What technology changes in romance scams

Reviewed, human‑curated evidence indicates that social‑network services, dating platforms, and social media are common initial contact venues for relationship‑based fraud and that offenders frequently shift conversations from public platforms to less‑monitored messaging channels to continue grooming and seek payments. Readily available generative‑AI and deepfake tools can produce realistic synthetic faces, voices, and video at lower technical thresholds than earlier methods; investigators and reviewers report these capabilities can undermine provenance‑based checks (for example reverse‑image searches) because synthetic images may have no external source to match. Evidence about automation is heterogeneous: experimental and corpus studies document signatures consistent with partial automation in initial acquisition phases (templated or nearly identical replies), while offender‑facing and convicted‑case analyses document organized, multi‑worker operations that employ algorithms, worker assignment, background files, VPNs, and foreign phone numbers, indicating substantial human mediation in many high‑loss or sustained scams. Hybrid romance→investment schemes (“pig‑butchering”) repeatedly use fabricated trading dashboards, fake apps/websites, and offender‑controlled crypto flows; blockchain forensic case studies describe three‑tier transaction role structures and extensive recirculation and swapping that make deterministic attribution inferential. Researchers have developed OSINT and prototype detection workflows with promising in‑sample metrics in some studies, but authors emphasize that algorithm effectiveness depends critically on input‑data quality and representativeness and highlight privacy and acceptability concerns for message‑scanning approaches. Common methodological patterns across the corpus—reliance on purposive, complaint‑derived, forum‑sourced, or small qualitative samples and enforcement‑selected forensic cases—provide detailed operational insight but limit population‑level generalization and precise prevalence estimates. Key evidence gaps identified by authors include the empirical measurement of how extensively LLMs/generative AI are used in live scams and whether they operate end‑to‑end, real‑world validation of detection systems with platform and private‑message data, and field‑tested deepfake‑detection and platform mitigation strategies that preserve privacy and legal rights (Cross, 2022a, 2022b; Dickerson et al., 2020; Dominguez Castillo, 2026; Franceschini et al., 2023; Grace Carvalho Fernandes et al., 2023; Griffin & Mei, 2024; Gujarathi et al., 2026; Huang et al., 2015; Jakobsson, 2016; Kassem & Carter, 2023; Ko et al., 2026; Pope & Seto, 2026; Rabby & Chowdory, 2024; Sorell & Whitty, 2019; Stern & Choi, 2026; Vedhanayagam et al., 2026; Wang & Topalli, 2024).

Platform ecology and channel shifting

Multiple studies using complaints, corpora, and case evidence identify social‑network services and dating platforms as common initial contact venues for relationship‑based fraud; after rapport is established, scammers frequently redirect conversations to less‑monitored messaging apps or phone/VoIP channels to continue grooming and seek payments (Ordekian et al., 2024; Pope & Seto, 2026; Wang & Kelsay, 2025; Yan, 2024).

Platform affordances—public profile content, friend networks, and searchability—are repeatedly reported to enable tailored targeting and personalized scripts; social‑media visibility of apparent wealth or shared connections can also be used to normalize deceptive behavior or recruit accomplices in some operations (Barnor, 2024; Cross, 2022a; Kopp et al., 2016; Ohu & Jones, 2025).

Evidence is context sensitive: platform rankings and dominant contact channels vary by country and over time (for example, Facebook, Instagram, and country‑specific messengers such as KakaoTalk or WeChat are salient in regionally focused studies), so platform‑specific prevention measures should reflect local usage patterns (Amirkhani et al., 2024; Cantin et al., 2025; Choi et al., 2024; Corpuz et al., 2025; Yan, 2024).

Authors report that offenders commonly use multiple applications during a single scam and often move victims to less‑surveilled channels (for example WhatsApp, Telegram, or other messaging services) after trust is established, a tactic linked to isolation from platform monitoring and continued solicitation (Ordekian et al., 2024; Wang & Kelsay, 2025).

