Skills
Everyday digital competence
Measured digital competence is falling while daily device use rises, and the adult figures most often quoted as current were collected between 2011 and 2015.
The sharpest finding here is the distance between what people believe about their competence and what they demonstrate. In ICILS 2023 the great majority of grade 8 students said they could judge whether information online is trustworthy; a twenty-fifth could describe two ways of checking a website (IEA, 2026). The "digital natives" idea that made this seem unlikely was never evidence-based — a 2008 review found it rested on assertion rather than data — and measurement now runs against it, with computer and information literacy falling across cycles while daily device use at school rises steeply. The adult figures in most policy documents are worse than dated: they come from an assessment fielded between 2011 and 2015, in a domain the OECD retired and says is not comparable with its replacement. For a practical example of software used to turn activity into operational metrics, see this overview.
Confidence runs far ahead of demonstrated ability
ICILS tests students rather than asking them to rate themselves: the same students rated their judgement and were then asked to show it.
84% of grade 8 students said they could judge the trustworthiness of internet information at least moderately well, but only 4% could describe two ways to verify a website's reliability and 24% could describe one.
Direction: Decrease. Strength of evidence: Strong.
Caveat The verification tasks need written explanation, and confidence and performance were measured on different scales.
The failure is specific, and not about operating the machine. In the same assessment 94% could open a specified link from a document, 87% could download a file from a chat message and 32% could assess crowd-sourced information (IEA, 2026). Mechanical fluency and judgement are different skills, and judgement is the one that can be taught and tested.
Measured competence fell while device use rose
The refutation of the generational story comes from comparing cycles of the same test.
The share of grade 8 students reaching computer and information literacy Level 2 or above fell by about 11 percentage points between 2013 and 2023, and by about 5 points between 2018 and 2023.
Direction: Decrease. Strength of evidence: Strong.
Caveat Trends cover only the countries in all relevant cycles, and the 2023 cycle overlaps with pandemic schooling disruption.
Over the same period the share of students reporting daily ICT use at school rose by 39 percentage points while measured computer and information literacy fell.
Direction: Decrease. Strength of evidence: Strong.
Caveat An ecological correlation across cycles rather than a within-student design; curriculum and device policy changed too.
That is not proof that devices reduce competence, since several things changed at once. It is evidence against the assumption that exposure produces skill, the basis of the digital natives claim. Bennett, Maton and Kervin reached the same conclusion in 2008, reviewing the original literature and finding no empirical basis for it. Calling young people naturally digitally literate repeats a slogan from the early 2000s that still shapes decisions about technology in the classroom.
The adult figures in circulation are a decade old
The statistic behind most "X% of adults cannot solve problems with technology" headlines comes from PIAAC's problem solving in technology-rich environments.
Only 8% of adults reached the top level of PIAAC's problem solving in technology-rich environments assessment, 37% reached Level 2 and 16% scored below Level 1.
Direction: Decrease. Strength of evidence: Strong.
Caveat A decade old, and retired — PIAAC Cycle 2 replaced it with adaptive problem solving, which the OECD says is not comparable.
Three cautions attach to it. US adults did worse than the participating-country average, at 23% below Level 1 and 5% at the top level (NCES, 2016). Roughly 15% of respondents declined or could not take the computer-based assessment and are excluded entirely, and the report does not separate the inexperienced from those who refused. And the domain no longer exists: the current cycle is Do Adults Have the Skills They Need to Thrive in a Changing World?, covering 31 countries and published on 10 December 2024. Its proficiency percentages could not be retrieved, and estimating them would be worse than leaving the gap open.
Self-reported skill counts something different
Europe's headline indicator is built from activities people say they have done recently, not from tasks performed under test conditions. In 2023 56% of people aged 16-74 in the EU had at least basic overall digital skills, meaning 44% did not, with the Netherlands highest at 83% (Eurostat, 2025 edition). That figure and the tested results above cannot share a chart: one counts activities people report, the others score tasks people attempt. At the excluded end the counting gets softer still. A House of Lords committee reported that around 2.4 million people in the UK cannot complete a single basic task such as opening a browser, without attributing the figure to a named study (House of Lords Communications and Digital Committee, 2023). Lloyds Banking Group put 95% of UK adults online in 2025 using its own banking customers, a base that under-represents the unbanked — the group the figure claims to size, and the problem running through who is still offline.
Scam recognition is going backwards
One everyday competence has a trend line, and it points downwards.
82% of UK adults responded safely to a scam email scenario in 2025, down from 88% in 2022, and 14% felt confident they could spot scams yet did not take the safe action.
Direction: Decrease. Strength of evidence: Mixed.
Caveat A survey scenario rather than behaviour under a real phishing attempt, where success rates are typically far worse.
That 14% is the students' confidence-ability gap again, in adults with money attached; what reduces losses is covered in scams, phishing and what stops them.
The short version
- ICILS 2023 found 84% of grade 8 students confident they could judge online trustworthiness and 4% able to describe two ways to check a website (IEA, 2026).
- Those students were fluent operators, 94% able to open a link from a document, so the deficit is judgement, not mechanics.
- Across countries with comparable data the share reaching the middle competence level fell about 11 percentage points from 2013 to 2023 as daily school device use rose 39 points.
- "Digital natives" was never evidence-based: Bennett, Maton and Kervin found in 2008 that it rested on assertion, and tested competence has fallen since.
- The adult evidence has a hole in it: PIAAC's figures are from 2011-2015 in a retired domain, the EU's 56% is self-reported, and no current proficiency percentage is available.