How many languages does a museum audio guide actually need?

KEY TAKEAWAYS

The right number of languages isn't a vendor question; it's a visitor-research question. Three inputs drive the answer: your local language-at-home Census data, your city's international tourism arrivals, and your own gate-side visitor survey.
For a US art museum in a major city, the conventional list is seven to ten — English, Spanish, Mandarin, French, German, Italian, Korean, with Japanese, Portuguese, and Arabic added at the top of the international tourism funnel.
European institutions floor higher, usually ten to twelve, because the cross-border audience and the in-region language base both demand it.
The historical reason most museums shipped English-only — per-language production cost in the studio model — has collapsed. The constraint is now editorial review, not studio time.
The wrong answer for almost every institution is "as many as possible." Ten well-reviewed languages serve visitors better than twenty unsupervised ones.
Convo ships ten languages by default; the work this article is meant to support is deciding whether those ten are the right ten for your specific audience.

The question comes up on roughly half the discovery calls I take. The honest answer is rarely the single defensible number a curator wants for a board memo. It's a short list derived from three inputs: the language demographics of the population around you, the international tourism funnel into your city, and what your own visitors tell you when you ask. This piece walks through how to do that work in an afternoon, what the standard answer looks like for US, European, and Asian institutions, and where the math points when you actually run it.

I'm Eric Duffy. I run Convo, a platform that ships ten languages by default. That number isn't magic — it's the empirical center of what mid-size cultural institutions on three continents have been asking us for. The point here is to help you reason about whether ten is the right ten for your building.

Why the language question has changed since 2020

For most of the audio-guide era, the number of languages a museum offered was decided by a budget line, not by visitor research. Per-language production in the studio model ran roughly 60–80% of the English cost — script translation, native voice casting, studio booking, editing, mastering. A mid-size institution with one English tour and a Spanish track had already exhausted what its multilingual budget could absorb. Three languages was the realistic ceiling.

That constraint has collapsed. On modern AI-narrated platforms, an approved English script regenerates and re-voices in ten languages in roughly a minute, and the marginal cost of the eleventh is software-only. The decision has moved upstream — from "which two languages can we afford?" to "which ten serve our actual visitors?" That's a research question, not a procurement question. The AI audio guides pillar covers how the production math shifted; this piece is about what the new math lets you do.

What three inputs determine the right language list?

The right list comes from three sources, weighted by which question you're trying to answer. For an institution serving primarily a local audience, the local language-at-home data dominates. For a destination institution with mostly out-of-region visitors, the tourism funnel dominates. For most museums, the answer is a blend.

Local language demographics. The US Census American Community Survey reports languages spoken at home for every county, metro area, and tract. Per the Bureau's 2025 release on the 2017–2021 ACS, 22% of the US population age five and older speaks a language other than English at home, with Spanish accounting for roughly 61% of that group. The remaining 39% varies wildly by metro — Mandarin in San Francisco and New York, Vietnamese and Korean in Los Angeles, Haitian Creole in Miami, Somali in Minneapolis. Pull the table for your county.
International tourism arrivals. Your city's tourism bureau publishes annual source-country data for international visitors. For a destination museum, this is the dominant input. In 2024, Paris drew its top international cohorts from the United States (2.7M), the United Kingdom (2.6M), Italy (1.6M), Germany (1.5M), and Spain (1.4M) (RoadGenius / OTCP). The Louvre's own 2024 mix tracked closely: 13% US, 6% China, 5% combined Italy/UK/Germany, 4% Spain (Louvre press release).
Your own visitor survey. Most institutions skip this and shouldn't. Ten questions at the gate, two weeks across weekday and weekend visits, with one ask: "What language would you most prefer to take an audio tour in?" The AAM's 2025 Annual Survey of Museum-Goers — 98,904 respondents across 202 museums — is the national methodology benchmark, and it generalizes well to a single-institution language survey.

Weight the inputs by your audience mix. A neighborhood museum in Queens leans on Census data. A destination institution like the Met or the Louvre leans on the tourism funnel. Most museums sit in between, and the answer is the union of the top languages from both, sanity-checked against the visitor survey.

What does the standard list look like for a US art museum?

For a mid-size to large US art museum in a tier-one city, the conventional list runs seven to ten. The base seven, in roughly the order they get added:

English — the floor.
Spanish — non-negotiable, given that Spanish accounts for roughly 61% of the non-English language-at-home population per the Census ACS.
Mandarin (simplified Chinese) — second-largest non-English language nationally, and the largest international tourist segment for many US museums.
French — international tourism plus Canadian visitors.
German — European tourism, particularly to history and decorative arts museums.
Italian — European tourism, particularly to art museums and historic sites.
Korean — substantial Korean-speaking population in coastal US cities plus inbound Korean tourism.

The next three, added when audience justifies:

Japanese — inbound tourism from Japan; collections with Japanese holdings.
Portuguese — Brazilian inbound tourism (consistently top-five for major US cities) plus the Portuguese-speaking US population.
Arabic — institutions with Middle Eastern collections or Gulf-state audiences.

A handful of destination institutions have run their own visitor research for decades and converged on roughly this same ten-language set. It's not a coincidence that most modern platforms ship a list that overlaps with it significantly.

What changes for European institutions?

European institutions floor higher than US ones, usually ten to twelve languages, for two structural reasons. First, the in-region audience itself is multilingual at a scale the US doesn't have — a museum in Brussels serves a population that natively splits across Dutch, French, and German, and cross-border tourism adds another five or six languages on top. Second, the European tourism funnel is more evenly distributed across source countries. Paris's top five each contribute over a million visitors a year, and the next tier (China, Netherlands, Belgium, Brazil, Japan) is not far behind.

