# ElevenLabs' audiobook chat feature lifts listening time 24% By Steven Van · 2026-08-07 Voice Chat in ElevenReader, built on ElevenLabs' ElevenAgents SDK, has drawn over 50,000 users across a dozen languages. [ElevenLabs]() is now running its own agent platform, ElevenAgents, across its business, and the company just published the first hard numbers from that rollout. The showcase is ElevenReader Voice Chat, a feature that lets audiobook listeners pause and talk to the same voice that is narrating their book. Users who adopted it listen 24% longer on average, and readers with five or more Voice Chat sessions finish their books at a 78% completion rate. For anyone building voice products, or just watching where interactive audio is headed, this is one of the clearest case studies yet of an AI agent measurably changing how people consume content. [Source post](https://x.com/i/status/2085735303352828140) ## Ask the narrator a question, then drop back into the story Voice Chat lives inside the ElevenReader app. Mid-book, a listener can ask about a character they have lost track of, a plot point they missed, or a theme they want to dig into, and the agent answers in the same voice that has been reading to them. When the conversation ends, playback resumes where it left off. There is also a text mode, with what ElevenLabs describes as seamless switching between voice and text, so you can type a question on the train without talking to your phone. Since launch, more than 50,000 unique users across a dozen languages have used the feature. The query data says a lot about why people reach for it: 40% of conversations are plot summaries, essentially "catch me up," and another 22% are character and theme analysis. That maps to the real problem with long audiobooks. People drift, lose the thread, and quit. Giving them a zero-friction way to re-orient appears to keep them in the book. ## The retention math: 24% more listening, 78% completion The 24% listening lift is not a loose average across the user base. ElevenLabs compared each user's listening time in the month before their first Voice Chat conversation against the month after it, which controls for the fact that heavy listeners were always going to listen more. The 78% completion rate for users with five or more sessions is the number that should catch publishers' attention, since audiobook completion is a chronic industry weak spot. Correlation caveats apply, engaged readers ask more questions, but the pattern held strongly enough that ElevenLabs is using it as the flagship proof point for its agent platform. ## How the agent avoids spoilers The obvious failure mode for a "talk to your book" feature is the agent cheerfully revealing the ending. ElevenReader handles this by feeding the agent variables tied to the reader's exact position: the book title, description, and author, a summary of the full book, a summary of the current chapter, and the current paragraph. The system prompt frames the agent as a reading companion that helps you understand and enjoy the content without spoiling anything unless you explicitly ask. Before launch, the team stress-tested this by actively trying to break their own guardrails inside the app, then kept reviewing real conversations after release to tighten the protections. ## Built on the same SDKs ElevenLabs sells to everyone else Voice Chat runs on the ElevenAgents native mobile SDKs for iOS and Android, integrated directly into the ElevenReader app. That is the strategic point of the announcement: ElevenLabs is dogfooding the exact agent stack it offers customers, which spans voice and chat agents in 70+ languages, deployment across phone, web, WhatsApp, SMS, and email, and built-in testing and guardrail tooling. The company is signaling that the platform is mature enough to carry a consumer product with real engagement metrics attached. ## What creators should take from this If you publish audio content of any kind, serialized fiction, courses, long-form podcasts, the ElevenReader numbers sketch a playbook. Interactive Q&A layered onto passive listening is not a gimmick; it is a retention feature. The pieces that made it work are all replicable: position-aware context so answers stay relevant, a consistent voice so the experience feels continuous, and guardrails tested adversarially before launch. With ElevenAgents exposed through public SDKs, that playbook is no longer exclusive to ElevenLabs' own apps. ![ElevenLabs]() ElevenLabs The most realistic AI audio platform. [View ElevenLabs →]() Sources [ElevenLabs blog: How ElevenReader used ElevenAgents](), [ElevenAgents platform](), [ElevenLabs announcement on X]() [Read ElevenLabs' audiobook chat feature lifts listening time 24% on Creators Toolbox]() --- Canonical source: https://creatorstoolbox.com/blog/elevenreader-voice-chat-elevenagents