I recently tried to cancel an order from a popular food delivery app and somehow ended up playing a tiny, miserable escape room inside a chatbot. I couldn’t type my actual issue. I could only tap preset options, none of which matched what I needed. So I kept backing out and trying again, like the right answer was hiding behind one more bad menu choice.

Eventually, I Googled how to talk to a live person like an idiot who couldn’t solve a basic task. The answer, naturally, was to pretend I had a different request. I’m kidding, though. I know customers like myself aren’t exactly idiots. But the experience sure made me feel like one.

The consumer effort tax

That’s the consumer effort tax. It’s the extra cost we pay after the checkout screen is already done with us. It’s paid in your time, your patience, and the quiet shame of doing unpaid support work for a company that can process your payment instantly but somehow can’t build a working button for the thing you need fixed.

I’m not exactly alone in the chatbot swamp. The Guardian recently cited National Consumer Rage Survey data showing that nearly 80% of Americans had a service or product problem in 2025, and about two-thirds of them felt “rage” about it. The same report cited Groundwork Collaborative research estimating that US households lose $165 billion a year to the “annoyance economy.”

Tech companies tend to describe the fix as self-service. That sounds helpful, almost empowering. In practice, it often means the company has moved the work to the customer and dressed it up as convenience.

The chatbot isn’t the worker here

AI support makes the loop easier to scale. A chatbot can respond instantly, apologize politely, and still do nothing useful. A HubSpot and SurveyMonkey research reported that 53% of consumers dislike or hate AI in service interactions, while 82% would still prefer human support if the outcome and time spent were the same.

That tracks with the actual feeling of getting trapped in one of these systems. You’re not being helped so much as filtered. A 2025 paper on customer-service chatbots describes “gatekeeper aversion,” where customers resist chatbot channels because they expect an imperfect first step before possibly reaching a human.

Companies have softer words for this. Deflection. Containment. Automation. Inside a dashboard, those can look like success. From the customer side, it can mean the company successfully prevented you from reaching someone who could fix the issue.

The maze is the product

Bad support starts looking less accidental when you notice how much effort goes into making basic exits harder. The FTC finalized a click-to-cancel rule in 2024 meant to make canceling recurring subscriptions and memberships as easy as signing up. That rule sounds obvious, which is usually how you know the original design was probably rotten.

The same logic applies to support. If a company can measure how many people it deflects, it can measure how much time it burns before anyone actually helps. That should be public, if only so we can see which companies have turned customer service into a waiting room with branding.

The strange thing is, the app doesn’t even need to fight this hard. My attention is already scattered across its buttons, prompts, offers, and nudges. Wasting support time feels almost inefficient. Fix the problem, let me leave, and there will be another convenience waiting to collect me soon enough.