AI, deepfakes, and verification challenges

Several conceptual reviews, case reports, and forensic observations document that readily available generative‑AI and deepfake tools can produce realistic synthetic faces, voices, and video at lower technical thresholds than earlier methods; these capabilities can undermine provenance‑based checks (for example reverse‑image searches) because synthetic images may have no external source to match (Cross, 2022b; Cross & Layt, 2021; Kassem & Carter, 2023).

Investigative case material and media‑documented incidents attribute large losses to scams that reportedly used AI‑generated media (including deepfake videos and voice cloning) in some instances, and authors explicitly call for empirically validated, accessible detection tools rather than assuming existing provenance checks suffice (Choi et al., 2024; Cross, 2022b; Mouri, 2024).

At the same time, complaint‑based analyses and observational casework report many instances where reverse‑image searches or other identity checks exposed stolen or reused photos and helped substantiate suspicions, although authors also note cases with no positive matches where the absence of a footprint raised concern (Botha et al., 2025; Cross & Layt, 2021; Khukhunaishvili, 2024).

Because evidence shows both successful image‑search detections and plausible use of uniquely generated synthetic images that lack linkable provenance, authors call for development and validation of accessible deepfake‑detection methods and image‑authentication tools as a research and operational priority (Cross, 2022b; Cross & Layt, 2021).

Automation, semi‑industrialization, and LLM risks

Evidence of automation is heterogeneous. Large email‑corpus and experimental studies document signatures consistent with partial automation during acquisition phases—e.g., highly similar replies, low click/response rates, and template‑driven messages—while operational and convicted‑case analyses emphasize substantial human mediation for sustained grooming and high‑loss operations (Hu et al., 2021; Huang et al., 2015; Jakobsson, 2016; Robinson & Edwards, 2024).

Qualitative offender‑facing studies and court‑file analyses show organized, multi‑worker operations that use algorithms, worker assignment, background files, VPNs, and foreign phone numbers—practices that can be supported by but are distinct from fully autonomous LLM‑driven end‑to‑end scams; the reviewed corpus does not provide validated estimates of the proportion of operations that are end‑to‑end LLM‑automated (Abubakari et al., 2026; Dominguez Castillo, 2026; Ogunleye et al., 2020; Pope & Seto, 2026; Wang & Topalli, 2024).

Experimental evaluations of LLM safety in romance‑fraud contexts show model susceptibility to jailbreak prompts under specific conditions, but authors caution generalization is limited by prompt choice, single‑turn designs, and evaluation criteria—evidence that LLMs present both opportunities to scale malicious content and limits in current operational use (Parra et al., 2026).

Authors note an empirical gap: the precise proportion of romance‑fraud operations using LLMs rather than traditional scripted techniques remains unknown, and measurement of live operational LLM use requires targeted empirical designs reported as a research priority in multiple sources. This gap constrains confident claims about end‑to‑end automation prevalence (Dominguez Castillo, 2026; Sorell & Whitty, 2019).

Pig‑butchering, blockchain forensics, and crypto‑enabled schemes

Multiple case studies and mixed‑method analyses characterize pig‑butchering as a hybrid long‑duration trust‑building process that funnels victims into fabricated trading dashboards, fake apps/websites, or offender‑controlled exchanges; distribution commonly uses web links or QR codes rather than legitimate app‑store channels (Franceschini et al., 2023; Gujarathi et al., 2026; Wang & Zhou, 2022).

Blockchain‑forensics case studies document structured role‑based transaction architectures (for example hub, intermediary, and holding tiers) and extensive recirculation and swapping of funds across addresses and cryptocurrencies, tactics that impede tracing and make attribution inferential rather than deterministic (Griffin & Mei, 2024; Stern & Choi, 2026).

Authors recommend combining on‑chain and off‑chain evidence in living forensic collaborations and real‑time analyst partnerships to improve interpretability and to keep pace with evolving laundering techniques; several sources frame these cooperative approaches as necessary given enforcement‑selected or partial sampling frames in forensic case work (Gujarathi et al., 2026; Stern & Choi, 2026).

