How AI-generated voice fraud is quietly destroying one of your most trusted customer touchpoints and why two words are the industry’s entire response to it
Read these two words: Potential spam.
That badge — the small label that flashes on your subscriber’s screen before an unknown call rings through — is the telecoms industry’s current answer to one of the fastest-growing threats to consumer safety in Europe. Thirteen spam calls per subscriber per month in Spain and France. A 15% fraud call rate in the UK — the highest in Europe — meaning more than half of all unwanted calls are not merely annoying, they are actively malicious. AI-generated voices cloned from as little as three seconds of audio, impersonating energy providers, family members, government agencies. GenAI-enabled scams up 456% between May 2024 and April 2025. Deepfake fraud losses projected to hit $40 billion by 2027.
And the industry’s response is a label. The phone still rings. The customer still has to decide. And for millions of people — elderly subscribers who answer everything, professionals who can’t afford to miss a call, parents who have no way of knowing whether that voice on the other end is real — the label provides no meaningful protection at all.
For CX leaders, this framing error has a cost. Not just to subscribers who get defrauded — though that cost is real, and in the UK alone average losses per victim ran to £13,342 in 2024. The cost lands in your care queues, in your NPS scores, and in the slow, silent erosion of the one channel that still resolves complex customer problems faster than any other.
The phone call is broken. And “potential spam” isn’t fixing it.
A threat that has outpaced the industry’s response
The scale of what’s happening is not in dispute. France and Spain have held the worst spam call rates in Europe for seven consecutive quarters, with both markets running a 43% spam rate on unknown calls and residents receiving an average of 12 to 13 unwanted calls every month, according to Hiya’s H1 2024 Global Call Threat Report. More than one-third of consumers across the US, UK, Canada, Germany, France, and Spain have now encountered a deepfake fraud call. A 2024 McAfee study found that 1 in 4 adults have experienced an AI voice scam, with 1 in 10 personally targeted.
The composition of the threat is also shifting. A decade ago, spam calls were primarily robocalls and telemarketing — annoying, predictable, and reasonably well-served by a label. That era is over. Today’s attacks involve real-time conversation. Contextual information scraped from public sources to build credibility. Voices indistinguishable from a family member, a GP’s receptionist, a bank fraud team. Advancing technologies mean that fraudsters have been able to broaden their geographical and technological reach, all at a lower cost, as Tracey Wright, Chair of Comms Council UK, has noted. What was once a volume problem has become a sophistication problem — and a label does not address sophistication.
This is a CX problem. Not just a security problem.
Here is where I want to challenge a persistent framing error.
Voice fraud is typically categorised as a security issue — owned by fraud teams, escalated to regulators, discussed in risk forums. Customer experience leaders hear about it when a vulnerable subscriber has been victimised and the complaint lands in the care queue. By then, the damage is done. But the erosion of trust in the phone call is a customer experience problem from the very first ring — and it has downstream consequences that should concern every CX leader deeply.
Because here is what actually happens when a customer stops answering unknown calls.
74% of consumers say they do not answer calls from unknown numbers out of fear they might be scams. 92% assume an unidentified call is fraudulent before they even engage. And the consequences of that avoidance are not abstract. When consumers stop answering, hospitals struggle to confirm surgeries or coordinate urgent follow-up care. Financial institutions face delays delivering legitimate fraud alerts that require immediate response. Your contact centre agent calling to resolve an open complaint goes to voicemail. The appointment reminder from a specialist practice goes unanswered. Seventy percent of consumers have missed a call due to safety or fraud concerns, and afterwards learned it was a legitimate number — TransUnion, 2024. Nearly half of all unidentified calls now go unanswered entirely.
None of this shows up in your fraud reporting. All of it shows up in your CX metrics.
Yet nearly 80% of those same consumers say phone calls are important for communicating with businesses — especially for complex, sensitive, and urgent matters. The channel is still valued. It is simply no longer trusted. That gap between value and trust is where your CX problem lives.
The false dichotomy between customer experience and fraud protection has long plagued the industry. CX teams are rooted in the philosophy of providing effortless experiences, while security teams create intentional friction to deter bad actors — pitting the two groups against one another under the false assumption that you must choose one or the other. That dichotomy applies equally here. The assumption has been that protecting customers from spam means adding friction — warnings, labels, confirmation steps. What that assumption misses is that the friction of not answering at all is far more damaging to the customer relationship than any scam call that gets through.
The deeper damage: who is actually being hurt
The subscribers most affected by this erosion of trust are not distributed evenly across your base. They are concentrated in the segments you can least afford to lose.
