The hidden role of telephony channels in insurance fraud

Fraud now accounts for almost 44% of all crime reported in England and Wales and is estimated to cost the UK economy £219bn each year, according to the 2026 Cifas Fraudscape report.

To take one insurance sector example, across all Aviva brands including Direct Line, more than 18,400 suspect claims worth £233m were uncovered in 2025 – or £638,000 of fraud detected every day – according to figures from the insurer.

Motor insurance fraud still makes up the majority of fraudulent claims Aviva detects, representing more than seven in 10 cases. However, the nature of scams is evolving – and voice-based interactions are a key part of that evolution.

Fraudsters are increasingly moving away from staged collisions – which are riskier to orchestrate and easier for insurers to detect – and towards exaggerated claims for vehicle damage, repair costs, credit hire and injury, often using wider cost pressures as justification.

We’re seeing fraud become more sophisticated, from exaggerated claims to the use of AI‑generated documents.

Pete Ward, Aviva

Pete Ward, head of claims counter fraud at Aviva, said: “We’re seeing fraud become more sophisticated, from exaggerated claims to the use of AI‑generated documents, and we’re continuing to invest in the tools and expertise needed to identify and stop it.”

A common entry point for fraudsters is the telephony channel, which remains a critical acquisition, servicing and claims route. If insurers do not monitor this channel and treat caller activity as a source of risk intelligence, they could miss behavioural and intent signals that could be gathered long before a loss occurs – signals that can link multiple policy applications account takeovers or early reconnaissance attempts that would otherwise remain invisible.

Call patterns, including the numbers used, can provide early signals of “behavioural indicators, social engineering and reconnaissance that support wider fraud attempts”, said David Pritchard, claims fraud manager at NFU Mutual.

Call centre targets

Fraudsters increasingly use reconnaissance calls to test an insurer’s contact centre, learning how agents authenticate customers, what information they are willing to disclose and how claims are handled before attempting a more sophisticated fraud.

Many of these reconnaissance calls leave low‑level but repeatable traces that are routinely discarded: unstructured agent notes, unlogged caller numbers, brief test calls or failed interactive voice response (IVR) journeys. When incoming calls are not monitored, recorded, transcribed and linked to policy records, those early warning clues do not feed into fraud investigations – and the same actor can reappear across teams without detection.

Crucially, fraudsters often leave behavioural clues that can help security systems and fraud teams detect the scam. “In many cases, value comes from recognising patterns across multiple contacts rather than relying on a single interaction,” said David Phillips, claims validation technical manager at NFU Mutual.

Telephony channels are a rich source of fraud detection intelligence, says Rob Fallows, head of intelligence at DAC Beachcroft Claims. But the content of the calls can be unwieldy to parse and so can be neglected.

“While technology which highlights stress and adverse sentiment in calls is well-established, historically fraud analytics platforms have struggled to tap unstructured ‘qualitative’ sources like calls,” said Fallows.

However, he is excited by the opportunities large language models (LLMs) are creating to exploit intelligence concealed in calls. “LLMs are bringing structure to the unstructured for the first time,” Fallows said.

Converting voice interactions into structured intelligence is not just about speech‑to‑text, however. Treating the phone numbers used when calling and call metadata as persistent identifiers, and linking repeat calls, forwarding/spoofing indicators and IVR paths, lets insurers join the dots between apparently isolated incidents, and surface actionable fraud leads.

In many cases, value comes from recognising patterns across multiple contacts rather than relying on a single interaction.

David Phillips, NFU Mutual

The bigger picture

Matthew Crabtree, head of financial crime intelligence and investigations at Allianz, said the key, both at an individual insurer level and sector-wide, is not viewing telephony-based fraud intelligence in isolation.

“Across the UK we’ve seen a lot of cyber attacks, and therefore there’s a lot more compromised identities out there, which are then being used to facilitate other frauds like insurance fraud,” he pointed out.

“That sort of data available to fraudsters is then opening up opportunities for them within the insurance industry, so things like vishing calls are coming into us as an insurer and to our customers as people are trying to get extra data, which can be used to build up an identity and then be used for other things like ghost broking,” Crabtree said.

‘Ghost brokers’ – posing as middlemen for well-known insurance companies, claiming they can offer legitimate car insurance at a significantly cheaper price – will either forge insurance documents, falsify details to bring the price down, or take out a genuine policy for someone, before cancelling it soon after. The consumer won’t realise unless they get stopped by police or make a claim.

Aviva detected more than 105,000 fraudulent insurance applications in 2025. A growing proportion of this activity is linked to ghost broking, with the number of ghost-brokered policies identified rising by 7% year-on-year.

To combat ghost-broking Aviva has strengthened its defences at the quote and application stage, using data, behavioural insights and targeted interventions to detect suspicious activity earlier and disrupt fraud before policies are issued.

“Some insurers will use voice analytic tools to detect repeated voices where the [applicant’s] identity is different and more commonly to determine if [they are] being truthful,” said Richard Cliffe, counter fraud and recoveries manager at Collinson Insurance.

Organised criminal gangs are now using synthetic AI-generated audio to communicate with insurers, who counter this by using a combination of spectrographic analysis to identify anomalies, he added.

Exploring the metadata

Suspicious behaviour can be detected by insurers but only if they’re willing to invest in adequate tech-driven fraud detection controls.

Richard Cliffe, Collinson Insurance

Calls can also be screened for many non‑biometric signals that indicate organised activity: repeated phone numbers used across unrelated policy applications, abnormal call volumes or rapid repeat inbound attempts, geographical mismatches between claimed addresses and calling locations, and unusual IVR navigation patterns. 

These metadata‑driven signals, for insurers that can capture them, can surface organised crime networks or persistent actors.

According to Cliffe, some fraudsters seek to minimise direct interaction with insurers, as conversations with experienced claims handlers can expose inconsistencies that are harder to conceal in real time.

“Their recollection of events leading up to the claim are sketchy, they become defensive or aggressive quickly, and it can become apparent that they are using a call script which is often a tactic used by organised criminal gangs on multiple claims,” he said.

However, more organised or sophisticated perpetrators will often use the phone strategically – to build credibility, influence how a claim is handled, apply pressure for faster settlement, or probe an insurer’s processes before attempting a larger fraud.

“Suspicious behaviour can be detected by insurers but only if they’re willing to invest in adequate tech-driven fraud detection controls,” Cliffe added.

Ultimately, [alongside capturing your firm’s own data] collaboration is also crucial to bridge intelligence gaps, both across the insurance industry and other regulated sectors, said Fallows, including via voice channels.

A push towards real-time sharing of data and analytics is something the government recognised the importance of in its recent fraud strategy, with the creation and funding of the Online Crime Centre (OCC) and other mechanisms.

“Fraud,” Fallows pointed out, “respects no boundary nor sector – consequently no one organisation has the full picture”.

Sponsored content