AI Digital Divide: Who Really Benefits from Artificial Intelligence?

When I was researching the digital divide for my PhD, the internet was already being celebrated as a democratising force. Social networking sites, especially Facebook, promised new opportunities for social connection. I focused on self-representation, a concept that is now familiar to almost anybody who uses social media, but which was relatively new at the time.

But it led me to a rather different question: even when people have access to the same technology, do they have the same opportunities to benefit from it?

I examined how privileged Mexican users constructed their identities on Facebook, drawing on French sociologist Pierre Bourdieu’s concepts of economic, cultural and social capital.

The argument was that what people presented online was not simply a matter of personal choice. Their education, economic circumstances, cultural knowledge and existing social networks influenced how they represented themselves, the connections they established and the distinctions they made. For example, the trips they shared, the tastes they displayed and their connections with other users, including the early forms of what we now call influencers, all contributed to their online identities.

The internet offered opportunities to participate, but participation did not erase the social structures people brought with them.

From the Digital Divide to the AI Divide

Today, artificial intelligence is being presented with remarkably similar promises. Knowledge, expertise and sophisticated technological capabilities are said to be becoming available to anyone with an internet connection, opening the world to seemingly infinite opportunities.

The technology has changed. The underlying question has not.

The original digital divide was frequently understood through access: who had a computer, who could connect to the internet and who was being excluded. At the time, access to these technologies was still associated with economic privilege, even as internet use was becoming increasingly popular.

As digital technologies became more widespread, differences in skills, patterns of use and the benefits people obtained from being online became more important, rather than disappearing.

This distinction remains relevant in the age of artificial intelligence.

Why Access to AI Is Not the Same as Benefiting From It

A person who uses AI to produce a quick summary does not necessarily gain the same advantages as someone who can integrate it into their professional work, evaluate its limitations, develop applications or use it to establish a business.

And someone with access to a free chatbot is not necessarily operating under the same conditions as a company with substantial computing resources, proprietary data and specialised technical staff. Even the choice of AI models, including free and paid versions, can be linked to different economic resources and levels of knowledge.

Subscription costs, access to more advanced models, suitable devices and the time needed to learn how to use these systems all influence the opportunities available to users.

Here, the forms of capital I examined in my thesis still apply. Economic capital affects what people can afford. Cultural capital influences their knowledge and ability to make effective use of technology. Social capital shapes the networks, relationships and opportunities through which these capabilities can become valuable.

How Researchers Define the AI Divide: Access, Use and Outcomes

Recent academic research suggests that AI inequality operates across several interconnected dimensions.

A 2026 review of digital inequality identifies access, use and outcomes as distinct but related dimensions of the emerging AI divide, while emphasising institutional power and market concentration.

The World Bank’s 2025 research also shows how differences in connectivity, computing infrastructure, locally relevant data and digital skills are shaping countries’ opportunities to adopt and benefit from AI.

The concern is not only whether people have access to the technology, but whether they possess the resources and knowledge to use it meaningfully.

This matters because access alone tells us very little about who ultimately benefits.

Who Owns the Technology?

The question becomes even bigger when we take AI ownership into account.

The AI economy introduces a particularly significant issue: who controls the infrastructure on which everyone else depends?

Developing advanced AI systems can require enormous investments in computing power, specialised chips, data, engineering and energy.

These requirements create advantages for companies with substantial financial and technical resources.

How Concentrated Is the AI Market?

The same World Bank report identifies a concentration of AI innovation, computing infrastructure and investment in high-income economies. Its findings also highlight the dependence of many countries on cloud computing services supplied by a relatively small number of companies.

The implications extend beyond corporate competition.

Companies that control computing infrastructure, models and distribution platforms can influence the cost, availability and conditions under which other organisations develop and use AI.

The OECD’s analysis of AI markets raises concerns about vertical integration, where firms operate across several levels of the AI industry, potentially making it harder for competitors to enter or challenge established providers.

Does Ownership Explain AI Inequality?

However, the academic debate is not entirely one-sided.

Open-weight models and shared research tools can reduce some barriers to entry, while a 2024 survey of AI researchers found that many researchers were more concerned about access to talent and data than computing resources alone.

Ownership matters, but it does not explain every inequality.

The more useful question may be who has the capacity to develop, adapt, scrutinise and meaningfully challenge AI systems, rather than simply consume services built by others.

And this is where the enthusiasm surrounding artificial intelligence deserves to be questioned. The promise of democratised knowledge and technological opportunity can obscure the critical and ethical questions about who controls these systems, how they operate and whose interests they serve.

What the New AI Digital Divide Really Means

My PhD examined how established social, cultural and economic distinctions could persist online, and even acquire new forms of visibility, at a time when the debate was already moving beyond who could connect to the internet.

Artificial intelligence adds another dimension to that discussion.

The question is no longer only who can participate in a digital environment, but also who owns and controls the systems increasingly mediating that participation.

Yet it would be a mistake to replace one narrow understanding of inequality with another.

Ownership and market concentration matter. So do affordability, knowledge, education, social networks and the ability to turn technological access into meaningful opportunities.

The new AI digital divide is not simply between those who have access to artificial intelligence and those who do not. It is between those who can use, shape and benefit from these systems, and those who remain dependent on decisions made elsewhere.

Fifteen years later, the technology is profoundly different. The relationship between technological opportunity and social inequality remains just as relevant.

Perhaps more so.

References

1. Nessi García, L. (2011). Constructing Online Identities on Social Networking Sites: Social, Economic and Cultural Distinctions Made by Privileged Mexican Users. PhD thesis, Nottingham Trent University. British Library EThOS.
2. Nessi, L., & Guedes Bailey, O. (2014). Privileged Mexican Migrants in Europe: Distinctions and Cosmopolitanism on Social Networking Sites. Crossings: Journal of Migration & Culture, 5(1), 121–137. https://doi.org/10.1386/cjmc.5.1.121_1
3. Papaioannou, C. (2026). Dimensions of the AI Divide: Digital Inequality and Psychological Consequences. AI & Innovation, 1(2), e70002. https://doi.org/10.1002/aiv2.70002
4. Gelles, R., Kinoshita, V., Musser, M., & Dunham, J. (2024). Resource Democratization: Is Compute the Binding Constraint on AI Research? Proceedings of the AAAI Conference on Artificial Intelligence, 38(18), 19840–19848. https://doi.org/10.1609/aaai.v38i18.29959