More than one-third of UK consumers encountered deepfake voice fraud attempts in 2024. Elderly customers are the most exposed — most likely to have been targeted, most likely to have lost money, and most likely to have overcorrected into blanket call avoidance. Their families tell them not to answer. So they don’t — including when their GP calls, when their pharmacist calls, when your contact centre agent calls to resolve their complaint. Older Americans alone reported $3.4 billion in fraud losses in 2023, an 11% rise from the year before. The pattern is the same across Europe.
These are not marginal customers. They are, in many cases, your most loyal, longest-tenured, highest-spending subscribers. And they are being systematically pushed away from your most effective service channel by a threat your current strategy has not adequately addressed.
The professionals on your network who can’t afford to ignore calls have developed their own adaptation — answering reluctantly, warily, with one hand already hovering over the end-call button. That posture doesn’t generate satisfaction. It generates anxiety. And anxiety is not a foundation for loyalty.
The parents who have coached their teenagers not to answer unknown numbers — a reasonable instruction in 2025 — have now created a household where your legitimate outbound communications, your appointment reminders, your service notifications are structurally unreachable. Not blocked. Not declined. Just never answered.
Why the label was never enough
The spam label was a reasonable tool for a problem that was once primarily about volume. Flag the number. Let the subscriber decide. If they’re alert, they’ll decline. The model made sense when the threat was robocalls selling solar panels.
It does not make sense when the threat is an AI-generated voice that knows your subscriber’s name, their energy provider, and the name of their bank — and is conducting a real-time, adaptive conversation designed to extract money or personal information.
Only one of four free deepfake detection tools was able to correctly flag AI-generated audio in independent testing. If the technology built to detect this threat fails, a label on a call screen stands no chance. And yet only about one-third of consumers have downloaded phone fraud prevention apps, and 59% say they would be unwilling to pay extra for protection. The subscriber-action model — label the call, let the customer decide, offer an app if they want more — is failing at precisely the point where the threat has become most acute.
The European Parliament’s own research body has concluded that the path forward requires solutions that block illegal or unwanted calls by default, based on call analytics carried out before calls reach consumers. Not labels. Not apps. Intervention before the phone rings.
What a genuinely customer-centric response looks like
If the phone call is a broken channel, the customer-centric question is not how to warn people more effectively. It’s how to fix it.
A genuinely customer-centric response operates on the same principle that defines great CX in every other context: it removes the burden from the customer entirely. It intercepts the problem before it reaches them. It acts on their behalf and delivers the outcome they would have wanted — without requiring them to evaluate, decide, or worry.
In practical terms, this means moving protection to the network layer. Rather than labelling a call and letting it ring through, a network-level intervention answers the call on behalf of the subscriber before it ever reaches the handset. It engages the caller in natural conversation. It determines intent in real time. It routes accordingly — connecting legitimate callers immediately, blocking confirmed threats, and delivering a transcript of ambiguous calls for the subscriber to review at their own pace.
The subscriber does nothing. The doctor’s call gets through. The scammer does not.
This is not theoretical. The technology to do this at carrier grade — multilingual, GDPR-compliant, operating at the IMS/SIP layer, fail-open by design — exists and is in active deployment conversations across European markets. The commercial model is straightforward: an operator that moves protection to the network level can offer genuine, felt protection to every subscriber without a single one of them needing to take any action.
The strategic opportunity CX leaders shouldn’t miss
Every telecom operator in Europe currently offers some version of spam labelling. France and Spain have held the worst spam rates in Europe for seven consecutive quarters — and no operator in either market has yet given subscribers genuine, automated, proactive protection. That gap is a significant competitive opportunity. Not a technical feature story. A customer experience story.
The operator that solves this doesn’t just reduce fraud complaints. It restores the channel. 77% of consumers say they are more likely to answer a call if they know who is calling or feel protected. Every one of those answered calls is a resolved complaint, a confirmed appointment, a retained customer. The loyalty uplift is measurable, and it is concentrated in exactly the segments — elderly, high-usage, high-tenure — where churn is most costly and trust is most hard-won.
This is NPS-moving. It is churn-reducing. It is the kind of differentiation that competitors cannot replicate quickly, because it requires the network-level capability and the organisational will to move protection upstream rather than leaving it with the subscriber.
The phone call doesn’t have to be broken. But fixing it requires CX leaders to claim it as their problem — not leave it to the security team, not address it with a label, and not wait for a regulatory mandate that arrives after the damage is already done.
Your customers are already experiencing it. They’ve already learned not to answer. The question is whether your organisation is going to respond at the pace of the threat — or at the pace of the industry.