Keyword research methods written for English markets break in Thailand. Not slightly — they miss whole categories of demand. Three properties of Thai search cause most of the trouble.

Thai has no spaces between words

Thai script runs continuously. Word boundaries are inferred, not marked. Every keyword tool has to guess where one word ends and the next begins, and they guess differently.

The practical effect is that volume gets split across segmentations of what is really the same query, and tools under-report accordingly. A term that looks like it gets 2,000 searches a month may be closer to 8,000 once variants are consolidated by hand.

Our workaround is to treat Search Console as the authority. Third-party tools give us the candidate list; Search Console query data on a site already in the market tells us what people actually type. Where we have no site data, we use Google autocomplete and related searches as a sanity check on tool volumes.

Searchers switch scripts mid-query

Thai users routinely mix Thai and Latin script in a single search — a Thai verb with an English brand name, or a technical term left untranslated because there is no common Thai equivalent. In enterprise IT we regularly see Thai grammar wrapped around English product names.

This means the same intent has at least three query forms: fully Thai, fully English, and mixed. They often return different SERPs. We research all three and decide per keyword whether one page can serve them or whether they need separate targets.

Transliteration has no fixed spelling

English words written in Thai script have no standardised transliteration. Common product categories have four or five accepted spellings, all in real use. Tools treat these as unrelated keywords.

We build a variant map early: for each core term, every spelling we can find in Search Console, autocomplete, and competitor page copy. The variants get consolidated for volume estimates and used deliberately in on-page copy, because searchers who type the uncommon spelling still need to find the page.

The method we run

• Build the seed list in both languages from the client’s own product vocabulary, not from a tool.
• Expand with a keyword tool, accepting that volumes will be wrong and directionally useful only.
• Pull twelve months of Search Console queries where a site exists, and treat that as ground truth.
• Map spelling and script variants into single intent groups.
• Check the live SERP for each group — Thai results, English results, and mixed — because they frequently differ.
• Assign one page per intent group, with the dominant variant in the title and the alternates in the body.

Why it is worth the extra work

Most competitors in Thai markets run the English playbook and accept the tool numbers. That leaves whole variant families uncontested. In several of our projects, the fastest wins came from spellings the client had never targeted because no tool reported meaningful volume against them.