Why AI Requires a Resilient Internet

Artificial Intelligence (AI) has become one of the most discussed topics in the past three years. Organizational adoption patterns and the growth of AI-related job postings reveal that AI has moved beyond experimental phases into substantive workplace integration. According to McKinsey & Company surveys, over 75% of respondents report AI use, with more than 70% specifically using generative AI—a dramatic increase from roughly 30% to over 70% in recent years, especially in countries like China and India.

This dramatic increase in AI usage at companies (and among individuals) hinges critically on the availability of a stable, resilient, and free Internet. In this post, we show how one measure of AI usage (collected by Reuters for the Digital News Report 2025, the share of survey takers who reported seeking news from AI Chatbots in the past week) correlates with their score on the Pulse Internet Resilience Index (IRI). 

AI Usage vs Internet Resilience

One might assume that greater use of generative AI chatbots for news would go hand-in-hand with more robust Internet infrastructure, but the correlation plot suggests otherwise. Figure 1 shows a downward trend when comparing the Reuters Institute’s “fraction of respondents who use AI chatbots for news” (2025) with the most recent (2023) Pulse IRI country scores.

Figure 1 — Fraction of participants reporting using AI chatbots for news (AI Usage) and Pulse Internet Resilience Index scores.

This indicates that countries with a higher proportion of news consumers experimenting with AI tools do not consistently have the most resilient networks; there’s a slight, albeit erratic, inclination in the opposite direction. The data points are widely dispersed: for example, the United States shows very high AI usage alongside only average Pulse IRI scores, whereas smaller European nations (like the Netherlands or Denmark) exhibit lower AI adoption but possess top-tier connectivity.

It is also important to note that these data are only available for a few countries and may not fully represent the potential links between AI use and Internet resilience.

AI Usage vs Internet Shutdown Risk

Paradoxically, Internet shutdowns are becoming increasingly common and could threaten the rise of AI use and adoption in work and entertainment. This could impact economic growth and development downstream, as AI use becomes a marker of competitive advantage and innovation globally.

For example, where companies may deploy chatbots for customer relations, a national-level Internet outage might imply that such companies must institute fallback measures that rely on human-centred customer support. Where Internet disruptions are erratic, companies may incur larger costs in mitigating the losses associated with such shutdowns.

There is already broad evidence that such shutdowns can affect economic activity, but it remains to be seen how much these impacts correlate with increasing AI usage. Figure 2 shows a positive correlation between shutdown risk (as measured by Pulse) and AI usage. Countries at the greatest risk of facing an artificial Internet disruption also appear to be where residents rely on AI chatbots for news.

Facing a higher probability of Internet shutdowns means that the greater reliance on AI could significantly affect their ability to access news reliably.

Figure 2 — Fraction of participants reporting using AI chatbots for news (AI Usage) and Internet shutdown risk (score).

Measure This Space

Although this is only one specific measure of AI usage, it captures many consistent and comprehensive ways in which AI–human interactions are being captured in data. Ideally, country-level data on usage statistics of large-language models (LLMs) or similar would be best to understand if there is any correlation between AI use and Internet resilience and availability.

The dynamics between AI use and the Internet have yet to be fully understood. Still, it is increasingly evident that an uninterrupted and free Internet will be required for countries looking to gain a competitive edge and benefit from AI’s rise. Perhaps AI developments will progress enough so that most AI can be run locally or on consumer-grade hardware. Still, until we reach such a stage, the Internet will play an essential role in AI’s accessibility.

Contributors: Anirudh Tagat

Abhishek Vajjala is a senior research assistant in the Department of Economics at Monk Prayogshala.

The views expressed by the authors of this blog are their own and do not necessarily reflect the views of the Internet Society.