The Attention Landscape · established evidence
The Denominator Problem: What ITU's Connectivity Data Can and Cannot Tell You About Attention
The International Telecommunication Union is the only body in the audience-measurement field producing a government-reported, near-census statistic: roughly 6 billion people, about 75 percent of the world, online in 2025, up from 5.8 billion in 2024. That number is genuinely foundational, and genuinely misused. It answers exactly one question, who has the means to be exposed to media at all, the denominator. It says nothing about what those 6 billion people actually see, for how long, or with what attention, the numerator every other instrument in this space is built to estimate. Treating connectivity as a proxy for attention is the single most common category error in casual readings of "internet statistics," and understanding why the two are different is the necessary first step before any attention map can be built responsibly.
The one near-census in a field of samples
Every other source in the media-measurement field, Nielsen's panels, GWI's survey waves, Pew's recruited panel, the BLS time-use diary, is built from a sample and extrapolated to a population. The ITU's connectivity statistics are structurally different: they are compiled from government and telecom-operator reporting of actual subscriber and infrastructure data across member states, which makes them the closest thing among these sources to a near-census rather than a sampled estimate.
That distinction is worth taking seriously rather than treating as a technicality. A near-census figure carries a different kind of authority than a sampled one, and it is precisely because ITU data carries that authority that it gets misapplied so often, quoted as if its precision about access implies an equal precision about attention, which it was never built to measure.
What the ITU actually reports
The ITU's Facts and Figures 2025 report estimates 6 billion people, roughly 75 percent of the global population, were online in 2025, up from 5.8 billion in 2024. The report also breaks that headline figure down by region, with usage ranging from 88 to 93 percent across the CIS, Europe, and the Americas, down to just 36 percent in Africa. This is foundational denominator data: it establishes who, in principle, could be exposed to any given piece of digital media, before any question of what they actually choose to look at is even asked.
It is worth being precise about the boundary of what this figure claims. "Online" in ITU's reporting means having internet access and, typically, having used it within a defined recent period, a connectivity and basic-usage measure. It is not a measure of how much time a connected person spends on any medium, which sources they trust, what they read, or whether they were exposed to a given business's presence at all. Access is the precondition for attention. It is not attention.
Why the denominator is necessary but never sufficient
Every attention or media-consumption figure produced anywhere else in this field, eMarketer's time-spent estimate, GWI's panel survey, Reuters Institute's news-consumption report, is implicitly built on top of a population that had to be online to be measured in the first place. In that sense, ITU's connectivity data functions as the base layer, the universe of people who could theoretically appear in any of those other studies' samples.
But a universe of potential attention is not a map of actual attention. Knowing that 6 billion people are online tells you nothing about how those 6 billion people allocate the finite hours in their day, which platforms, engines, or sources capture that allocation, or whether a specific business, in a specific vertical, in a specific market, is present anywhere those people's attention actually lands. The gap between "connected" and "attending" is not a rounding error. It is the entire subject matter of every other measurement instrument in this space, and no connectivity statistic, however precise, can substitute for the work those instruments exist to do.
Where the denominator problem causes real damage
This is not a purely academic distinction. It shows up as a specific, recurring failure mode in how connectivity statistics get used in business and marketing contexts: citing "X billion people are online" as if it were evidence of reach, opportunity, or exposure for a specific campaign or business, when the correct reading is only that X billion people are eligible, in principle, to be reached, with everything about whether they actually were left completely unaddressed.
The regional breakdown in the ITU data makes the stakes of this error concrete. A market with 93 percent connectivity in the CIS region and a market with 36 percent connectivity in parts of Africa present radically different denominators for any attention strategy, and neither number says anything about which specific surfaces, search engines, AI answer systems, social platforms, within either market actually carry the attention that matters for a given business. Treating a high connectivity percentage as if it were a high-opportunity signal, without separately measuring where attention concentrates inside that connected population, is building a strategy on the wrong half of the equation.
What a responsible attention map has to add on top
The ITU's contribution to any serious attention-measurement effort is to supply the boundary condition, not the answer. A model of where attention actually sits has to be built on top of the connectivity denominator, using instruments that observe behavior directly: survey panels like GWI or Pew's American Trends Panel, time-use diaries like the BLS ATUS, passive device data like Nielsen's hybrid panel, or, for the newest surfaces, direct sampled observation of what buyers see when they ask an AI answer engine a real question.
No source anywhere in this field, including the most rigorous instruments named above, currently measures attention to AI answer engines using a classical, disclosed method. That is the genuine frontier gap this entire body of measurement history points toward, and it is exactly where the denominator problem becomes most acute: the population that could be exposed to a synthetic answer is, in principle, bounded by the same ITU connectivity figures as everything else, but nothing in the existing record yet tells anyone, reliably, what share of that connected population's attention actually lands inside a given synthesized answer, for a given business, in a given market.
