Skip to main content
Back to blog
Privacy • October 21, 2026

Privacy-First Reading Tools: Why On-Device Processing Matters

Privacy-First Reading Tools: Why On-Device Processing Matters

Most reading and accessibility tools ask for a lot of trust. They need to see the pages you visit, sometimes the text you select, sometimes your account details. For a tool that's supposed to help you read, that's a strange bargain — you're trading privacy for the ability to read more comfortably.

This post is about why that matters, what "on-device" and "privacy-first" actually mean in practice, and how to tell whether a reading tool is doing the work locally or sending your reading somewhere else.


Why Reading Tools Are Unusually Sensitive

A reading tool sees more of your life than almost any other extension. It sees the articles you read, the documents you open, the emails you skim, the medical pages you look up, the financial sites you browse. If it has AI features, it may also see the specific paragraphs you ask it to summarize or explain.

That's a lot of data about a person. And unlike a search engine — where you knowingly type a query — a reading tool operates passively, in the background, on pages you didn't explicitly hand over.

The risk isn't hypothetical. Extensions have been caught logging browsing history, selling data to advertisers, or being acquired by companies that then changed the privacy policy. The Chrome Web Store's review process catches some of this, but not all, and "established publisher" badges don't guarantee future behavior.

For a tool whose whole pitch is "we help you read," asking it to also phone home with your reading list is a contradiction.


What "On-Device" Actually Means

The term gets used loosely, so here's the practical breakdown:

Fully on-device means the extension does all its work — font rendering, text extraction, TTS, AI inference — inside your browser, using your computer's resources. Nothing leaves the device. The extension has no server to send data to, and even if it wanted to, there's nowhere for the data to go.

On-device with optional cloud features means the core reading tools (fonts, spacing, ruler, Reader Mode) run locally, but some features (AI summaries, cloud sync) send data to a server when you use them. This is the honest middle ground: the reading itself stays private, and you opt into data sharing only for specific features.

Cloud-dependent means the extension sends page content to a server to do even basic work — extracting text, rendering fonts, or running TTS. These tools are convenient but require trusting the operator with your reading history.

The distinction matters most for AI features. Text-to-speech can run entirely on your device using the browser's built-in voices — no network required. AI summarization, by contrast, usually needs a large language model, which either runs locally (slow, limited) or in the cloud (fast, requires sending text). There's no free lunch here, but there's a big difference between "sends text only when you press the button" and "sends text continuously in the background."


Questions to Ask Before Installing a Reading Tool

You don't need to read privacy policies cover to cover. A few targeted questions get you most of the way:

  1. Does it work offline? If the core reading features (fonts, spacing, ruler, Reader Mode) work with no internet connection, they're running on-device. If they stop working offline, they're cloud-dependent.

  2. Where do the fonts come from? Some extensions load fonts from a CDN (Google Fonts, Adobe Fonts). That's not a privacy disaster, but it means the font provider sees which sites you visit. Bundled fonts — included in the extension package — don't leak anything. FocusFlow bundles all eight of its fonts for this reason.

  3. Does TTS use your browser's voices or a cloud service? Browser voices (the ones in Chrome, Firefox, Edge) run locally. Cloud TTS (Google Cloud, Amazon Polly, etc.) sends your text to a server. Both can sound good; only one is private.

  4. For AI features: when does it send data? The good answer is "only when you press the button, and only the text you selected." The bad answer is "continuously" or "we may use your data to improve our services." Look for explicit statements about what gets sent and when.

  5. What does the privacy policy say about data sales? "We don't sell your data" is a start. "We don't collect browsing data" is better. Look for specifics, not just assurances.

  6. Can you use it without an account? Requiring an account isn't automatically bad, but it means the tool can link your reading to your identity. A tool that works fully without sign-in is harder to track.


What FocusFlow Does (and Doesn't Do)

Since this is our tool, here's the honest breakdown:

On-device, no network required:

  • All eight reading fonts (bundled in the extension; no font CDN sees your browsing).
  • Typography controls (size, spacing, line height, weight, alignment).
  • Reading ruler, focus line, color overlays, visual filters (contrast, saturation, big cursor, reduce motion).
  • Bionic Reading (bolds word starts locally).
  • Reader Mode (extracts and reformats article text locally).
  • Text-to-speech using your browser's own voices. Nothing is sent anywhere to be spoken.
  • Per-site settings, profiles, Reading List — all stored in your browser's extension storage.

Cloud, only when you explicitly ask:

  • AI tools (summarize, simplify, define, check my understanding, words to know, key points). These send text to an AI service only when you press the button. Summarize sends the article (or the section you're reading in Reader Mode). Define and Words to know send single words. Check my understanding sends the section you're reading. Your quiz answers are never sent.
  • Pro settings sync across devices (opt-in; requires an account).

What we don't do:

  • No analytics, no telemetry, no usage tracking.
  • No browsing history collection.
  • No ads, no data sales.
  • No account required for the free tier.

The free tier is genuinely free and unlimited — not a trial. Pro ($2.99/month or $29.99/year) adds the AI tools, PDF export, immersion reading, and a few visual extras. The privacy model is the same for both tiers: reading stays local, AI runs only on demand.


Why This Matters More for Accessibility Tools

Accessibility tools are used by people who may have more to lose from data exposure. A reading tool for dyslexia or ADHD can reveal a disability. A tool used to read medical or financial pages reveals health or financial concerns. A tool used to read in a second language reveals immigration status or travel patterns.

The people who most need reading tools are often the people least well served by "we may share your data with partners" privacy policies. That's not a hypothetical ethical concern — it's a practical reason to choose tools that don't collect in the first place.

It's also why we're skeptical of tools that require accounts for basic functionality, or that offer "free" tiers supported by data monetization. The reading tool market has plenty of both.


How to Evaluate a Tool You're Already Using

If you already have a reading extension installed, a quick audit:

  1. Open it offline. Turn off Wi-Fi and try the core features. If fonts, spacing, ruler, and Reader Mode still work, you're in good shape. If everything breaks, the tool is cloud-dependent.
  2. Check the network tab. In Chrome, open DevTools (F12), go to the Network tab, and use the extension on a page. If you see requests to external servers during normal reading (not during an explicit AI action), data is leaving your device.
  3. Read the permissions. In chrome://extensions, click "Details" on your extension and look at the permissions. "Read and change all your data on all websites" is normal for a reading tool (it needs to modify page content). "Access your browsing history" or "Communicate with cooperating websites" is more concerning.
  4. Check the privacy policy. Search for "sell," "share," "third party," and "analytics." The absence of those words is a good sign; their presence with vague qualifiers is not.

The Bottom Line

Reading tools should make reading easier, not create a new kind of exposure. The good news is that on-device processing is now fast enough and capable enough that privacy-first reading tools don't have to compromise on features. The trade-off between "works well" and "keeps your data local" is shrinking, and for core reading features (fonts, spacing, ruler, Reader Mode, TTS), it's essentially gone.

When choosing a reading tool, the question isn't "does it have the most features?" It's "does it do what I need without sending my reading somewhere I didn't intend?" For a lot of readers, those two questions have the same answer now.


Related reading: Privacy-First Accessibility · FocusFlow Dashboard Tour · Core Features