A ChatGPT bot that exists just to waste telemarketers' time

Most people deal with scam calls by hanging up, blocking the number, or swearing at the caller. None of that really solves the problem, and you still end up annoyed.

Roger Anderson, a California resident based in Monrovia, took a different route. His company, Jolly Roger Telephone Company, offers a $25-per-year subscription that turns ChatGPT and voice cloning into an active counter-offensive against telemarketers and scammers.

Instead of blocking robocalls, the system aims to keep scammers on the line as long as possible, draining their time and money.

Turning ChatGPT into an anti-scam weapon

Despite FCC crackdowns, Americans still endure an average of 14 spam calls per month per phone number. Most are predatory pitches hunting for credit cards, often targeting the elderly.

Anderson's setup is straightforward. Once subscribed, users get a dedicated forwarding number. You can route unwanted calls directly to the bots, or use your phone's merge feature to join a three-way call and silently listen in on the chaos.

The scammers aren't speaking directly to ChatGPT in real time. Instead, Jolly Roger uses OpenAI's bot to listen, analyze what the caller is saying, and trigger matched canned phrases recorded through cloned voices.

Users can choose from multiple personas. Whitey Whitebeard is an irritable old man who rambles endlessly and gets sidetracked by background noise. Salty Sally plays an exhausted stay-at-home mom constantly yelling at loud kids in the background while asking the caller to repeat everything.

The delivery can occasionally sound repetitive or talk over the caller, but it works surprisingly well. Scammers, blinded by the prospect of closing a deal, regularly stay hooked for up to 15 minutes. In one recording, Whitey kept a fake Dish Network scammer strung along for nearly 10 minutes until the caller had a full, expletive-filled meltdown.

Beating the autodialers at their own game

Scamming the scammers is not entirely new. A bot named Lenny has been doing this since 2008 and remains free to use via call forwarding.

However, modern scammers have evolved. As noted by The Wall Street Journal, automated dialers blast out roughly 100 calls per second, filtering for real people by demanding a key press before routing the call to a human.

Lenny cannot handle keypad prompts, but Jolly Roger's bots can. They recognize automated IVR scripts and actively simulate the most common pass-through keypresses, tricking the autodialer into handing the line over to a live scammer.

Why it matters

We think this is one of the cleverest consumer applications of language models yet, simply because it flips the unit economics of telephone fraud on its head. Robocall operations rely on pure velocity; forcing a live agent to burn 15 minutes on a synthetic grandmother makes the entire enterprise unprofitable.

For everyday users, $25 a year is a small price to pay for turning daily annoyance into entertainment, and it offers practical utility for protecting vulnerable family members.

The trade-off is friction. Setting up call forwarding or manually merging three-way calls still requires active user effort.

It also highlights an inevitable arms race. As telemarketing operations adopt generative voice agents of their own, we are likely heading toward an absurd future where conversational bots spend all day bickering with each other on the phone while telecom carriers collect the tolls.