Is AI Replacing Translators? U.S. Jobs, Wages & Industry Data (2026)

"Replacement" can mean several different things: fewer paid assignments, lower rates per word, fewer entry-level openings, fewer people working in the occupation, or a shift from translating to reviewing machine output. Each shows up differently in the data, and no single government series measures all of them.

This study puts three public sources side by side: Census Bureau revenue for firms with employees, Census Bureau receipts for self-employed businesses, and the Bureau of Labor Statistics occupational outlook. The Census industry, NAICS 541930, covers businesses that mainly translate written material, interpret speech or provide sign language services, so it can't separate document translation from in-person interpreting. [7]

The Census Bureau's Annual Integrated Economic Survey (AIES) publishes a revenue series for translation and interpretation services from 2013 to 2024. It covers businesses with paid employees, and the figures are not adjusted for price changes. [1][4] We converted each year to 2024 dollars using the annual average Consumer Price Index for All Urban Consumers (CPI-U). [2]

In current dollars, revenue rose from $3.85 billion in 2013 to $6.93 billion in 2024, up 79.9%. After inflation, the rise was 33.6%. Inflation-adjusted revenue peaked in 2021 at $7.41 billion, slipped in 2022, fell 9.0% in 2023 and recovered partly in 2024. (our calculations) [1][2]

The employer series leaves out the many translators and interpreters who work for themselves. The Census Bureau's Nonemployer Statistics cover businesses with no paid employees that are subject to federal income tax and have receipts of $1,000 or more. [6] We read the national files for 2016 to 2024 for NAICS 54193. [5]

In 2024 there were 56,914 such businesses with $1.60 billion in receipts, equal to about 23% of the employer revenue figure (our calculation). 97% were sole proprietorships. Their number dipped in 2020 and 2021 and has risen every year since. [5]

Total receipts in current dollars were the highest of 2016–2024. After inflation they were 10.4% above 2016 but 2.2% below 2019. Because the number of businesses grew about as fast as receipts, average receipts per business, $28,040 in 2024, were essentially unchanged from 2016 in 2024 dollars ($28,229). (our calculations) [2][5]

The Bureau of Labor Statistics counts about 73,900 interpreter and translator jobs in 2025 and projects 75,400 in 2035, growth of 2%, which it describes as slower than the average for all occupations. It still expects about 6,000 openings a year, mostly to replace workers who change occupations or retire. [3]

On technology, the BLS outlook says computer tools, including AI, are making the work of translators and localization specialists more efficient, but that "many of these jobs cannot be entirely automated" because computers can't yet produce work comparable to human translators in most cases. It also expects demand from a diverse U.S. population, globalization, national security needs and sign language video relay services. [3]

The median wage doesn't describe freelancers. It comes from the Occupational Employment and Wage Statistics survey, which doesn't include the self-employed. [10] And the BLS self-employed share (about 20,700 jobs at 28% of 73,900, our calculation) is far smaller than the Census count of 56,914 nonemployer businesses, in part because the Census count includes any qualifying business with at least $1,000 in receipts, however small, and the two sources define their units differently. [3][6]

Pedro Llanos-Paredes and Carl Benedikt Frey of the Oxford Martin School studied how machine translation affected the translation profession and the demand for foreign language skills. Using differences in machine translation use across 695 local labor markets in the United States after 2010, when Google Translate was released as an app, they report a negative relationship between Google Translate adoption and translator employment, backed by an instrumental variable approach and placebo tests. They also found that better machine translation reduced demand for every foreign language skill they examined, including Spanish, Chinese, Japanese, French and German. [8]

That study is about machine translation over the 2010s. It doesn't measure the effect of the generative AI tools released since late 2022, and the Census and BLS series in this study can't isolate AI's effect either.

The stakes of an error differ by setting. NIST's Generative AI Profile lists confabulation, meaning confident but false output, among the risks of generative AI. [11] In immigration matters, a federal regulation and USCIS policy both assume a person who takes responsibility for the language work:

• Translated documents. A foreign-language document submitted to USCIS must come with a full English translation that the translator certifies as complete and accurate, along with a certification that the translator is competent to translate. [12] See our certified translation services.
• Interpreted interviews. At an interpreted USCIS field office interview, the interpreter must be fluent and impartial, interpret fully and accurately, and sign Form G-1256 under oath in the officer's presence. [13] Our guide to Form G-1256 explains the declaration.

Neither one says anything specific about AI, and neither means AI has no role in the work. They do mean that a named person answers for the result.

All three files are free to use with credit. Each row carries the URL of its source.

1. Employer revenue. FRED series AIES54193RCT00US, "Translation and Interpretation Services, All Establishments: Revenue", published by the Census Bureau's AIES, in thousands of dollars, 2013–2024 (last updated September 10, 2026). The AIES covers private nonfarm businesses with paid employees; estimates are not price-adjusted and are subject to sampling and other error. [1][4]
2. Inflation. Annual average CPI-U, U.S. city average, not seasonally adjusted (1982–84=100). 2015–2024 values match the BLS summary table; 2013 and 2014 are the averages of the 12 monthly values of the same index. CPI-U measures consumer prices, not the prices of translation services. [2]
3. Self-employed businesses. Census Bureau Nonemployer Statistics, national files for 2016–2024 (NAICS 54193 throughout), all legal forms, receipts in thousands of dollars. Nonemployer figures are noise-infused for confidentiality. Earlier years were not in the same file location and are not included. [5][6][9]
4. Two different universes. Employer revenue and nonemployer receipts come from different programs with different definitions, so we show them side by side and give their ratio only as a rough scale comparison. Neither series counts workers.
5. Jobs and wages. BLS Occupational Outlook Handbook, 2025–2035 projections and May 2025 wages. Employment includes the self-employed; wages don't. [3][10]
6. Causation. These are descriptive national series. They don't identify the effect of AI, and we don't claim they do. The Oxford study is described on its own terms. [8]
7. Checks. All figures were read from the original sources on November 6, 2026. In every nonemployer year, the four legal-form rows add up to the all-forms total for both businesses and receipts, and the 2024 receipts-size classes add up to 56,914.
8. Updates. We plan to add the 2025 AIES and Nonemployer figures when the Census Bureau releases them.

About the publisher. Rush Translations and Immigrant Services provides certified document translation and in-person interpreters for USCIS interviews. This study uses public data only and contains no company figures. Related research: U.S. immigration language statistics, the U.S. Language Access Atlas and federal court interpreter statistics, FY2016–FY2025.