AI Visibility in the Philippines: GEO for Tagalog Queries

Almost every article about generative engine optimisation is written for a single-language market. The advice is sound as far as it goes: write clear answers, structure your pages, make your entity easy to identify.

None of it addresses the situation most Philippine businesses are actually in. Your customers do not all ask in the same language, and neither do the people deciding whether to hire you.

I have watched a Manila company with a well-built English site come up in AI answers for English questions and disappear completely when the same question was asked in Tagalog. Nothing was wrong with the site. The question was simply in a language the page did not serve.

This article covers what changes when the query language changes, and what to do about it.

Why the Answer Changes With the Language

Three things happen when someone asks in Tagalog or Japanese rather than English.

The source pool narrows. The model draws more heavily on material written in the language of the question. There is far less Tagalog business content on the web than English, so a smaller set of sources carries more weight. Being in that smaller set is easier than being in the English set — and being absent from it is more damaging.

The intent shifts. A question asked in Tagalog often carries different assumptions than the same question in English. "Magkano ang website?" tends to come from a small business owner asking about a first site. "How much does a website cost?" more often comes from someone comparing agencies. Same words, different customer.

Localisation gets guessed. When the model has no local source, it fills the gap from wherever it can — often with figures, regulations or practices from another country. If nobody in your market has written the answer down, the answer your customers get will be someone else's answer.

That third point is the one worth acting on. It is not that you are outranked. It is that the space is empty and something else has filled it.

Related: What Is GEO (Generative Engine Optimization)? The Customer Acquisition Strategy After SEO explains this in detail.

The Four Languages That Matter Here

The fourth row is the one Japanese-affiliated companies in the Philippines consistently underestimate.

When a decision-maker in Tokyo asks an AI assistant whether it is realistic to have work done in Manila, the answer comes from Japanese-language sources. If your company has nothing in Japanese, you are not part of that conversation — regardless of how strong your English presence is.

Taglish deserves a note too. Mixed-language queries are extremely common in the Philippines and are usually treated by models as closer to English than to Tagalog. In practice, a well-written English page with a few clearly stated local specifics tends to serve Taglish queries reasonably well. Pure Tagalog queries do not behave the same way.

Related: How AI-Powered Multilingual Marketing Helps Philippine Businesses Reach Global Markets explains this in detail.

Four Fixes, In Order of Effort

Fix 1: Write the answer in the language of the person deciding

This sounds obvious and is routinely skipped.

Pick the ten questions your customers actually ask. Ask each one in English, then in Tagalog, then in Japanese if you serve Japanese clients. Record whether your company appears.

Where you appear in English but not in Tagalog, the gap is a content gap, not a technical one. Write that answer in Tagalog. It does not need to be a full translated site — a single well-structured page answering one question in the local language often does more than translating fifty pages of marketing copy.

Start with pricing and process questions. Those are asked most often in local languages, and they are the questions where a missing local answer gets replaced with a foreign one.

Related: How English-First AI Projects Help Philippine SMEs Accelerate Adoption explains this in detail.

Fix 2: State the local specifics explicitly

Models cannot infer what nobody wrote down.

If your prices are in pesos, write the peso figure. If your working hours overlap with Japan, say which hours. If you handle BIR filings or DOLE requirements, name them. Vague phrasing such as "competitive rates" and "flexible schedules" gives the model nothing to attach to your business, so it attaches a generic answer instead.

This applies in every language, but it matters most in the languages where local sources are thin. In a sparse source pool, one page that states a concrete local fact can become the source.

Fix 3: Keep the same entity across languages

If your Japanese page calls the company one thing and your English page another, the two do not reliably connect.

Use the same company name, in the same script, on every language version, alongside the local rendering. Keep the address, phone number and registered entity name identical across pages. Link the language versions to each other.

This is unglamorous work and it is the difference between three pages that reinforce one company and three pages that describe three unrelated ones.

Fix 4: Check quarterly, not once

The source pool for non-English queries is small, which means it moves. A single new competitor page can displace you in Tagalog results in a way it never would in English.

Run the same ten questions in each language once a quarter and record the result. Fifteen minutes per language. If you appear one quarter and not the next, look at what changed rather than assuming the effort failed.

What This Costs

Fix 2 is the cheapest and usually the highest-yield, because most sites are already vague about the things their customers most want to know. Start there, then do Fix 1 for whichever language shows the largest gap.

None of this requires paid tools. It requires deciding which languages your buyers actually use, which is a business question rather than a technical one.

Common Mistakes

Machine-translating the entire site. A wholesale translation produces pages that read as translations and rarely answer the questions asked in that language. Ten well-written local pages beat two hundred translated ones.

Assuming Japanese head offices search in English. They do not, in general. If the buying decision happens in Tokyo, the query happens in Japanese.

Writing local-language pages that avoid numbers. The reason a local page wins is that it contains local specifics. A translated page that says "please contact us for pricing" adds nothing the model can use.

Treating Taglish as a separate language to write in. It is not, for this purpose. Serve it with clear English plus explicit local detail.

Testing once and concluding. Answers vary between runs. Ask each question three times before deciding whether you appear.

FAQ

Q: Do we need a full Tagalog version of our website?

A: Not usually. Start with a small number of pages answering the questions most often asked in Tagalog — typically pricing, process and eligibility. A full site translation is a much larger commitment with a lower return per page.

Q: We are a Japanese company operating in Manila. Which language should we prioritise?

A: Japanese, if your work is commissioned from Japan, and English for local corporate buyers. Tagalog matters most if you sell to local small businesses or hire locally at scale. Decide by who signs, not by who you talk to most often.

Q: Will publishing in multiple languages hurt our search rankings?

A: Not if the language versions are properly linked and each page is genuinely written for its language. Problems arise from near-duplicate machine translations, not from multilingual content as such.

Q: How do we test queries in a language nobody on the team speaks?

A: Have a staff member or client who speaks it write the ten questions the way they would actually ask them. Do not translate your English questions — the phrasing people really use is different, and that difference is the point.

Q: How long until this shows results?

A: Two to three months before you see consistent appearances, in our experience. Non-English pools move somewhat faster than English ones because there is less competing material, but the underlying sources still take time to be picked up.

Start With the Language of the Decision

GEO advice written for single-language markets misses the situation most Philippine businesses are in. Your visibility is not one number. It is a different number for each language your customers use, and an English-only presence can be invisible in the language where the decision is actually made.

Pick the ten questions your buyers ask. Ask them in each language your market uses. Where you are absent, write that answer, with the local numbers stated plainly.

At PH AI Works we help Japanese-affiliated and local businesses in the Philippines work out which query languages matter for their buyers, and build the small set of pages that closes the gap. If you already have an English site performing well and want to know what you are missing elsewhere, that check is a good place to begin.

References

No external statistics are cited in this article. The behaviour described — answers shifting with the language of the query, and gaps being filled from foreign sources — is drawn from query testing carried out for client engagements in Metro Manila, run in English, Tagalog and Japanese. Client names and figures are withheld under the relevant agreements rather than estimated. Readers are encouraged to run the same ten-question test on their own business rather than take these observations on trust.