BrowserTools

Site updates

7 entries · 23 recorded changes · most recent 2026-10-01

BrowserTools is maintained by one person, and this page is the record of it. Every entry below describes work that has actually shipped — not plans. Where something was wrong, the entry says what it was and what fixed it, because a changelog that only lists improvements is not a changelog.

If you think a tool is broken, or a page is wrong, or something is missing, tell me. Corrections get made and get an entry here.

  1. Corrected false privacy claims, removed duplicated copy, fixed a real accessibility bug

    • The privacy policy said no analytics were collected and that Cloudflare Web Analytics was switched off, while Cloudflare was in fact injecting its beacon into every page and recording page views. Verified in a real browser, corrected the policy and the cookie page, and rewrote the two tests that had encoded the incorrect conclusion.
    • Speech to Text was missing from the policy's list of tools that contact a service, even though it streams microphone audio to a cloud recogniser in Chrome and Edge. It is now listed as the eighth.
    • Every page was reachable at two URLs because the trailing-slash version was served instead of redirected. 218 permanent redirects added, generated from the route registry and guarded by a test.
    • Two guides were ranking for their tools' queries rather than their own. Six guide introductions re-aimed at explanatory searches, each now linking to the tool it explains.
    • Tool pages repeated 'free, no signup, no upload, in your browser' four times. The repeated lines were replaced with per-tool detail.
    • Opening the first FAQ on each tool page moved the rendered word count of every page above 740; 56 pages had been under 500 once collapsed answers were excluded.
    • The header search button had an accessible name that did not contain its visible text, which fails WCAG 2.5.3 and breaks voice control. Fixed.
  2. Removed 256 KB of unused advertising script and all speculative prefetching

    • Google's ad script was loading on every page and contacting four Google hosts, while no ad slot existed to fill. It now loads only when a slot is configured, which removed roughly 256 KB per page view — about 39% of total page weight — and eliminated the ad cookies it was setting on pages that showed no ads.
    • Every page was eagerly downloading the full 186-tool catalogue because a bookmark bar that usually renders nothing imported it statically. It is now loaded on demand.
    • RSC prefetch requests on a tool page went from 7 to 0; homepage prefetch traffic fell by about 40%.
    • A test that failed only on Windows was traced to line-ending differences in the working tree and fixed at both the repository and the test level.
  3. Removed structured data Google no longer supports

    • FAQPage and HowTo markup were being emitted on every tool page and guide. Google discontinued the FAQ rich result in May 2026 and retired HowTo in 2023, so all of it was dead weight in the page head. Removed, and a test now fails if either reappears.
    • Collapsed FAQ answers were not present in the served HTML at all, because the accordion unmounted them. Replaced with native disclosure elements so the answers ship in the page.
    • Sitemap dates were asserted at build time to prevent a page advertising a modification date in the future or older than the change it describes.
  4. Every tool page given individually written content

    • The remaining templated tool pages were rewritten individually. All 186 tools now have their own overview, step-by-step instructions, capabilities and FAQs; the fallback generator that produced them is no longer used by any page, and a test enforces that.
    • Twelve guides were written covering Base64, URL encoding, NIST password guidance, JSON pitfalls, PDF structure, BMI limitations, loan maths, unit conversion, randomness and lorem ipsum.
    • An About page, a Contact page, and full Privacy, Terms, Cookie and Disclaimer pages were written rather than stubbed.
  5. Category pages rebuilt as real hubs

    • The eleven category pages were rebuilt as curated directories with category-specific editorial copy describing what each group of tools actually does and which ones need a network connection.
    • Guides became reachable from the header and from the tool pages that they explain, so no page is orphaned.
  6. Every tool audited for what it actually sends over the network

    • All tools were reviewed to confirm which ones genuinely require a network connection. Eight do; the rest run entirely on the device. Those eight are now named in the privacy policy, and each tool's own page repeats what it sends.
    • The privacy policy was rewritten to match what the code does rather than what the code was assumed to do.
  7. First public release

    • BrowserTools went live with 44 browser-based tools: image converters, a PDF toolkit, text utilities, encoders, calculators and random generators.
    • Everything runs client-side. No upload endpoint, no account, no usage quota.

How this site is built

Every tool runs in your browser. For most of them that means your file is processed on your own device and never uploaded anywhere, which is why there is no account, no queue and no usage limit. The privacy policy names the eight tools that do contact a service, and about explains the architecture. How the site is funded is on the privacy page too — currently it is not serving any ads at all.