Estimating the Impact of “Free” AI Services on Growth and Productivity
The seismic rise in artificial intelligence (AI) rekindles an existing problem in economic measurement. When for-profit firms provide a good or service that does not reflect its cost of production, it is not consistently recorded either in the gross domestic product (GDP) accounts as an output or in the productivity accounts as an input.
In a recent working paper, U.S. Bureau of Economic Analysis authors Jon Samuels and Rachel Soloveichik, along with Leonard Nakamura from the Federal Reserve Bank of Philadelphia, make an imputation to the GDP and productivity accounts to include “free” digital content.
Free content is defined as content for which users pay zero out of pocket and which thus appears to be free from the perspective of the user and the national accountant. The paper first conceptualizes free content as barter, then values free content consistently with other components of the National Economic Accounts, and finally calculates what GDP and productivity would be if free content was tracked. To avoid double counting, the authors exclude the portion of free content that is bundled with other paid products or otherwise already included in the National Economic Accounts.
Using their barter model and currently available data, the authors produce preliminary estimates of the effects of free content on output, inputs, and productivity at the aggregate and industry levels. They use this barter transaction methodology to measure the impact of advertising-supported media and marketing-supported information, including AI, on GDP.
The authors find that including free content in consumer entertainment has a substantive impact on both recent GDP quantity and recent productivity growth, with trend breaks around 1995 and 2022.
[View larger chart]
Including free content increases average GDP quantity growth between 1929 to 1995 by 0.04 percentage point per year, increases average growth between 1995 to 2022 by 0.09 percentage point per year, and increases average growth between 2022 and 2025 by 0.22 percentage point per year (chart 1).
The trend break around 1995 is likely due to the development of the internet, and the trend break around 2022 is likely due to the development of AI. These adjustments noticeably ameliorate the recent growth slowdown but do not reverse it.
[View larger chart]
Including free content increases productivity growth between 1948 to 1995 by 0.01 percentage point per year, increases average growth between 1995 to 2022 by 0.08 percentage point per year, and increases average growth between 2022 and 2025 by 0.18 percentage point per year (chart 2).
The authors also estimate a valuation of user-generated digital content. In 2025, they estimate that users generated $407 billion of content. This is lower than the $1,489 billion of free content produced by for-profit firms, but substantial enough to warrant tracking and discussion, with the intention of improving the measures.
In conclusion, the authors argue that their paper addresses one of the many challenges of understanding and measuring economic growth in the free digital economy: how to account for content when there is no explicit payment from the user to the supplier. Their paper demonstrates that the relatively simple tweak of accounting for free content as a barter transaction can address many of the measurement issues.
