Trackless Links: Building a Safari Extension for Automatic URL Cleaning

#privacy #safari #showdev #ios

Every link you click carries invisible baggage. Those cryptic strings at the end of URLs -- utm_source, fbclid, gclid, twclid -- are tracking parameters. They tell advertisers where you came from, what you clicked, and how to follow you across the web.

Most people never notice them. But once you do, they are everywhere.

The Problem

A typical tracked URL looks like this:

https://example.com/article?utm_source=twitter&utm_medium=social&utm_campaign=spring2026&fbclid=IwAR3x...

Strip the tracking, and all you need is:

https://example.com/article

Those extra parameters tell the destination site exactly which campaign brought you, which platform you came from, and sometimes which specific ad you clicked. When you share a tracked link with someone else, you pass along that tracking context and potentially link their activity to yours.

How Trackless Links Works

Trackless Links is a Safari extension for iOS and macOS that intercepts URL loads and strips tracking parameters before the request reaches the server.

The extension uses Safari's declarativeNetRequest API (content blockers) combined with a curated ruleset that matches known tracking parameter patterns:

UTM parameters (utm_source, utm_medium, utm_campaign, utm_content, utm_term)
Facebook (fbclid, fb_action_ids, fb_action_types)
Google (gclid, gclsrc, dclid)
Twitter/X (twclid)
TikTok (ttclid)
And dozens more from various ad networks and analytics platforms

The key architectural decision: everything runs on-device. The extension uses Safari's native content blocking API, which means URL matching happens in Safari's process without any network calls or external dependencies.

Bonus: Domain Credibility with CRED-1

Trackless Links also ships with the CRED-1 dataset -- 2,672 domains with five independent trust signals aggregated from established credibility databases. When you visit a site with known credibility issues, the extension can flag it.

CRED-1 grew out of my doctoral research on AI-driven disinformation at Frankfurt University of Applied Sciences. The dataset is open source (CC BY 4.0) and documented in a peer-reviewed paper accepted at ACM WebConf 2026.