A blockchain-forensics case-study using OFAC sanctions as an enforcement-based sampling frame analyzed a 29-address OFAC sample and found extensive recirculation and swapping of funds across addresses and cryptocurrencies, which impedes tracing and obscures fund origins. Because the analysis is limited to that partial OFAC-based sample, the authors caution against generalizing this single-case forensic structure to all pig-butchering networks without comparative replication (Griffin & Mei, 2024; Stern & Choi, 2026).

Detection prototypes, OSINT workflows, privacy, and methodological patterns

Researchers have developed prototype detection approaches and OSINT workflows—examples include Maltego‑based link analysis and TF‑IDF/logistic‑regression classifiers—that report promising in‑sample performance or high inter‑annotator agreement in validation datasets, though dataset, labeling, and external‑validation details vary across studies (Ko et al., 2026; Vedhanayagam et al., 2026).

Detection researchers caution that algorithm effectiveness and accuracy depend heavily on the quality and relevance of input data. Proposed research directions include developing detection profiles from communication and payment patterns, while a scoping review recommends living evidence syntheses and real-time collaboration with forensic analysts to maintain relevance. These are cautions and research recommendations, not demonstrations of real-world detection effectiveness (Gujarathi et al., 2026; Sorell & Whitty, 2019; Vedhanayagam et al., 2026).

At the same time, dynamic analysis of private messages or behavioral profiling raises clear privacy, acceptability, and governance concerns; the evidence emphasizes participatory evaluation of user acceptability, transparency, and bias mitigation before deploying message‑scanning or automated warning systems (Dickerson et al., 2020).

Across the corpus, common methodological patterns include reliance on purposive, complaint‑derived, or forum‑sourced samples, small qualitative samples, single‑case or forensic case studies, and secondary‑data syntheses; these designs provide rich operational detail but limit population‑level generalization and precise prevalence estimates (Abubakari, 2023; Dickinson et al., 2023; Grace Carvalho Fernandes et al., 2023; Vedhanayagam et al., 2026).

Evidence gaps and recommended future directions

Authors repeatedly identify key evidence gaps that constrain confident conclusions: (a) empirical measurement of how extensively LLMs/generative AI are used in live scams and whether they operate end‑to‑end, (b) real‑world validation of detection systems with platform and private‑message data, and (c) field‑tested deepfake‑detection and platform mitigation strategies that preserve privacy and legal rights (Cross, 2022b; Dominguez Castillo, 2026; Rabby & Chowdory, 2024; Sorell & Whitty, 2019; Vedhanayagam et al., 2026).

Recommended future directions supported in the evidence include: developing living evidence syntheses and real-time collaboration with forensic analysts to keep pace with evolving tactics and laundering technologies; investing in native-language datasets and addressing platform moderation gaps that may limit AI-driven detection; acknowledging that dynamic analysis of private conversation content raises privacy concerns; and encouraging researchers and civil-society actors to produce comprehensive, technology-focused analyses to establish scale and guide anti-trafficking initiatives (Amirkhani et al., 2026; Dickerson et al., 2020; Gujarathi et al., 2026; Luong & Ngo, 2024).

Reported experimental evaluations of the prototype logistic‑regression classifier achieved 85% accuracy, 88% precision, 80% recall, and an F1 score of 0.84. The authors note that algorithm effectiveness and accuracy depend heavily on the quality and relevance of input data. Inter-annotator agreement was very high across identifier categories (Cohen’s κ: wallet addresses 0.978, platform URLs 0.988, loss amounts 0.969). The authors also invite NGOs, civil‑society organizations, and academics to provide comprehensive analyses—particularly on technology-related issues—to help establish scale and guide anti-trafficking initiatives (Ko et al., 2026; Luong & Ngo, 2024; Vedhanayagam et al., 2026).

Related-work reviews characterize romance-scam research as limited, fragmented, and inconsistent because victim–scammer data are not readily available. Future interdisciplinary research should develop detection profiles from recognisable scammer strategies, jurisdictions, grooming stages, communication patterns, and payment patterns (Sorell & Whitty, 2019; Vedhanayagam et al., 2026).

