A Google.org-backed study of 28.6 million fraud signals overturns the idea that older adults are most at risk online. Here is what the data actually shows about age and scams.
Most people, if asked who scammers target online, would say the elderly. That assumption turns out to be wrong - at least when it comes to the internet. A study published on 18 March 2026 by Oxford Information Labs Research (OXIL Research), backed by Google.org, analysed 28.6 million domain-name-based fraud signals collected throughout 2025 and found that older adults rank 11th among targeted groups. Working-age adults between 18 and 60 are, by a wide margin, the most reached by online scam infrastructure.
This does not mean older people are safe. The picture is more nuanced than a simple ranking suggests, and the study's findings on how scammers adapt their tactics by age group are worth examining in detail.
The signals used in the study came from the Global Signal Exchange, a platform co-founded in 2024 by Oxford Information Labs, the Global Anti Scam Alliance, and Google to pool scam and fraud intelligence in real time. By January 2025, the exchange had launched formally as a UK-based non-profit. The 28.6 million signals analysed include URLs, domain names, and host names reported to the platform across the full year of 2025. The authors describe it as the largest dataset of its kind in academic literature.
Each signal was scored by Google's Gemini 2.5 Flash Lite AI model on a relevance scale from 0 to 4. Human reviewers independently scored a sample of the same signals. The agreement between the AI and the human panel was exceptionally high - 0.9481 for target group scoring, 0.8954 for age bracket scoring, and 0.8932 for thematic lures - placing all three well into what statisticians classify as "very high agreement" territory using Gwet's AC2 coefficient. The results, in other words, are not simply the output of an algorithm running unchecked.
Working-age adults between 18 and 60 account for 58.4% of all age-targeted signals. Broken down further: people aged 29-39 represent the single largest slice at 21.86%, followed by the 18-28 bracket at 20.74%, and middle-aged adults between 40 and 60 at 18.61%. Teenagers aged 13-17 account for 15.80% of signals. Early old age (61-70) accounts for 14.45%, middle old age (71-79) for 4.10%, and the very elderly (80 and above) for just 3.24%.
The research is careful about what this means. It does not imply scammers have decided to leave older people alone. It reflects where the online population is most concentrated. Scam infrastructure, the study argues, mirrors the demographic distribution of the internet itself. Where the most people are, the most scam domains are built.
Ranking 11th in domain-based signals is not the same as being safe. The study explicitly flags that older adults are likely targeted heavily through offline and analogue channels - phone calls, postal mail, in-person fraud - none of which appear in a dataset built from internet domain names. The signals analysed capture digital scam infrastructure. They do not capture the full picture.
When older adults are reached by digital scams, the tactics used are distinctive. The study cross-referenced scam signals with age brackets to identify which thematic lures correlate most strongly with each group. For adults aged 61 and above, "relief from difficulty" dominates. This category covers scams that promise miracle cures, financial relief, housing solutions, or fixes for chronic health problems. The Pearson correlation coefficient for this lure type reaches 0.5413 for adults aged 61-70, and 0.5272 for those aged 71-79. Both figures fall into the "strong" correlation range as defined by the study's methodology, where anything above 0.5 is considered strong.
The logic is straightforward once explained. Scammers targeting older adults exploit what the study calls "life-stage vulnerabilities" - fixed incomes, chronic health conditions, and anxieties about long-term financial stability. A domain promising an affordable solution to a persistent health problem or a housing scheme for retirees does not look like an obvious scam. It looks like something that fills a real need. Examples from the dataset include senior-bank-accounts-449.bond and affordableseniorapartmentsdc.com.
"Informational fraud" - scams disguised as helpful information about brands, local services, or community resources - ranks as the second most prevalent lure for older adults. Correlation coefficients in the 0.4 range were found for adults aged 61-79. The study links this to research on how older adults tend to navigate the internet, relying more heavily on branded searches and local service lookups, which makes convincing-looking fake local business sites effective against that group. Authority-based scams, where the fraudster impersonates a government agency or a bank, rank third - lower than media narratives about elderly fraud might suggest.
The study's most striking age-related finding concerns teenagers. When the dataset is filtered to include only the highest-relevance signals - those scoring 3 or 4 on the 0-4 scale - gambling-themed content emerges as the dominant category, accounting for 48.1% of high-relevance signals. Examples include 577betsport.com, 597betonline.com, and pinupcasino-cz.com.
