This article does one thing: clearly explain what happened in the Read-Later market in 2025–2026, why it happened, and what key metrics users should watch. It explains structural industry changes: business models, data security, AI audio, self-hosting renaissance, and real-world challenges for users.
1. Core Events of 2025: Pocket Shutdown & Omnivore Closure Shattered "Free Forever"
Pocket: Transitioning from Independent Service to Data Cleanup
Mozilla's support documentation provided critical timeline milestones: Pocket stopped service on July 8, 2025; followed by an export window; and starting November 12, 2025, exports were disabled and user data entered the permanent deletion queue.
The impact on users wasn't just "switching tools," but rather the exposed migration costs of years of accumulated reading states, tagging systems, notes, and highlights: whether export formats are complete, whether context is preserved, whether searchability can be restored—these determine the real losses.
Omnivore: Hosted Service Stopped After Team Acquired by ElevenLabs
The Omnivore user community and related discussions repeatedly mentioned one fact: users needed to export data before November 15, 2024, after which data would be deleted.
This "rapid shutdown post-acquisition" directly changed user expectations around open source and free hosting: open-source code doesn't equal sustainable service. Ordinary users who don't self-host remain subject to the commercial decisions of hosting providers.
2. Business Model Realignment: Subscriptions Become Mainstream Because Cost Structures Changed
Early Read-Later apps had main costs in client development and minimal sync; now the main costs are: web parsing and cleaning, anti-scraping handling, full-text indexing, cross-platform sync, AI processing, and audio generation (if doing TTS).
Take Readwise Reader as an example—its pricing page clearly states subscription prices: annual billing at $9.99/month, monthly at $12.99/month.
This pricing isn't about "what reading view is worth," but "what continuous service and continuous iteration is worth." After Pocket and Omnivore, many users began accepting a more direct logic: paying buys sustainability and predictable service boundaries, not "free but uncertain."
Meanwhile, another route is also growing: shifting backend costs to users themselves (such as iCloud, local databases, or self-hosting). This isn't emotional anti-subscription sentiment, but an engineering choice to reduce "vendor service termination" risk.
3. Product Evolution: Read-Later Being Absorbed by PKM
In 2025–2026, user needs shifted from "saving links" to "turning content into usable knowledge assets." So common Read-Later feature upgrade directions are:
- Highlights, annotations, notes, and tagging systems more like note-taking software
- Export workflows to Obsidian/Notion more like "knowledge middleware"
- RSS, newsletters, PDFs, video transcripts all treated uniformly as "readable objects"
This shift brings two outcomes:
- Heavy tools become more complex, with rising learning curves.
- Lightweight tools that don't solve "digestion and recycling" become content dumping grounds.
To judge whether a Read-Later has entered the "PKM track," two points suffice:
- Can it save and migrate structured outputs from reading (highlights/notes/tags/metadata)
- Can these outputs be searched, reviewed, and reused in the future (not just remaining in the original text)
4. AI Audio (TTS) Expands "Reading Time," But Also Brings New Lock-in and Cost Issues
TTS extends reading from screens to commute, exercise, and household chore scenarios, expanding usable time, not information volume itself. The problem is: high-quality TTS is expensive, and products easily go to two extremes:
- Good listening experience, but weak in note-taking, highlight export, and secondary processing
- Strong knowledge management, but mediocre audio experience
From a user perspective, whether audio is truly valuable doesn't depend on "can it read aloud," but whether this chain is complete: can the listening process leave searchable assets (highlights, timestamps, voice marks, positions that can jump back to original text), and can these assets be exported.
5. Self-Hosting Renaissance: Motivation Isn't "Geek," But Data Sovereignty and Verifiable Sustainability
The core reason self-hosting regained popularity after Pocket and Omnivore is straightforward:
- Data stays on machines/accounts you control
- Service won't suddenly disappear due to company strategy changes
The costs are equally real: deployment and maintenance barriers, mobile experience, crawling success rates (especially for dynamic sites and anti-scraping environments). So what's more likely to grow in 2026 are compromise solutions: local-first + usable sync + clear data export, rather than pure SaaS or pure self-hosting dominating everything.
6. Challenges for Chinese Users: Not "Interface Translation," But Ecosystem and Underlying Technology
Common pain points for Chinese users concentrate in four areas:
- Content walls: especially closed platform article crawling and image display issues
- Chinese search: word segmentation, indexing, hit rate, and performance
- Fonts and typesetting: rendering consistency across different devices, especially e-ink screens
- Chinese TTS: polyphonic characters and sentence breaks increase comprehension costs
This determines a practical conclusion: Read-Later tools targeting Chinese users must treat "crawling + local storage + full-text search" as first-class capabilities or struggle to retain users long-term.
7. InfoFlow Chose a More "Local-First" Route
After events like Pocket and Omnivore, what users worry about most isn't features, but two things:
- Can data be retained
- Can it be migrated in the future
So InfoFlow's design orientation is: try to place key data in user-controlled storage, work offline, save web content locally to reduce losses from broken links, and provide fast full-text and in-page search.
Don't treat InfoFlow as "a bookmark box," but as the entry layer of a personal knowledge processing pipeline: first reliably save content, then let users turn it into reusable information through search and notes.
8. 2026 Read-Later Checklist (More Effective Than "Feature Comparison Tables")
Just check these 6 items to avoid most pitfalls:
- Export & backup: Can you export structured data like Markdown/HTML/JSON, including highlights, notes, tags, and metadata
- Local capabilities: Is it fully usable offline (reading, search, notes)
- Full-text search quality: Does Chinese reliably hit results, is speed acceptable
- Original content fidelity: Does it save webpage content copies, reducing unreadability from 404/redesigns
- Migration cost: Does it provide import from Pocket/Omnivore etc., does import preserve structure (not just links)
- Business sustainability: Can the fee structure cover long-term maintenance (subscription/one-time/self-hosted backend), are boundaries clear
Conclusion
The main thread of the 2025–2026 Read-Later market isn't "who has more features," but three words: sustainable, portable, searchable. Pocket's shutdown and Omnivore's closure only exposed the problem early: what users truly need is a system that turns content into long-term assets, not a temporary favorites folder.