References

• Abubakari, Y. (2023). The Espouse of Women in the Online Romance Fraud World: Role of Sociocultural Experiences and Digital Technologies. Deviant Behavior, 45(5), 708–735. https://doi.org/10.1080/01639625.2023.2263137 Publication record
• Abubakari, Y., Lazarus, S., & Oseh-Ovarah, V. (2026). A crime script perspective on mapping the entry, continuation, and exit pathways of online romance fraud. International Journal of Comparative and Applied Criminal Justice, 1–32. https://doi.org/10.1080/01924036.2026.2645937 Publication record
• Amirkhani, S., Alizadeh, F., Randall, D., & Stevens, G. (2024). Beyond Dollars: Unveiling the Deeper Layers of Online Romance Scams Introducing “Body Scam”. Extended Abstracts of the CHI Conference on Human Factors in Computing Systems, 1–6. https://doi.org/10.1145/3613905.3651004 Publication record
• Amirkhani, S., Alizadeh, M. F., Randall, D., Stevens, G., & Zytko, D. (2026). My Parents Expectations Were Overwhelming: Online Dating Romance Scams Targeting Minors in Iran Through Exploitation of Parental Pressure. Publication Server of Bonn-Rhein-Sieg University of Applied Sciences (Bonn-Rhein-Sieg University of Applied Sciences). https://doi.org/10.48550/arxiv.2601.16321 Publication record
• Barnor, J. (2024). The charade of discreetness: Exploring the paradoxical lifestyles of romance fraudsters. Journal of Investigative Psychology and Offender Profiling, 21(3). https://doi.org/10.1002/jip.1637 Publication record
• Botha, J. G., Singh, K., & Leenen, L. (2025). Analysis of a Cryptocurrency Investment Scam: Pig Butchering. European Conference on Cyber Warfare and Security, 24(1), 61–70. https://doi.org/10.34190/eccws.24.1.3558 Publication record
• Cantin, P.-M., Ream, F., & Dupont, B. (2025). The Online Mutual Help Practices of Romance Fraud Victims. Les Presses de l’Université d’Ottawa | University of Ottawa Press eBooks, 179–194. https://doi.org/10.2307/jj.37022854.16 Publication record
• Choi, S. W., Lee, J., & Choi, Y. J. (2024). Unveiling the Patterns of Romance Scams in South Korea. International Journal of Cyber Behavior, Psychology and Learning, 14(1), 1–15. https://doi.org/10.4018/ijcbpl.357152 Publication record
• Corpuz, D. R. R. N., Galang, M. J. G., Gueco, K. H. S., & Pamintuan, K. D. (2025). ROMANCE AND RUIN: THE INTERPLAY OF FINANCIAL LOSS AND PSYCHOLOGICAL WELL-BEING AMONG LOVE SCAM VICTIMS IN THE PHILIPPINES. EPRA International Journal of Multidisciplinary Research (IJMR), 675. https://doi.org/10.36713/epra25491 Publication record
• Cross, C. (2022a). Meeting the Challenges of Fraud in a Digital World. The Handbook of Security, 217–238. https://doi.org/10.1007/978-3-030-91735-7_11 Publication record
• Cross, C. (2022b). Using artificial intelligence (AI) and deepfakes to deceive victims: the need to rethink current romance fraud prevention messaging. Crime Prevention and Community Safety, 24(1), 30–41. https://doi.org/10.1057/s41300-021-00134-w Publication record
• Cross, C., & Layt, R. (2021). “I Suspect That the Pictures Are Stolen”: Romance Fraud, Identity Crime, and Responding to Suspicions of Inauthentic Identities. Social Science Computer Review, 40(4), 955–973. https://doi.org/10.1177/0894439321999311 Publication record
• Dickerson, S., Apeh, E., & Ollis, G. (2020). Contextualised Cyber Security Awareness Approach for Online Romance Fraud. 2020 7th International Conference on Behavioural and Social Computing (BESC), 1–6. https://doi.org/10.1109/besc51023.2020.9348307 Publication record
