For years, major technology firms told child safety advocates that technical obstacles made it impractical or risky to restrict access to online services for teenagers. Today, a widening set of national and regional regulators conclude that the difficulties can be overcome and are introducing strict new age-verification obligations for social media platforms, AI chatbots and providers of adult content.
Three months after Australia implemented a landmark prohibition on under-16s holding social media accounts, lawmakers and regulators in parts of Europe, Brazil and several U.S. states are considering similar measures. Political leaders on both sides of the aisle have signaled interest; one high-profile U.S. governor publicly endorsed the idea last month, and another national figure has been reported to be looking into age limits for online services.
These policy moves are being propelled by intensifying worries over online abuse and teenage mental health, a recent surge of concern about AI-generated sexual images of minors, and growing confidence among regulators about the capabilities of so-called age-assurance systems. Supporters of such technology say it can estimate a user’s approximate age by combining facial analytics, parental approval workflows, identity documents and other online signals.
Technological improvements and market maturation
Vendors and independent researchers report that recent developments in artificial intelligence have enhanced the accuracy of age-gating tools while reducing costs, making deployment feasible at scale across many types of online services. Industry participants interviewed for this article said trade associations, technical protocols and certification schemes have contributed to a more standardized market for age-assurance solutions.
Age-inference methods that draw on "digital breadcrumbs" - such as account creation dates, patterns of content interaction and other behavioural signals - can often place a user in an approximate age band without active biometric checks. Companies that specialize in age assurance, including several firms serving large platforms, supplement inference with automated checks such as facial scans and machine analysis of government ID documents.
At the app-store level, major mobile platform operators have added features allowing parents to pass an age-range indicator to app developers. Analysts note that broader advances in identity verification technology have spilled over into the age-checking market, lowering the per-check costs and expanding use cases beyond high-value transactions to routine account gating.
Industry executives say basic machine-only checks typically cost well under $1 per verification, and at high volumes can fall to single-digit cents. More labor-intensive verification steps - for example, human review or comprehensive cross-referencing of personal data - remain available but are used less frequently and command higher fees.
Empirical measures of progress
Independent testing supports descriptions of measurable improvement in facial-age estimation. A long-running study by a U.S. standards agency found that face-scanning software submitted to its tests had an average age-estimation error of 4.1 years in initial 2014 assessments; that average declined to 3.1 years by 2024 and has moved further toward 2.5 years in the most recent testing window referenced by participants.
Some vendors report even tighter average errors within regulators’ target adolescent ranges. One provider scheduled to release a new face-analysis model in April said its latest version achieves an average error of about one year for users between 14 and 18. Another identity-verification firm cited an average error of roughly 1.77 years for the 13-to-17 cohort. An Australian government-commissioned report reached a broadly similar view, concluding that photo-based age estimation products were generally accurate but warning of a higher-uncertainty "grey zone" for users within three years of the legal cutoff.
Practical implementation and early outputs
Australia’s eSafety regulator said it will gather population-level data for two years to evaluate the impacts of its social media account ban for teens and plans to publish initial findings later this year. Within weeks of the ban’s entry into force in December, regulators reported that companies had blocked millions of suspected underage accounts. The eSafety office said 4.7 million accounts suspected of being underage were locked, although some industry sources cautioned that part of this figure likely reflected Google accounts that were prevented from signing into YouTube regardless of account activity.
Individual platforms have also disclosed large numbers of removals. One major social-media owner reported taking down about 550,000 accounts across its branded services suspected to be underage in the law’s opening weeks; another platform stated it removed approximately 415,000 accounts.
Regulators in Europe and the U.K. are closely observing the Australian experience. Officials from one continental body intend to discuss age verification during an upcoming visit to Canberra. The United Kingdom, which already imposes age checks for pornography sites and is considering stricter safety rules for social networks and AI chatbots, has been exchanging notes with Australian counterparts.
How services and vendors approach checks
Executives at age-assurance firms say social media platforms typically perform fewer biometric scans or ID checks than industries such as online gambling or adult content, because social platforms already accumulate large volumes of user data they can use to infer age. That allows social networks to rely more heavily on inference methods - analysing account activity, linked financial information and other signals - to meet regulatory expectations.
Privacy-preserving methodologies are one key variation in implementation. "On-device" processing, which performs checks entirely on a user’s device without transmitting data to cloud servers, provides enhanced privacy but can reduce the systems’ ability to detect users attempting to appear older than they are. According to vendor executives, common evasion techniques used by teens include masks, heavy makeup, fake facial hair or presenting toy faces for scanning.
Vendors argue facial-age estimation can serve as a virtual analogue to the age checks performed daily in brick-and-mortar settings. If a person looks younger than a location's age threshold in the physical world, they may be asked for ID; similarly, online services can prompt further verification when an initial signal indicates someone may be underage.
Limitations and areas of uncertainty
Despite clear improvements, technology providers and independent reports acknowledge persistent limitations. Systems show increased error rates for users whose ages fall near statutory cutoffs - a "grey zone" where supplementary assurance, such as parental consent or ID checks, may be required. Performance also degrades for certain skin types and with lower-quality images from older phone cameras.
Vendors report on-device checks are less effective at detecting deliberate attempts by youngsters to appear older. In addition, independent and government reviews highlight that automated photo-based estimation is not uniformly reliable across all contexts, reinforcing the need for layered verification strategies where uncertainty is greatest.
Industry behaviour and regulatory signaling
Trade groups representing verification vendors caution that early figures from Australia should be interpreted cautiously because several major platforms appear to have implemented only the minimum technical steps necessary to comply. One industry association executive said some platforms even asked vendors to disable controls that would have strengthened age checks, suggesting the social networks may be cautious about the prospect of similar measures being adopted widely and therefore not motivated to make the policy look demonstrably successful.
That source characterized parts of the industry response as a test of regulators’ willingness to enforce higher standards. He said platforms are "extremely worried" that a successful policy could become contagious and spread globally, and that reluctance to assist robust enforcement might be driven by that concern.
Where this leaves policymakers and market participants
Regulatory momentum appears to be building as governments and lawmakers weigh the balance between protecting minors online and the technical, privacy and commercial challenges of implementing age assurance at scale. Vendors point to shrinking costs and improved accuracy as enabling conditions for broader deployment, while independent testing and government-commissioned reviews show measurable progress alongside specific weaknesses that require supplementary measures.
As regulatory experiments proceed, transparency about how platforms implement checks and the quality of evidence they generate will be central to assessing effectiveness. Australia’s two-year data collection window will be watched closely by officials and vendors in other jurisdictions, and early lessons should inform how inference, biometric checks and parental consent mechanisms are combined to manage the technology’s known limitations.
For companies operating in affected sectors - including social media, online advertising, app stores, identity-verification vendors, and businesses dependent on youth engagement - the coming period will test technical integration, compliance costs and reputational management as more jurisdictions consider age-verification mandates.