Research PaperVisit our research →
Rethinking Scam Prevention: Large-Scale, AI-powered analysis for a Safeguarding Approach

Abstract
This report, authored by Oxford Information Labs Research (OXIL Research) with support from Google.org, uses AI-powered semantic analysis of 28.6 million domain-based signals collected through the Global Signal Exchange (GSE) to analyse scammer behaviour in 2025. The analysis measures semantic relevance of domain-based signals for target groups, age groups and thematic lures employed by scammers. The findings challenge traditional assumptions about victimhood and advocate for a transition from individual awareness to a collective safeguarding approach.
Categories:
Key Findings
- Scammers utilise a “scattergun approach” to target the general population. The general population is the most targeted group, accounting for 34% of total weighted relevance in the signals. Scammers prioritise scale rather than precision, using high-volume, generic communications (e.g., fraudulent dental or restaurant sites) to target the average internet user.
- Scammers weaponise situational vulnerabilities (such as stress, cognitive load or social isolation) over static traits. Signals targeting situational vulnerabilities account for 38.7% of weighted relevance across all groups. Financial distress or aspiration is the leading driver at 24.8% followed by emotional dysregulation and situational crises (11%), and vulnerabilities linked to emotional connection (2.9%).
- Gambling content dominates high-relevance signals, with tactics mirroring those used in teenage targeting. When filtering for the most explicit, high-relevance signals, gambling-themed content dominates (48.1%). This content shows a moderate correlation with teenagers (aged 13– 17) –the highest across all age brackets– suggesting scammers use similar linguistic tactics to exploit gambling addictions and to target teenagers.
- Older adults are not the most targeted group. Contrary to media narratives, older adults are not the most targeted group, ranking 11th in targeted vulnerability. Instead, 58.4% of signals target working-age adults (18–60), mirroring the most active internet population. When older adults are targeted, scammers primarily use lures offering "relief from difficulty" (e.g., solutions to chronic health or income issues). Adults are still at risk and presumably targeted through a combination of online and offline means.
- Protected characteristic groups are targeted with high-precision “sniper” tactics: While smaller in volume (2.73%), groups with protected characteristics (e.g., neurodiversity or mental health conditions) are targeted with high-precision "sniper" tactics. This includes dual targeting of both the individual and their support networks.
Access Publication
Publication Document
rethinking-scam-preevention2 (1).pdf
Research Details
- Published
- March 17, 2026
- Last Updated
- March 16, 2026
- Pages
- 20