• Dickinson, T., Wang, F., & Maimon, D. (2023). What Money Can Do: Examining the Effects of Rewards on Online Romance Fraudsters’ Deceptive Strategies. Deviant Behavior, 44(9), 1386–1400. https://doi.org/10.1080/01639625.2023.2197547 Publication record
• Dominguez Castillo, L. (2026). Industrialized heartbreak: how generative AI enables romance fraud at scale. AI and Ethics, 6(3). https://doi.org/10.1007/s43681-026-01129-9 Publication record
• Franceschini, I., Li, L., & Bo, M. (2023). Compound Capitalism: A Political Economy of Southeast Asia’s Online Scam Operations. Critical Asian Studies, 55(4), 575–603. https://doi.org/10.1080/14672715.2023.2268104 Publication record
• Grace Carvalho Fernandes, S., BASTOS DO NASCIMENTO, V., & Castelo Branco Do Nascimento, V. (2023). ROMANCE SCAM: A MANIPULAÇÃO DAS VÍTIMAS E O TRÁFICO HUMANO [Romance Scam: Victim Manipulation and Human Trafficking]. Anais do Congresso Internacional de Relações Internacionais do Amazonas: a Amazônia no mundo e o mundo na Amazônia. https://doi.org/10.29327/congressoriamazonas.699587 Publication record
• Griffin, J. M., & Mei, K. (2024). How Do Crypto Flows Finance Slavery? The Economics of Pig Butchering. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.4742235 Publication record
• Gujarathi, P., Verma, S., & Nair, V. V. (2026). Pig Butchering Scams as Cyber-Enabled Financial Crime: A Scoping Review of Dimensions, Modus Operandi, and Victim-Offender Dynamics. Deviant Behavior, 1–16. https://doi.org/10.1080/01639625.2026.2677691 Publication record
• Hu, Y., Wang, H., Zhou, Y., Guo, Y., Li, L., Luo, B., & Xu, F. (2021). Dating with Scambots: Understanding the Ecosystem of Fraudulent Dating Applications. IEEE Transactions on Dependable and Secure Computing, 18(3), 1033–1050. https://doi.org/10.1109/tdsc.2019.2908939 Publication record
• Huang, J., Stringhini, G., & Yong, P. (2015). Quit Playing Games with My Heart: Understanding Online Dating Scams. Lecture Notes in Computer Science, 216–236. https://doi.org/10.1007/978-3-319-20550-2_12 Publication record
• Jakobsson, M. (2016). Understanding Social Engineering Based Scams. https://doi.org/10.1007/978-1-4939-6457-4 Publication record
• Kassem, R., & Carter, E. (2023). Mapping romance fraud research – a systematic review. Journal of Financial Crime, 31(4), 974–992. https://doi.org/10.1108/jfc-06-2023-0160 Publication record
• Khukhunaishvili, D. (2024). Romance Scam as one of the main Challenges of cybercrime. https://doi.org/10.13140/rg.2.2.23659.14880 Publication record
• Ko, S., Kim, J., Jang, J. Y., & Kim, G. (2026). Human‐supervised LLM triage of pig butchering complaints: A validation study of multi‐model identifier extraction. Journal of Forensic Sciences. https://doi.org/10.1111/1556-4029.70411 Publication record
• Kopp, C., Layton, R., Sillitoe, J., & Gondal, I. (2016). The Role of Love stories in Romance Scams: A Qualitative Analysis of Fraudulent Profiles. Zenodo (CERN European Organization for Nuclear Research). https://doi.org/10.5281/zenodo.56227 Publication record
• Luong, H. T., & Ngo, H. M. (2024). Understanding the Nature of the Transnational Scam-Related Fraud: Challenges and Solutions from Vietnam’s Perspective. Laws, 13(6), 70. https://doi.org/10.3390/laws13060070 Publication record
• Mouri, M. (2024). Romance Scam and Legal Interpreting/Translation. IICAH Official Conference Proceedings, 471–475. https://doi.org/10.22492/issn.2432-4604.2024.41 Publication record
• Ogunleye, Y. O., Ojedokun, U. A., & Aderinto, A. A. (2020). Pathways and Motivations for Cyber Fraud Involvement among Female Undergraduates of Selected Universities in South-West Nigeria. Zenodo. https://doi.org/10.5281/zenodo.3702333 Publication record
