Adverse Media Screening Moves From Search to Risk Intelligence

Financial institutions have accepted adverse media screening as a core element of financial crime compliance, but the next challenge is more exacting: proving that the right entity is being screened, across the right sources, with a decision trail that can withstand scrutiny.

Responding to RegTech Insight, Ripjar CEO Matt Mills said the open questions raised by the report point to a more fundamental challenge than search efficiency. Mills said the problem with manual adverse media checks is that “searching for a name you’ve been given is deceptively risky”. The fundamental challenge, he said, is entity understanding — accounting for name variants, aliases, misspellings and transliterations before risk signals are missed. “Get the entity permutations wrong, and you miss critical risk signals,” he said.

Manual Search Risk

The report’s manual-search finding sits alongside evidence of technology adoption. Among firms conducting adverse media screening, 81% use an integrated screening platform, 74% use an automated adverse media vendor solution and 72% draw on real-time media feeds. Yet 58% still use manual internet searches as part of their adverse media process.

Mills said some firms use manual searches as an informal primary control during pre-onboarding, before a formal record is created. That is high risk because it lacks a diagnostic audit trail and depends on the terms chosen by the individual analyst. Others use manual search as a fallback for perceived blind spots, such as local-language blogs, niche regional media or harder-to-reach content.

For compliance teams, this shifts the issue from source availability to evidential quality. A public-web search may find material, but it does not automatically create a structured, repeatable and auditable screening decision. It also cannot support continuous monitoring, deduplicate noisy hits or surface material that has been removed, suppressed or buried in search rankings.

Integration Test

The report also shows strong demand for unified screening. It found that 96% of respondents regard a platform combining sanctions, politically exposed persons, watchlists and adverse media as critical or very valuable for regulatory compliance.

Mills said best-in-class integration is “more than four tools in one interface”. In his view, the test is whether a single, correctly resolved entity can be evaluated across all risk factors continuously and in context.

That distinction matters. A firm can have adverse media, sanctions, PEP and watchlist data available through one front end while still relying on separate data models, alert structures and workflows underneath. Where that happens, analysts may still need to reconcile whether different alerts refer to the same person or organisation, whether risk signals are connected, and whether the audit trail gives a complete view of the decision.

Mills said many implementations fall short because they fail to integrate at the entity layer. Some organisations operate with siloed databases behind more modern interfaces, producing inconsistent alerts and manual reconciliation. Others rely on AI-driven summaries that do not provide the granular explainability required for audit and decision-making.

“You can have every dataset in the world,” Mills said. “But if your integration layer doesn’t understand entity permutations and resolve the same entity consistently across all risk domains, you won’t know if you’re looking at risk or at a reconciliation gap.”

The persistence of fragmented screening is also a procurement and operating-model problem. Sanctions, PEPs, watchlists and adverse media tools have often been bought at different times, under different contracts and for different regulatory or business drivers. Replacing several functioning tools can feel more disruptive than living with the reconciliation overhead they create.

That explains why unified screening can be widely desired while still difficult to achieve. Consolidation requires more than a technology selection. It requires migration of data, workflows, alert histories, model assumptions, escalation routes and audit evidence.

AI in Workflow

The report also identifies a gap between AI investment intent and production deployment. It says 79% of firms cite AI as an investment driver, while only 28% report full deployment in screening operations. That gap reflects the difficulty of moving AI from experimentation into governed compliance workflows.

Mills argued that the term AI needs more precision in this context. He distinguished between frontier large language models layered over a dataset and purpose-built machine learning, adaptive machine learning and generative AI applied to different stages of the screening process.

Mills said Ripjar uses specialised machine learning for entity extraction and name matching, adaptive machine learning for article classification and relevance, and generative AI for risk summarisation and alert triage. He framed this as a co-pilot model for investigators, intended to surface signal, reduce noise and support next-best action. Mills described the approach as “not replacing judgment. It’s amplifying it.”

Decision Evidence

The next phase of adverse media screening will be judged by the quality of decisions firms can evidence. Regulators, auditors and senior management will want to know how an entity was identified, what sources were screened, how matches were assessed, why alerts were escalated or closed, and how those decisions were recorded.

Ripjar’s report shows that the market has accepted the importance of adverse media screening. Mills’ explanation of the open questions points to the next test for firms and vendors: whether screening processes can resolve entities accurately, monitor risk continuously, integrate signals across domains and generate decision evidence that stands up to review.