The cruelest thing about AI is not that it denies answers to the poor.

Quite the opposite – it gives everyone an answer.

It gives students paper outlines, employees email templates, entrepreneurs business plans, and ordinary people legal explanations, investment advice, and career guidance. For the first time, answers are so cheap, so abundant, and so convincing.

But herein lies the problem: when answers are available to everyone, what becomes truly scarce is no longer the answer itself, but the ability to judge those answers.

The new information poor are not those locked out of using AI, but those who have already received the answers yet lack the ability to evaluate them or the means to turn those answers into real opportunities.

1. The Information Gap in the AI Era

In the internet era, the information poor were those excluded from the network. The solution seemed clear: connect cables,普及 devices, and improve literacy. In the search engine age, things got slightly more complex – you needed to learn how to refine keywords, filter sources, judge credibility, and preferably know some English. But the barriers were visible and quantifiable.

The information gap in the AI era has a fundamentally different structure.

Large language models are not search engines; they directly generate conclusions for you. You no longer need to "find" answers – they are organized into fluent paragraphs, clear steps, and confident tones, delivered right to your screen. On the surface, the barrier seems dramatically lowered. But hidden within is a harsh structure: when answers become cheap, errors become equally cheap; and the ability to discern "is this answer trustworthy?" becomes rarer and more valuable than ever.

Throughout history, the diffusion of every general-purpose technology has followed the same logic: new technology first rewards those who already possess complementary capital. The printing press first benefited the literate; computers first benefited those proficient in office software and programming; the internet first benefited those with strong English skills and advanced search techniques. AI's complementary capital includes educational background, professional knowledge, critical thinking, organizational endorsement, purchasing power, and the one quality hardest to quantify – judgment.

New technology rarely rewards those who need it most first. It usually rewards those who can best utilize it.

2. The First Divide: Access to AI

The first crack of inequality appears before you even open the application.

In April 2026, AI research institute Epoch AI, in collaboration with polling firm Ipsos, released a survey of approximately 5,000 American adults. The three rounds of questions posed a seemingly simple question: Which AI services have you used in the past week? But the answers revealed not a simple preference for products, but rather a map woven from income, access points, and distribution.

Approximately 80% of Claude's weekly active users come from households with an annual income exceeding $100,000; for Meta AI users, that figure is only 37%. Conversely, about 32% of Meta AI users come from households earning less than $50,000 annually, compared to just 7% for Claude users.

These numbers matter not simply because they prove "the rich use advanced AI, the poor use free AI." That is the most superficial reading. What deserves deeper questioning is: Why do different people encounter different AIs in their daily lives?

One person asks AI to pair dinner from fridge leftovers, brighten a photo's background, or polish a text message. Another asks AI to synthesize client interviews, compare supplier quotes, or identify weak assumptions in a report. Both are invoking the same technology. But one invocation ends at convenience; the other enters a cycle of income, position, and negotiation power.

The difference lies not only with users but also with access points. The path to using Claude requires proactive searching, product comparison, understanding capability differences, choosing to pay, and embedding the tool into workflows – each step filters users. Meta AI's path is almost the opposite: it is built into a social platform, free, low-friction, and users often encounter it passively while scrolling feeds, sending messages, or viewing photos.

This is not a market of taste, but a market of distribution. Users appear to be choosing tools; tools' prices and access points are also choosing users.

Source: epoch.ai

3. The Next Divide: Usage Scenarios for AI

Even if you find a good AI tool, the second divide awaits you at the workplace.

In ordinary offices, AI rarely arrives in the form of a "layoff notice." It first takes over meeting minutes, email drafts, spreadsheet organization, client classification, and report drafts. For managers, this automation frees up time for judgment calls; for new hires and junior employees, however, this automation takes away precisely the entry points they need to prove themselves, practice judgment, and move to higher-level work.

The data is colder than this scenario: An Anglo-American workforce AI tracking survey (Feb-Mar 2026, covering over 4,000 respondents in the UK and US) conducted by the Financial Times and research institutions shows that 63% of workers in the highest salary bracket use AI on a typical workday, compared to only 17% and 16% in the two lowest brackets. This is not a gentle slope; it is a cliff.

The more crucial finding lies in the driving factors. Regression analysis of this workplace survey reveals that the effect of salary on AI usage nearly disappears after controlling for other variables – what truly matters are four factors: age, seniority, industry, and training. Among these, training has the largest effect: in companies offering formal AI training, employees' daily AI usage rate is 37 percentage points higher than in untrained comparable companies. Even informal guidance yields a 24 percentage point increase.

