak and gets stronger. Give a new site three to six months of publishing before mining the query list. If Search Console is not connected at all, that is the first item on any technical SEO checklist worth following.

Deciding whether a topic needs a Malay page, an English page, or both

Not every topic deserves two pages. Publishing both by default doubles cost and creates internal competition for no gain.

Use three tests, in order. First, search the topic in each language from a Malaysian IP and look at what ranks. A Malay query returning ten English pages means Google cannot find good Malay content, which is an opening.

Second, check whether the two queries return the same URLs. The same set means one intent, so one page serves both. Different URLs means two intents and two pages.

Third, ask who signs the cheque. Consumer services, home repair, food and tuition skew Malay. Software, B2B services and anything with a procurement process skew English.

Keeping your Malay and English pages from competing

Two pages on one topic cannibalise each other if you let them. Make them genuinely different documents, not mirrors.

Give each version its own primary keyword and its own slug in its own language. A Malay page at /kos-service-aircond/ and an English page at /aircond-service-cost/ send a clear signal. Near-identical slugs do not.

Write them separately rather than translating. Different opening, different examples, different order of arguments, same facts and same prices. If the second version reads like the first run through a translator, Google has to choose between two versions of one page, and it will suppress one.

Link them to each other once in the conclusion. Everything else stays within language: Malay pages link to Malay pages, English to English. The same discipline applies to your on-page work, where titles and headings carry the language of the page and nothing else.

If both versions exist and one has clearly won, do not delete the loser. Rewrite it around a narrower keyword, or merge it and redirect.

Keyword difficulty and why a new domain should start long-tail

Difficulty scores estimate how hard ranking will be, built mostly from the backlinks of pages already there. Every tool calculates it differently, so read the number as a band, not a measurement.

For a new Malaysian domain with no link history, ignore anything competitive for the first six to twelve months and target queries of four or more words. “SEO” is unwinnable. “Kos SEO untuk kedai online” is rankable with a good article and patience.

You can judge difficulty without a paid tool. Search the query and look at who holds page one. Ten national brands and directories means leave it. Two or three small business pages and a forum thread means there is room.

Malay long-tail queries are usually the softest targets, because most competitors publish in English only. Low volume each, but there are many, and they convert better because the intent is specific. If this is new to you, start with the basics of how ranking works.

Mapping keywords to pages so one intent gets one page

Keyword research fails at mapping more often than at collection. People end up with 300 keywords and no plan, build a page per keyword, and create twelve pages competing for one intent.

1. Group by intent, not wording. “Kos service aircond”, “harga servis aircond” and “aircond service price” are one intent in two languages, so at most two pages.
2. Name the page before writing. One URL, one primary keyword, one language, written into a sheet before anyone drafts.
3. Assign secondaries to it. Everything in the group that shares intent goes into headings and body copy of the same page.
4. Check for an existing page. If something already ranks for anything in the group, improve it instead of publishing a competitor.
5. Mark the gaps. Groups with no page in that language are the publishing queue, ordered by how close the buyer is to paying.

Sanity check: if two pages would answer the same question with the same first paragraph, they are one page. Merge before publishing, because merging later costs rankings while the redirect settles.

Free tools that do the job without a Semrush budget

Paid tools are convenient, and their Malaysian volume data is thin enough that no strategy should rest on it. Mixed-language local queries are where sampled estimates fail hardest, so an RM0 tool budget costs you little.

Google Trends earns its place here. It gives no volume, but comparing “servis aircond” against “aircond service” filtered to Malaysia tells you which phrasing wins, the exact question a translated list cannot answer.

A worked example: aircond servicing in Puchong

Collect. Seeds are “service aircond” and “aircond service”. The alphabet run returns “kos service aircond”, “servis aircond rumah”, “aircond service near me”, “harga cuci aircond” and “aircond service puchong”. People Also Ask adds “berapa kerap kena service aircond” and “how much is aircond chemical wash”.

Group. Four intents: price, maintenance frequency, local hiring, and chemical wash as a specific job.

Decide language. Searched from a Malaysian IP, price and frequency return mostly Malay content, local hiring returns mixed results, chemical wash returns English. Price and frequency get Malay pages, local hiring gets both, chemical wash gets English.

Map. Five pages, not nine. A Malay price page, a Malay maintenance guide, a Malay and an English Puchong service page, and an English chemical wash page. The Puchong pages need local SEO work behind them, because they compete in the map pack too.

Check. Six months on, Search Console shows the Malay price page collecting impressions for “harga cuci aircond”, a phrase nobody optimised for. That becomes the next page, or a section on the existing one.

Frequently asked questions about bilingual keyword research

Conclusion: keyword research in a two-language market

Keyword research in Malaysia is a listening job before an analysis job. Collect the phrasing people actually use, from autocomplete, People Also Ask and the 16 months of query data in your own Search Console. Decide language per topic by looking at what already ranks. Give each intent one page in one language, and start with four-word long-tail queries while the domain is young.

Versi Bahasa Melayu di sini: Keyword Research Malaysia.

For a second opinion on which topics deserve a Malay page, talk to the Lesgo Media team and we will go through your query export.

Baca lagi

• Kenapa Website Tak Keluar Dalam Google? Cara Diagnose Sendiri
• Technical SEO Checklist untuk Business Malaysia (2026)
• Local SEO untuk Business Malaysia: Panduan Praktikal

Sumber rasmi

• Google Search Central
• Google Business Profile