What we're seeing in California robocall cases (2026)

Two patterns stand out in the California robocall cases we've been screening this year. They're worth flagging, even at the edges of what's defensible to publish before we have a full year's data.

Volume is up, but the mix changed

Do-Not-Call complaint data shows roughly a 12% year-over-year increase in unwanted calls to California area codes. That tracks what we're hearing in intakes.

What's changed is the type of caller. A few years ago, the most common kind of robocall case we screened involved solar leads and home-improvement pitches: the kind of high-volume B2C pipelines that exploded after California's residential solar push. That's quieted. The current high-volume offenders we see are:

• Crypto and DOGE-token promotions, often with offshore call centers
• Auto-warranty extension scams (perennial)
• "Lower your interest rate" prerecorded campaigns targeting older numbers
• Debt-collection wrong-party calls (these have always been a steady flow)

The shift matters strategically. Crypto-promotion defendants are often shell entities. Recovery requires more upfront investigation to identify the principal, which changes the case economics.

After Facebook v. Duguid, more cases lead with the recorded-voice angle

The 2021 Supreme Court decision in Facebook v. Duguid narrowed what counts as an "automatic telephone dialing system" (an "ATDS") under the TCPA. After that ruling, the ATDS angle got harder. The dialing system has to actually generate numbers randomly or sequentially, not just call from a stored list.

The part of the TCPA that bans recorded-voice calls without permission wasn't touched. So practically, most live cases now lead with the recorded-voice angle. Two implications:

First, the question we ask in screening, "did the calls use a recorded voice or appear automated?", is more important than it used to be. A live human caller can still violate the Do-Not-Call rules, but the case is structurally weaker.

Second, defendants have been more aggressive about contesting "prerecorded voice" (meaning a recording made in advance), arguing the call used a voice synthesized from text rather than pre-recorded audio. Courts have generally not accepted that framing, but it's a fault line worth knowing about.

What this means for our intake

We're being a little more selective about Do-Not-Call cases without any sign of automation. They're still worth bringing, but the hours of work per dollar recovered are less favorable than they were three years ago.

For prospective clients, the practical implication: if you're not sure whether the calls were automated, that's a "yes" worth investigating. The tells aren't always obvious. Modern AI-driven calls can sound nearly human for a few seconds before falling out of script.

If you'd like a quick check, CallCheck is the fastest path. We pull these patterns through the screener regularly so its outputs reflect what we're actually seeing.

This is a placeholder data piece. We'll publish real numbers from our case intake at the end of Q2.