Fraud Calls and Pay-Per-Call in 2025: Where Do We Go From Here?

Spam and fraud calls are on the rise. According to YouMail, spam call volume to consumers is up roughly 20% year-over-year in 2025. Most of us have felt this firsthand: the unknown caller IDs, the “Hi, is this the homeowner?” scripts, the increasingly realistic AI-generated voices.

But the impact isn’t only on consumers’ patience.

For companies that rely on inbound calls through performance marketing, fraud is becoming harder to ignore. Pay-per-call has long been considered an efficient channel for high-intent customer acquisition, especially in verticals like insurance, financial services, legal, and home services. When a consumer chooses to dial in, they’re often closer to making a decision than someone who clicks an ad or fills out a form.

That high intent is exactly what makes the channel valuable — and exactly what attracts fraud.

This raises a natural question many teams are asking in 2025: how do we eliminate fraudulent calls?

There may not be a simple solution, but the strategies being explored today are shaping the future direction of the channel.

Why Pay-Per-Call Attracts Fraud

Fraud follows incentive and pay-per-call campaigns often carry some of the highest lead payouts in digital marketing. A single qualified call may be worth $15, $50, even $200 depending on the industry.

This creates opportunity for bad actors to manipulate the system.

Some forms this takes:

• Auto-dialed bot calls handed off to live agents to mimic consumer inquiries.
• Call looping, where a call is passed repeatedly to generate multiple billable events.
• Caller ID spoofing to make calls appear local or more trustworthy.
• Human “fraud farms” trained to keep calls active for the minimum duration.

But fraud lives in the gray area of human behavior and system exploitation. Sometimes calls slip through that may look like fraud without any intention of wrongdoing, such as someone calling for the wrong service, a real human caller with zero purchasing intent, or a consumer who misinterprets the ad.

And here’s where things get tricky:

Not every low-quality call is fraud. And not every fraud call is obvious.

The Question Behind the Question

When we ask “How do we eliminate fraudulent calls?”, we’re really asking: what counts as a valid call in the first place?

Is it:

• Time on call?
• Transfer event completed?
• Intent expressed?
• A qualifying question answered correctly?

Two different buyers could classify the same call differently, and this ambiguity is part of why fraud detection remains complex.

Approaches Emerging in 2025

1. AI Call Scoring

AI call scoring models evaluate:

• Speech cadence and tone
• Caller behavior patterns
• Call intent signals
• Device metadata
• Repeated number usage
• Call routing paths

These systems attempt to label calls in real time: high quality, likely low intent, or likely fraud.

This solution scales fast and adapts with more data. However, the model is only as good as the quality and diversity of the training sets. If the training data excludes new fraud tactics, AI won’t recognize them until it has already been exploited.

2. Training AI to Identify “Fraud Signatures”

This approach moves beyond scoring into pattern recognition.

AI listens not for what the caller says, but how they say it:

• Script-reading tone patterns
• Repeated pauses used to avoid disconnection penalties
• Identical phrasing across thousands of calls
• Suspiciously consistent call lengths clustered around payout minimum thresholds

This is promising, but remember that fraud adapts quickly. When AI finds a pattern, fraud operators change the pattern.

3. Network-Level Fraud Blockers

Telecom carriers and platform providers continue implementing controls like:

• STIR/SHAKEN caller ID authentication
• Call reputation scoring
• Spam labeling on mobile devices

These help filter out the worst offenders, but calls can be accidentally classified as spam.

4. Human Quality Assurance

Despite advances in automation, human call monitoring teams are still widely used.

Humans can detect:

• Confusion vs. deception
• True buyer interest vs. filler conversation
• Subtle conversational manipulation to run out the clock

But human review is costly and not infinitely scalable.

Why This Is Hard

Fraud evolves faster than the systems designed to stop it.

Fragmented publisher and network ecosystems, limited data-sharing incentives, variability in how buyers define a “qualified” call, and operational pressure to scale volume quickly all contribute to complications in finding a true solution.

Where the Industry Seems Headed

The emerging trend is not one silver-bullet solution, but layering multiple signals:

• AI scoring for speed
• Behavioral pattern recognition for adaptability
• Network-level authentication for broad filtering
• Human review for nuance

The organizations that will adapt best may be those focused on precision and transparency, not just volume.

Closing Thought

Fraud in pay-per-call isn’t going away in 2025 and it likely won’t ever fully disappear. But the industry is shifting toward something more collaborative: shared data, clearer definitions of call quality, and systems that evolve alongside fraud rather than reacting after the fact.