Yet the reality is: as of early 2026, only 14% of employees reported receiving formal AI training provided by their employers, and two-thirds received no training of any kind.

AI training is not a technical issue; it is a distribution issue. Who gets selected for training gets permission to enter the track of productivity growth; for those who don't, the tool remains just an icon on the screen that they are not authorized to open.

AI is an application on the consumer side, but a form of permission on the workplace side. And permission has never been distributed equally.

Source: Focaldata

4. The Final Divide: The Ability to Judge AI

This is the most hidden divide, and the most fundamental one.

Imagine a fresh graduate entering a consulting firm. They use AI to generate a first draft of an industry analysis report – complete structure, ample data, confident tone. Their supervisor – someone with a decade in the industry – glances at it and points out that two cited data sources have methodological flaws in their original research, and the causal inference for a third conclusion is problematic. The supervisor isn't more capable because they worked harder; they have that foundational layer – knowing where errors are likely, knowing when fluency is genuine and when it's just the machine filling gaps.

This is the real meaning behind the counterintuitive finding in the workplace survey: the heaviest users of AI at work are not the youngest employees, but those who have been in their current role for 2 to 10 years. The relationship between AI usage and seniority remains significant even after controlling for age. This isn't because young people don't want to use it, but because the value of AI is highly dependent on the user's pre-existing judgment ability.

Experience is AI's most important complementary capital, and experience cannot be subscribed to.

AI lowers the cost of "sounding knowledgeable," but it does not equally lower the cost of "being truly knowledgeable." There's even a more dangerous consequence: users who lack the foundational layer are more likely to accept AI's output uncritically; and the more they accept it uncritically, the harder it becomes for their judgment to grow. When an agent judges for you, you are consuming intelligence, not accumulating it.

Nobel Prize-winning economist and MIT professor Daron Acemoglu is blunt on this point: Using AI tools requires a certain level of education, abstract thinking, quantitative skills, and familiarity with technology. "That AI will increase inequality is almost certain," he says.

The new information poor take shape here: they are not those without AI, but those with AI, with access, with answers, yet lacking the training to judge those answers; with tools and scenarios, yet lacking the permission to turn tool outputs into opportunities; consuming intelligence daily, yet never accumulating any.

5. The Limits of the Equalizing Effect

But the relationship between AI and inequality is not just about widening gaps.

Multiple experimental studies have found that, under controlled conditions, AI often boosts the performance of lower-skilled individuals more – for call center workers, junior writers, entry-level consultants. This is easy to understand: top experts gain limited marginal benefit from AI; someone who could never afford professional services reading a contract correctly for the first time with AI's help represents a qualitative leap.

But a key distinction must be noted: experimental studies measure "improvement after usage," while real-world data measures "who is actually using it," "who is allowed to use it," and "who, after using it, can turn results into opportunities." Neither set of data is lying; they measure entirely different things.

A technology can narrow gaps in the lab while widening them in the real world – if adoption itself is unequal, if the contexts themselves are unequal, if judgment itself is unequal.

AI possesses equalizing technical characteristics yet operates within unequal social structures. Both realities are true simultaneously, and that is the true shape of the problem.

6. Technology Will Diffuse, But the Dividends Won't Arrive Simultaneously

Every generation tends to believe that its era's general-purpose technology will break the old order.

After the printing press, the literate benefited first for centuries. At the dawn of personal computers, they amplified the capabilities of those who already knew office software and coding. The early dividends of the internet flowed to those who knew English, knew how to search, and had the time and motivation to arbitrage. In every technological wave, the cry of "this time is different" is loud, yet structural divides often take decades to become visible.

AI's divides may form faster and run deeper. Because it affects not just one type of task, but almost all work reliant on judgment and language. And these are precisely the capabilities hardest to standardize and redistribute.

Some believe the gap will eventually close. Economic historian and Oxford Internet Institute professor Carl Benedikt Frey holds this view, citing historical precedent: the inequality brought by PC普及 narrowed over decades as usage barriers fell. This analogy is not without merit.

The problem is, even accepting this optimistic historical analogy, Frey himself acknowledges the crucial qualifier: "It depends on how long the gap takes to close. If it's a decade or two, that's much more worrying."

A decade or two is not a timeframe one can wait for lightly – especially for those who need to find jobs, negotiate salaries, and accumulate experience during that period.

Conclusion

This is a peculiar historical moment: for the first time, we have a technology that can make everyone feel they are getting smarter.

Often, that feeling is the destination itself.

The problem is, in an era truly defined by judgment, mistaking the feeling for the destination might be the most expensive mistake of all.