Can Google’s AI Stop Scam Texts & Calls? Experts Have Doubts

For every person who openly recounts falling for a scam, countless unreported cases of others being scammed go unnoticed.

In fact, over the past year, 1 in 3 US adults experienced some form of financial fraud or scam, according to a recent Bankrate survey. Of those affected, nearly 2 in 5 reported actual financial loss.

Scam detection efforts have long been deployed to counter the threat, but it’s been a cat-and-mouse game. There have been attempts from spam filters to keyword blockers, but scammers often stay one step ahead, slipping past systems with practiced dexterity. The result has been a persistent feeling that most scam detection tools are either too late or too blunt to slice through the layers scammers wrap around themselves.

But that might be changing with Google’s on-device AI scam detection for smartphones – that is, if it delivers on its promises.

Table of Contents Table of Contents

1. Key Takeaways
2. What Is Google’s AI Scam Detection?
3. Will Google’s AI Tool Be Enough to Stop Scammers on Android?
4. The Bottom Line
5. FAQs
6. References

Show Full Guide

Table of Contents

1. Key Takeaways
2. What Is Google’s AI Scam Detection?
3. Will Google’s AI Tool Be Enough to Stop Scammers on Android?
Show Full Guide
4. The Bottom Line
5. FAQs
6. References

What Is Google’s AI Scam Detection?

In March 2025, Google began testing a new AI-powered Scam Detection feature in its

Messages and Phone apps as part of a broader push to strengthen on-device security and reduce the growing wave of SMS and call-related fraud.

Ahead of Google’s Android 16 launch next week, the company said on May 13th that it has made this feature more intelligent to improve its scam detection capabilities.

Before now, the model was trained to identify mostly package delivery and job scams. However, the Android maker says this has been expanded to enable the model to pick up patterns common in other scams, such as crypto scams, toll road and other billing fee scams, financial impersonation scams, technical support scams, and more.

Similar to SMS scam detection, Google says the feature can also detect and block suspicious prompts that often accompany scam calls, such as encouraging users to disable Play Protect, sideload unsafe apps, or grant sensitive permissions. These protections run entirely on-device and only apply to conversations with unknown contacts.

Just like Google, other companies are beginning to lean on AI to tackle scammers and shield users from fraudulent communications.

For instance, British telecom O2 employs “dAIsy”, an AI bot that engages scammers in prolonged calls to waste their time, alongside “Call Defence,” which uses AI to analyze call patterns in ​real-time​​ to determine if it’s spammy or genuine.

Major US telecoms like AT&T, T-Mobile, and Verizon all deploy AI-driven call filtering to help with scam detection.

Will Google’s AI Tool Be Enough to Stop Scammers on Android?

The promise of AI scam detection is nothing new. In recent years, several phone makers and cybersecurity companies have made similar moves, all with the goal of curbing the steady rise in digital fraud.

But despite these efforts, scam messages and calls still manage to slip through.

So while Google’s new Scam Detection tool feels like a step forward, it also raises the question: Will this finally work where others have stalled?

To understand how much of a solution this is, we spoke to some AI and cybersecurity experts who raised the veil on what to expect from Google’s AI scam detection feature.

Gaurav Tendolkar, Senior Data Scientist at Microsoft, told Techopedia he recently fell for a phone call scam last year and welcomes Google’s recent initiative as “a great step in terms of security.”

However, Tendolker points out that it would be impossible for the tool to avoid raising false negatives or false alarms. He said:

“Even with today’s AI sophistication, it is impossible to get a 100% correct scam detection paired with 0% false alarms.”

Tom Tovar, CEO of mobile security and development platform Appdome, noted that while device detection is “superior to a cloud approach,” it is limited by “isolated data access and lack of behavioral context across users, apps, and environments.”

He told Techopedia:

“Effective fraud prevention should focus on mobile brands adding AI-native security architectures to their mobile apps’ security frameworks – systems built from the ground up to integrate real-time telemetry, behavioral analytics, and adaptive risk models into the mobile application itself.”

For him, anything short of this approach means the feature will be limited to just the user’s screen, rather than seeing the “entire battlefield.”

According to Tovar, this is the same reason other AI-powered scam detection solutions have failed. He said:

“Previous AI-driven scam detection solutions have fallen short because they were treated as bolt-on features – cloud APIs or on-device widgets – rather than foundational layers of a dynamic defense architecture.

“They often lack the telemetry breadth, contextual understanding, or the real-time adaptability required to counter modern fraud patterns like impersonation, social engineering, or polymorphic mobile malware.”

The Bottom Line

Despite growing investment in AI for fraud detection, experts caution we can’t count on it yet. Scammers move quickly, changing tactics, language, and formats to outsmart filters. Therefore, what appears to work today might fail tomorrow.

Co-founder and CEO at Bitmind, Ken Jon Miyachi, points out that AI is inherently probabilistic, meaning it can’t guarantee perfect accuracy.

“AI-driven fraud detection can be evaded in some cases,” he said. “Out-of-distribution data may result in false positives or negatives, frustrating users and missing nuanced scams.”

Miyachi believes AI has real value, especially when paired with regular updates and strong monitoring, but warns that relying on it alone could risk both user trust and effectiveness.

FAQs