If you want to make your content available internationally, you’ll eventually ask yourself: Which languages are worth translating into?

That question came up for me while I was considering automatic translation of blog posts via LLM.

Available data (as of 2020, researched in 2025)

Wikipedia provides an overview of Internet users by language (March 2020) (Table 1). So far, I haven’t found more recent figures.

The same numbers show up in other places as well, e.g., at wéb-tränslatiôns, StarAvis, Optimational (where they’re even labeled as 2025), and in a publication on ResearchGate. Wikipedia cites Internet World Stats as the original source. That site no longer appears to publish these data, but the exact numbers are still accessible via the Internet Archive.

Rank Language Internet users Share of all internet users
1 English 1,186,451,052 25.9 %
2 Chinese 888,453,068 19.4 %
3 Spanish 363,684,593 7.9 %
4 Arabic 237,418,349 5.2 %
5 Indonesian 198,029,815 4.3 %

How current are these numbers?

The list has likely shifted since 2020. For one, the world population grew by 300 million between 2020 and 2024. Over the same period, the global share of internet users rose from 59% to 68%. This growth was especially strong in emerging economies such as India, Indonesia, Nicaragua, and the Philippines. That means hundreds of millions of new internet users have come online in recent years—many with a first language other than English. The top spots for English and Chinese likely remain stable due to sheer scale. But from rank 3 onward, the order may well have changed.

It’s also striking that Hindi is missing from the table. In 2020, India’s internet-user share was just over 43%. At the end of 2019, Ethnologue estimated about 612 million Hindi speakers in India. That works out to roughly 263 million Hindi-speaking internet users in 2020, which should have put Hindi in the top 5:

1
2
3
612 million Hindi speakers
x 43% internet users (India)
≈ 263 million Hindi-speaking internet users

In the table’s notes, there’s a mention of a “Zero Sum” approach— assigning each person to exactly one language— even though many people are multilingual. It’s quite possible that much of India’s population was simply counted under English. One clue: there aren’t 1.5 billion people worldwide with English as a first language (L1). Apparently English as a second language (L2) was included, while other languages in the table were counted only as L1.

Attempting to derive up-to-date numbers

Since newer statistics on internet users by language don’t seem to be available, I tried to infer current figures from other data.

The World Bank provides both the share of internet users by country (based on ITU data, a UN specialized agency) and population figures. Both datasets are available as CSV and thus easy to use programmatically. With these, we can first compute the absolute number of internet users per country, for example for Hong Kong:

1
7,524,100 population × 96% internet users = 7,223,136 internet users

Next, we need the language distribution by country to estimate internet users by language. Of course this is only an approximation, since internet use isn’t evenly distributed across language groups. From the country-level results, we can then aggregate to global totals.

Potential sources for language shares per country include:

However, manually compiling country language shares from Wikipedia looks like a fairly involved project, and it’s not realistic for me right now.

Plan B

Back to the original question: Which languages are worth translating online content into?

Lacking reliable, up-to-date data, I’ll start by following the most-spoken languages: English, Mandarin Chinese, Hindi, Spanish, and Standard Arabic.

As internet penetration rises, the share of internet users by language should naturally converge toward each language’s share of the population. If and when robust new statistics on internet users by language appear, I can adjust my selection.