Zhihu++: A Local-First, Ad-Free Android Client for Zhihu

What Zhihu++ Solves and Who It Is For

Zhihu++ targets users who are frustrated with the official Zhihu Android app, which the README describes as over 110MB and full of ads, promoted content, and paid columns. The project offers a third-party client that strips out ads, sponsored soft articles, and sales pitches, and it blocks Zhihu's paid 'Yanxuan' content. It is for people who want control over their feed: the ability to block keywords, users, topics, and even AI-generated content, and to see recommendations not dictated by Zhihu's algorithm. The intended audience is privacy-conscious users and those who dislike algorithmic lock-in, as the README frames the project as a way to reclaim 'the right to choose your own life' from large companies.

The Local Recommendation Engine: How It Works

The core differentiator is a 'local recommendation algorithm' that runs entirely on the device. According to the README, this algorithm is independent of Zhihu's own recommendation system and relies on a crawler to fetch content. Users can customize recommendation weights freely, meaning the feed is generated locally based on user-set parameters rather than Zhihu's server-side ranking. The client supports multiple recommendation modes: Web, Android, local, and hybrid. This is a significant architectural choice: instead of trusting Zhihu's feed, the app fetches raw content and re-ranks it locally. The trade-off is that the quality of recommendations depends on the crawler's ability to retrieve content and the user's willingness to tune weights. The README notes the project is 'not complete yet,' so this engine may be rough around the edges.

Getting It Running: Releases, Sizes, and Requirements

You download Zhihu++ from the GitHub Releases page. There are two Android variants: Lite and Full. The Lite APK is under 4MB, a stark contrast to the official app's 110MB+. The Full version includes an ONNX framework for on-device AI inference, which enables LLM-embedding-based intelligent content filtering. The README warns that the Full version is larger and may have lower performance, and that many AI features are still underdeveloped, such as a local knowledge base that the author admits is unfinished. For desktop, there are experimental JAR files for macOS arm64, Linux x64, and Windows x64, requiring Java 17 or higher. You can also install the nightly build for the latest changes. The project is written in Kotlin and uses Kotlin Multiplatform to share core code between Android and desktop.

Feature Set: Beyond Ad Blocking

Zhihu++ is not just an ad remover. The README lists a wide range of features: login via phone verification, QR code scanning, or manual cookie setup; browsing answers, articles, questions, and ideas (pins); viewing comments and notifications; and even content creation on Android, including writing and publishing answers with Markdown support. Reading features include text-to-speech, watermarked image saving, immersive reading mode, AI summarization, and exporting content as PDF, image, Markdown, or HTML, including entire collections. The blocking system is particularly extensive: keyword blocking with regex, NLP-based blocking using LLM embeddings (Full version only), user and topic blocking, and the ability to export and import blocklists across devices. There is also a feature to mark content as suspected AIGC, with voting by users. This breadth suggests a client that aims to be a full replacement, not just a filter.

Limitations and Wrong Tool Cases

The most obvious limitation is that Zhihu++ is not a finished product. The README states 'this project is not complete enough' and invites pull requests. The AI features in the Full version are experimental, with the author admitting that many are undeveloped. The reliance on a crawler and unofficial API access (via the zse96 v2 signature algorithm) means that Zhihu could change its endpoints or add anti-scraping measures, breaking the client. The desktop version is experimental and requires Java 17, which may be a barrier. For users who want a stable, polished app with guaranteed support, this is the wrong tool. Also, the telemetry, while optional, collects your IP address and a hashed account ID, which some privacy-focused users may find off-putting, even though it can be disabled. The project's AGPL-3.0 license means any derivative work must also be open-sourced under the same license, which could deter commercial use.

Alternatives: Hydrogen and Zhihu--

The README itself lists two alternatives: Hydrogen and Zhihu--. Hydrogen is another Zhihu client that does not require root, but the README does not provide details on its approach. Zhihu-- is described as being in 'very early development' with missing features. There is also an iOS port called Zhihu++ swift, but the README explicitly states it is unrelated to this project. The key difference is that Zhihu++ focuses on local recommendation and AI filtering, while Hydrogen may offer a different philosophy, though the README does not elaborate. If you need a more mature alternative, Hydrogen might be worth trying, but you would need to research it separately. The existence of multiple clients indicates a demand for third-party Zhihu access, but each has its own trade-offs.

Maintenance, License, and Upgrade Cost

The project is actively maintained, with recent releases including a nightly build and version 0.28.3 as of August 2026. The license is AGPL-3.0, which has strong copyleft implications: if you modify the code and distribute it, you must make your changes available under the same license. This is important for anyone considering forking or embedding the code. The upgrade cost is low for users: you download new APKs from the releases page. For developers, the project is based on Kotlin Multiplatform, which means shared code across platforms, but the codebase is likely to change rapidly given the nightly builds. The README does not provide migration guides or upgrade notes, so users must be prepared for potential breaking changes. The telemetry feature is opt-out, so you must remember to disable it in settings if you do not want to send usage data.

Editorial conclusion

Adopt Zhihu++ if you want a lightweight, privacy-conscious Zhihu client with local recommendation and robust filtering, and you are comfortable with AGPL-3.0 and a project that is still under development. Do not adopt it if you need a polished, stable app for production use or if you rely on official API guarantees, since the project depends on scraping and unofficial API access. Before committing, verify that the nightly build works on your device and that the features you need, such as AI filtering in the Full version, are functional. The project is actively maintained with recent releases, but its experimental desktop version requires Java 17 and is not a primary focus.

Official sources

1. Official README
2. Project repository
3. Release notes