The Machines Are Calling: Measuring Automated and Synthetic Voices in Unwanted Inbound Calls

Recent research addresses the previously undocumented prevalence of automated and synthetic voices in unwanted inbound calls, a gap identified despite the U.S. Federal Communications Commission's (FCC) February 2024 ruling classifying AI-generated voices under the Telephone Consumer Protection Act (TCPA). Using an interactive voice honeypot over 66 days, researchers recorded 10,987 calls and employed audio fingerprinting and commercial synthetic-speech detection to quantify machine-generated and synthesized speech.

Why it matters

The proliferation of AI-generated and synthetic voices in unsolicited communications poses significant challenges for regulatory compliance, consumer protection, and operational efficiency. Quantifying the scale of this phenomenon is crucial for developing effective countermeasures and informing policy decisions regarding automated communication systems.

Key insights

• The U.S. FCC classified AI-generated voices under the TCPA in February 2024.
• There was no prior peer-reviewed measurement of the volume of unwanted call traffic placed by machines or the proportion of synthetic speech within it.
• A research pipeline utilizing an interactive voice honeypot on real U.S. numbers recorded 10,987 calls over 66 days.
• The methodology involved an audio fingerprinting system and a commercial synthetic-speech detector to analyze call openings.
• The source text is a summary of research, not the full findings, so specific measurements are not available beyond the methodology.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2609.11137