The model: access is the floor, not the map
The discipline this comparison should instill is simple to state. Any connectivity statistic, however authoritative its near-census pedigree, answers the question "who could see this," and only that question. It should never be cited as if it answered "who did see this," "how much attention did this get," or "is my business visible to this population." Those are separate, harder questions, and every other source in this field exists because someone had to build a different instrument to answer them, each with its own disclosed sample and its own known bias.
For a business trying to understand its own standing, the practical translation is direct: a large connected population in your market is a necessary condition for opportunity, never a measurement of whether you are actually being found within it. The only way to close that gap is direct observation of where your specific buyers' attention concentrates, not an inference drawn from how many people, in general, have internet access.
The evidence
Key findings, with their sources
-
The ITU estimates 6 billion people, roughly 75% of the world population, were online in 2025, up from 5.8 billion in 2024, with usage ranging from 88-93% across the CIS, Europe, and the Americas down to just 36% in Africa.
established International Telecommunication Union, "Facts and Figures 2025."
-
ITU connectivity statistics are compiled from government and telecom-operator reporting across member states, making them closer to a near-census than the sampled or panel-based estimates every other source in the media-measurement field relies on.
established International Telecommunication Union, "Facts and Figures 2025," methodology framing.
-
No source in the current measurement field, including Nielsen, Ofcom/BARB, Reuters Institute, eMarketer, GWI, or DataReportal, currently publishes a rigorous, disclosed method for measuring attention to AI answer engines or generative search.
established Cross-source synthesis of the measurement-lineage research record, 2026.
-
Connectivity data measures access and eligibility to be exposed to media, not actual attention, consumption, or exposure to any specific source; conflating the two is a distinct and common category error.
established Raveneye Global analysis, applying ITU's disclosed scope against the attention-measurement literature.
Reference
Glossary
- Denominator (attention measurement)
- The base population that could, in principle, be exposed to a given medium or source, typically measured by connectivity and access data such as the ITU's. Necessary for any attention estimate but not itself a measure of attention.
- Near-census
- A statistic compiled from thorough administrative or operator-level reporting rather than a sample, as with ITU connectivity data, as distinct from the sampled, extrapolated figures most other media-consumption sources produce.
- Digital divide
- The gap in internet access and usage between regions or populations, illustrated in ITU data by the range from 88-93% connectivity in the CIS, Europe, and Americas down to 36% in Africa.
- Attention frontier gap
- The absence, across every source in the current measurement space, of a rigorous, disclosed method for measuring population attention to AI answer engines and generative search, the specific gap the Visibility Corpus is positioned to address.
Straight answers
Frequently asked questions
How many people are online according to the ITU?
The ITU's Facts and Figures 2025 report estimates roughly 6 billion people, about 75% of the global population, were online in 2025, up from 5.8 billion in 2024. Usage ranges from 88-93% in the CIS, Europe, and the Americas to just 36% in Africa.
Is internet access the same as internet usage or attention?
No. ITU's "online" figure measures access and basic recent usage, a connectivity statistic. It says nothing about how much time a connected person spends on any medium, which sources or platforms capture their attention, or whether they were exposed to any specific business. Access is a necessary precondition for attention, not a measurement of it.
Why is ITU data described as closer to a census than other sources?
ITU connectivity figures are compiled from government and telecom-operator reporting across member states, rather than from a survey sample extrapolated to a population, which is how nearly every other source in this field, from Nielsen panels to GWI surveys, produces its figures. That gives ITU data a different, generally more authoritative evidentiary basis for the specific question of access.
Can connectivity statistics tell me if my business is visible to buyers in a market?
No. A high connectivity percentage in a market establishes that a large population could, in principle, encounter your business online. It says nothing about whether they actually do, which surfaces their attention concentrates on, or whether you appear on those surfaces. That requires direct, dated measurement of your own standing, not an inference from a national connectivity rate.
Does any source measure attention to AI answer engines the way Nielsen measures TV?
Not yet, as of the current research record. No source in this field, including Nielsen, Ofcom/BARB, Reuters Institute, eMarketer, GWI, or DataReportal, publishes a rigorous, disclosed method for measuring population-level attention to AI answer engines or generative search. This is the frontier gap that requires new, primary observation rather than an existing vendor figure.
Provenance
Sources
- International Telecommunication Union, "Facts and Figures 2025" (established)itu.int
Every figure above is attributed to a real, dated source and tagged with its evidence tier. Where a claim could not be verified to a primary source, it is not stated as fact.