What does gambling content have to do with teenagers? The correlation analysis provides an answer. Among all age groups, teenagers aged 13-17 show the highest correlation with gambling-themed scam signals, with a Pearson coefficient of 0.3369. That is described as moderate in the study's framework, where 0.3 to 0.5 is moderate and above 0.5 is strong. The study's interpretation is that the linguistic hooks used in gambling-adjacent fraud - urgency, reward, risk, excitement - are structurally similar to the tactics used to draw teenagers into fraudulent digital environments more broadly. Whether scammers are consciously copying gambling-style language when targeting young people, or whether the overlap is an artefact of how both types of content are constructed, the data does not fully resolve. But the correlation is there in 28.6 million signals.
This has a direct connection to broader advertising concerns. Meta reported removing 134 million scam advertisements from its platforms in 2025, with gambling-related content among the categories drawing the heaviest enforcement. Google's AI-powered systems suspended 39 million advertiser accounts in 2024, a 208% increase from the year before.
The study's central argument is that age is a less reliable predictor of scam vulnerability than the situation a person is in at a given moment. Three categories of situational vulnerability together account for 38.7% of weighted relevance across all signals.
Financial state is the most common driver at 24.8%. This covers both people in acute financial distress and those who are financially aspirational - chasing an opportunity to make money or grow a business. Both states create need that scammers can exploit. Emotional dysregulation accounts for 11%. Stress, fatigue, bereavement, relationship breakdown, or the disorientation of navigating an unfamiliar bureaucratic system all reduce a person's ability to spot manipulation. The study references the psychiatric concept of the "window of tolerance," developed by Dan Siegel, to explain why people under emotional pressure struggle to apply knowledge they would otherwise use correctly. Need for connection accounts for 2.9%, targeting people who are socially isolated or seeking relationships.
The implication is important. A 35-year-old going through a divorce, recently bereaved, or under severe financial pressure is more vulnerable than a calm, financially stable 70-year-old. Scammers, according to OXIL Research, engineer the conditions of exploitation. They do not simply wait for the right demographic to come along.
According to Emily Taylor, CEO at OXIL Research: "Consumers are not the weakest link when it comes to online crime - and education and awareness need to be complemented by additional interventions."
According to Haviva Kohl, Senior Program Manager at Google.org: "Effective digital safety begins with evidence, not blame. Google.org's support of OXIL Research is rooted in the need to better understand the 'how' and 'why' behind modern scam tactics, shifting the focus from individual error to systemic exploitation."
The study proposes replacing the dominant "awareness campaign" model of fraud prevention with a safeguarding approach - a framework already in use across UK health, education, and social care. Under a safeguarding model, the responsibility for protecting people does not fall primarily on individuals. It falls on the organisations and systems those individuals interact with: banks, platforms, social care services, government agencies.
In practice this means banks designing systems that flag unusual transactions before they complete. It means platforms adding contextual warnings when users are about to interact with a suspicious domain. It means community organisations being trained to recognise when someone in their network may be at risk - not because of who that person is, but because of what they are currently going through.
Singapore became the first government to join the Global Signal Exchange in September 2025, bringing the platform's total signals tracked in real time to 380 million. That kind of cross-sector, public-private signal sharing is precisely what the OXIL Research study argues should become the norm.
The research is described as an initial phase of a longer programme. Future iterations will incorporate other signal types beyond domain names, including IP addresses, and will introduce longitudinal analysis. The research team has committed to publishing the Gemini prompts used in the analysis to allow external review of the methodology.
Who: Oxford Information Labs Research (OXIL Research) produced the study, with financial support from Google.org. The Global Signal Exchange - co-founded by Oxford Information Labs, the Global Anti Scam Alliance, and Google - supplied the data.
What: An analysis of 28.6 million domain-based fraud signals from 2025, using Gemini AI and human review, examining which age groups scammers target and with what tactics. The headline finding is that older adults rank 11th online, while working-age adults account for 58.4% of signals - and that situational circumstances, not age, are the primary driver of vulnerability.
When: Signals were collected from January to December 2025. The report was published on 18 March 2026.
Where: Data came from the Global Signal Exchange, a UK-based non-profit clearing house for fraud signals. Findings have implications for fraud prevention policy and digital platform design worldwide.
Why: Existing fraud prevention strategies place the burden of protection on individuals through awareness campaigns. The study argues this model misunderstands how scams work, since scammers deliberately engineer situations in which people's defences are lowered. A collective safeguarding approach - where organisations take responsibility for protecting the people they serve - is proposed as a more effective alternative.