Regional Synonyms: One Product, Five Names Across Five Countries

A composite scene: a sandal brand’s Sydney distributor asks for “thongs” in the title. The change enters the global template, and the next morning the US storefront reads “Beach Thongs” — lingerie in North America — while the German page still says Zehentrenner. One SKU, three markets, three names, one pointing at a different object.

Regional synonyms are a retrieval key before they are a copy decision: the model finds the page with the buyer’s word and introduces you with it, so a word table without a market column fails both jobs.

Retrieval fails before the copy does

AI search starts with retrieval, and the first gate is whether the buyer’s word appears on the page. Lexical matching carries exact recall; dense retrieval carries semantic tolerance. Tolerance is not coverage: “thongs” and “flip-flops” rarely share a sentence in training data, and “Zehentrenner” shares almost no context with either.

Even when the page is found, generation answers in the asker’s vocabulary. A shopper who types “jandals” and reads “this sandal brand” concludes the answer is about something else. One local word in an extractable paragraph gives the model a token to reuse verbatim.

The test is narrow: in each market, the primary local word appears at least once in a paragraph that survives extraction. Meta keywords, image filenames, and navigation labels do not count. Metadata is a hint; a paragraph is evidence.

Build a market-tagged alias graph, not a translation table

The artifact is a concept–market–word table: the SKU family as row, the market as column, the local word in the cell. Two failure modes separate once it looks like that.

Alias gap. A Product entity has one name. Alternate names need a home: alternateName in structured data and, off-site, local media, dealer pages, and community threads. Without that, the model treats the local query as a different product. Product aliases and model numbers cover names inside one market; this is the cross-market layer, where the word survives and the meaning drifts.

False friends. In Australia and New Zealand, “thong” is footwear; in North America, lingerie. One string, two entities. Adding the word cannot fix that; a market-scoped record and one disambiguation sentence can.

The hard rule: every record carries a market field. A synonym table without one is a list of future incidents, not an asset.

Example: one sandal, five markets

Example — these are common local usages, to be confirmed with a local reviewer before publishing. No real brand or client is described.

Market	Common local word	Confirm before it goes on a page
United States	flip-flops	“Thong” means lingerie in North America; keep it out of aliases
Australia	thongs	Homograph with swimwear and lingerie; the page needs one disambiguation sentence
New Zealand	jandals	Local default; younger shoppers also recognize flip-flops
South Africa	slops	Local only; do not extrapolate to other English markets
Germany	Zehentrenner	Used alongside Flip-Flops; sizes, returns wording, and warranty terms change too

All five words in one opening paragraph teach the model that the page sells everything; the answer unit dies. One page per market: primary word in the title, the rest in explanation slots. The UK and Canada largely follow US usage, and that belongs in the table too.

Five steps to a table you can maintain

Maintain the table like product master data: an owner, a version, a validation step. Otherwise it becomes a spreadsheet nobody edits.

Step	Action	Owner	Passes when
1 Frame concepts	SKU families as rows, markets as columns; core categories first	Content, category	Every core concept has at least three market cells
2 Source words	Site search, marketplace suggestions, community posts, distributor feedback	Local operations	Every word traces to a real buyer question
3 Annotate	Register, ambiguity, trademark, country-specific status	Local reviewer	Each word gets a verdict: on-page, explanation only, or banned
4 Validate	Search the word; check whether the local result page’s category matches yours	Local reviewer, content	The top ten results point to one category
5 Place and retest	Primary word into title and first paragraph; the rest into explanation slots	Content, front end	The market’s prompts return an answer that uses the right word

Step 4 is the one teams skip and the only falsifiable one. Search “thongs” in Australia: if results split between footwear and lingerie, the word is explanation-only, never a title. A translation tool knows the mapping, not the register. Crawl and render access is the other precondition: a perfect table behind a broken locale is inventory you cannot ship.

Placement: one primary word, the rest make way

Synonyms on a page are evidence, not keywords. One primary occurrence plus two explanations beats five words in a hero paragraph.

• Title, opening sentence, and structured data name: the same primary word, once.
• Other names go in explanation slots: an “also known as” spec row, an FAQ question, a comparison header, image alt, alternateName.
• One template per market. A global template that syncs everywhere is the most common contamination source in multilingual stores.
• An ambiguous word gets a disambiguation sentence — “In Australia, thongs are flip-flops” — which is quotable and travels well.
• Keep competitor brands, “dupe,” and “same as X” out of aliases; they create false entity links and carry trademark risk.

Where this breaks

First, regulated categories where a standard or legal name wins: chemicals, medical devices, certified electronics. The standard name stays primary; synonyms only explain. Slang traffic does not pay for a compliance failure. Second, one market and one language: the return here is near zero, so fix the answer unit first.

Third, volume chasing: a variant that means something else pulls mismatched intent and binds your entity to the wrong category. Fourth, machine-translated tables. False friends are where machine translation fails, and an alias only sticks if your brand entity resolves in that market. Local review cannot be automated away.

Market order is a separate decision; six variables for choosing the first market covers it, and the rest of this month’s line sits on the content and semantics hub.

This month, pick one market outside your home country and your best-selling SKU. Pull 20 real local questions from site search, support tickets, and marketplace suggestions, and build a three-column sheet: concept, local word, ambiguous yes or no. Replace the primary word in that market’s PDP title and first paragraph, then re-run the same ten prompts in two weeks. Check which word the answer uses, and whether it still points to you.