Mangesh Yadav

Multilingual SEO as an “entity problem,” not a “page translation problem”

by

I have been thinking about this for a while and wanted to get some reactions from people who work in SEO or build for ecommerce.

The entire multilingual SEO industry is built around one assumption. Translate your pages, create /fr/ and /de/ URLs, set up hreflang, and you are discoverable in every language. That works for Google because Google ranks pages and crawls URLs.

But AI search engines like ChatGPT, Perplexity and Google SGE don't work that way. They don't rank pages in the traditional sense. They try to answer a specific question about every product they encounter which is: what is this thing, and what is it called in different languages and contexts?

The field that carries that answer in Schema.org is alternateName with language tags. It looks like this:

json

"alternateName": [
  { "@value": "Chaussures de course", "@language": "fr" },
  { "@value": "Zapatillas para correr", "@language": "es" }
]

Instead of three separate pages for running shoes, chaussures de course, and zapatillas you have one product entity that AI systems understand across all three languages. LLMs are trained on concepts and relationships, not URL structures. This matches how they actually think.

Here is what makes this a real gap and not just a theory. No translation plugin writes to alternateName. Not WPML, not Polylang, not TranslatePress, not Weglot. They all solve the page problem beautifully. None of them solve the entity problem. So even a perfectly translated multilingual store is invisible to AI-assisted discovery in non-primary languages.

Some SEO plugins like Rank Math support schema but none of them automate multilingual entity mapping at scale. The gap is real and nobody has really filled it yet for WooCommerce.

The way I see it there are two models now running in parallel:

Old model: translate pages, set up hreflang, rank in Google.

New model: define your product as an entity, embed multilingual identity in structured data, get cited in AI answers.

Most of the industry is still operating in the old model. AI search is already in the new one.

Curious whether others who work in multilingual SEO or ecommerce have noticed this shift or are thinking about it differently. Has anyone tried to address the alternateName gap specifically?

11 views

Add a comment

Replies

Be the first to comment