A typical European list:

The same base seven (English, Spanish, French, German, Italian, Mandarin, Korean), reordered so that French/German/Italian are higher in the priority stack.
Plus Portuguese (Brazilian tourism + intra-EU travel).
Plus Japanese (consistent inbound from Japan).
Plus Russian, Arabic, or Dutch depending on the audience.
Plus regional languages (Catalan in Barcelona, Swiss German variants) where institutional identity demands them.

The number to optimize for in Europe isn't really ten or twelve. It's "all of your top tourism source countries plus the in-region language base." That sums to ten to fourteen, depending on the city.

What changes for Asian institutions?

Asian institutions tilt their lists toward the regional tourism funnel and the cross-Asian visitor flow that dominates their gates. A Singapore museum, a Seoul museum, and a Tokyo museum will each ship a list that looks structurally different from a Western institution's:

English as the international floor.
Mandarin, Japanese, Korean as the regional base.
Bahasa Indonesia, Thai, Vietnamese, or Tagalog depending on the source-country mix.
One or two European languages — typically French, German, or Spanish — for Western tourism.

Western European languages drop in priority because the inbound mix is so different. A museum in Tokyo doesn't need German higher than Korean; a museum in Singapore doesn't need Italian higher than Bahasa. The same three-input method works; the weights are just different.

How do you actually run the visitor survey?

Most institutions skip the visitor survey because it sounds expensive or hard. It is neither. The minimum viable version is a laminated card at the front desk and ten weekday shifts of front-of-house staff asking visitors as they arrive.

The questions that matter:

What language would you most prefer to use for an audio tour today? Multiple choice with your candidate list plus an "other (please specify)" field.
Are you a local resident, a domestic visitor, or an international visitor?
If international: what country are you visiting from? Open text.

That's the whole survey. Run it for two weeks across weekday and weekend visits, separate the international and local responses, and the answer to "which languages do our visitors actually want?" will be statistically obvious. The cost is staff time and a few hundred dollars of printing. A 200-respondent single-institution survey is plenty of signal to make the language decision.

What's the right number when in doubt?

If you can't run the survey before you have to decide — common when an exhibition is opening in six weeks — the defensible defaults are: seven for a US art museum (English, Spanish, Mandarin, French, German, Italian, Korean), ten for a European institution (the same seven plus Portuguese, Japanese, Arabic), and three or four for a hyperlocal community museum (English, Spanish, plus whichever one or two dominate your neighborhood's Census data).

Convo ships ten on every paid tier because ten is the modal answer across the customer base — the union of "what a US destination museum needs" and "what a European institution needs" lands at ten. Specifically: English, Spanish, French, German, Italian, Portuguese, Chinese, Japanese, Korean, Arabic. If your institution justifies a different ten, the relevant vendor question is whether they can swap languages in and out, not how many they ship.

For how a single source collection serves multiple linguistic audiences, see the note on one collection, many audiences.

Where this work ends and editorial work begins

Choosing the language list is the easy part. The harder part is making sure each track is good — that the Mandarin reads with the right register for art-museum interpretation, that the Korean honorifics match the formality of the institution, that idioms translate as meaning rather than as words. That's a native-reviewer question, not a translation-model question. A platform that ships ten languages without a native-review workflow is shipping ten drafts, not ten finished tours.

For more on multilingual interpretation — translation versus localization, native review workflows — see the Multilingual interpretation hub. For how the language question fits into the broader visitor experience picture, see the Visitor experience pillar.

Frequently asked questions

Two: English and Spanish. The Census case for Spanish is overwhelming — Spanish is the home language of roughly 61% of the non-English-speaking US population, and most major-city museums see a Spanish-speaking visitor share well above the national average. Below two languages, you're declining to serve a measurable share of your existing audience.

Modern large-language-model translation is good enough for a first draft in all the major languages discussed here. But every track should have a final pass by a native reader before it ships — not because the translation is wrong, but because the register and idiom for museum interpretation matter, and a native reader catches what translates correctly but reads awkwardly.

Everything that has an English version. Translating only the highlights creates a tiered experience where non-English speakers get a shorter tour — exactly the inequity the multilingual program is meant to address. The cost of translating the long tail is now low enough that the case for partial translation has mostly evaporated.

Both are accessibility tracks, not language tracks — covered in the Accessibility & inclusion pillar. A complete multilingual program ships audio description alongside the standard language tracks, not as an afterthought.

No, but you do need ten reviewers somewhere in the pipeline — staff, volunteers, community partners, or contracted native readers. The review can be light (an hour per language per tour, focused on register and idiom) but it has to happen. Most institutions handle this with a mix of bilingual staff, university partnerships, and community advisors.

Default to the largest-audience dialect for your institution. In the US, that's typically Mexican Spanish, Brazilian Portuguese, and simplified-character Mandarin. In Europe it shifts toward Castilian Spanish and European Portuguese. For dialect-sensitive collections, the choice is curatorial, not technical.

What to do next

If you've read this far, the practical next move is one of two things. If you have time, run the two-week visitor survey before any vendor conversation — the data will sharpen every question you ask in a demo. If you don't, default to the seven-language US list or the ten-language European list and refine once you're live.

For the broader category context, the AI audio guides pillar covers why per-language cost is no longer the binding constraint. For the procurement view, the Buying & cost pillar is the next read. To see how this works at Convo, the product page describes the ten-language default and the native-review workflow that supports it.

About the author

Eric Duffy is the founder of Convo, a platform that lets museums and cultural institutions publish multilingual audio tours their visitors can have a conversation with. He writes about the practical decisions curators and directors face when they're sizing up the multilingual question for the first time, drawing on discovery conversations with institutions across the US and Europe. Reach him at eric@convo.app or on LinkedIn.