• Ohu, F., & Jones, L. (2025). An Examination of Digital Validation-Seeking Behaviors in Adolescents as Precursors to Romance Scamming. Scientia Moralitas Conference Proceedings. https://doi.org/10.5281/zenodo.14911844 Publication record
• Ordekian, M., Papasavva, A., Mariconti, E., & Vasek, M. (2024). A Sinister Fattening: Dissecting the Tales of Pig Butchering and Other Cryptocurrency Scams. 2024 APWG Symposium on Electronic Crime Research (eCrime), 136–148. https://doi.org/10.1109/ecrime66200.2024.00016 Publication record
• Parra, Y. J., Chi, H., Alo, R. A., & Lima, V. (2026). Evaluating Jailbreak Vulnerabilities in LLMs: A Taxonomy and Comparative Analysis in Romance Fraud Scenarios. 2026 IEEE 5th International Conference on AI in Cybersecurity (ICAIC), 1–6. https://doi.org/10.1109/icaic67076.2026.11395743 Publication record
• Pope, T., & Seto, C. H. (2026). From swipe to swindle: a narrative review of research on older adult victims of romance fraud. Journal of Elder Abuse & Neglect, 38(4), 423–443. https://doi.org/10.1080/08946566.2026.2690996 Publication record
• Rabby, F., & Chowdory, M. M. U. (2024). Romance Scamming: Uncovering the Transnational Crime and Legal Challenges. International Journal of Law and Societal Studies, 1(1), 48–62. https://doi.org/10.61424/ijlss.v1i1.134 Publication record
• Robinson, J., & Edwards, M. (2024). Fraudsters target the elderly: Behavioural evidence from randomised controlled scam-baiting experiments. Security Journal, 37(4), 1173–1196. https://doi.org/10.1057/s41284-023-00410-4 Publication record
• Sorell, T., & Whitty, M. (2019). Online romance scams and victimhood. Security Journal, 32(3), 342–361. https://doi.org/10.1057/s41284-019-00166-w Publication record
• Stern, M., & Choi, K.-S. (2026). Three-Tier On-Chain Transaction Architecture in a Sanctions-Linked Pig-Butchering Network: A Blockchain-Forensics Case Study. International Journal of Cybersecurity Intelligence & Cybercrime, 9(2). https://doi.org/10.52306/2578-3289.1249 Publication record
• Vedhanayagam, P., Singh, M., Dadhania, A. P., & Ikram, S. T. (2026). Forensic footprints in digital love: Unveiling romance scams with Maltego and machine learning. AIP conference proceedings, 3449, 020278. https://doi.org/10.1063/5.0298575 Publication record
• Wang, F., & Kelsay, J. D. (2025). The prevention of online romance scams using a crime script analysis from the victim’s perspective. International Review of Victimology. https://doi.org/10.1177/02697580251377100 Publication record
• Wang, F., & Topalli, V. (2024). The cyber-industrialization of catfishing and romance fraud. Computers in Human Behavior, 154, 108133. https://doi.org/10.1016/j.chb.2023.108133 Publication record
• Wang, F., & Zhou, X. (2022). Persuasive Schemes for Financial Exploitation in Online Romance Scam: An Anatomy on Sha Zhu Pan (杀猪盘) in China. Victims & Offenders, 18(5), 915–942. https://doi.org/10.1080/15564886.2022.2051109 Publication record
• Yan, Y. (2024). Explore the Causes of Crime and Social Impact of Romance Fraud in Mainland China from 2019–2021. Advances in Social Science, Education and Humanities Research, 601–608. https://doi.org/10.2991/978-2-38476-271-2_73 Publication record

This AI-assisted synthesis is based on reviewed evidence records. It may contain errors or omissions. Follow the publication and DOI links, consult the original works, and make your own judgment about the evidence.