What Skrypte captures on every ad visit.
2 July 2026
There's a difference between data that's collected and data that's used. Right now, several categories of data are collected on every visit that reaches an ad unit. None of it currently decides which ad is served. This is what's actually recorded, and where it comes from.
Page context
Every ad request carries the page URL, the page path, the referrer, the visitor's language and timezone, and basic screen and viewport dimensions. Separately, when a page loads, its title, meta description, and Open Graph type and title are also recorded. Together, this is enough to know what page an ad appeared on and roughly what it's about.
Engagement
As a visitor scrolls and interacts with a page, we track how far they scrolled, in fixed steps of 25, 50, 75 and 100 percent of the page, along with total time on page, time the tab was actually in focus, and how many clicks happened. Core Web Vitals for the page are recorded too. None of these figures currently influence which ad gets shown.
Geography
A visitor's IP address is looked up against a geographic database to get a coarse country, region and city. Today this is used for analytics and fraud review, not for choosing which ad to show a given visitor.
What's stored about the request itself
Being direct about this one: both a hashed and a raw form of the visitor's IP address are stored. The hash is what's used for fast lookups, like the rate limiting described in our bot filtering article. The raw address is kept for fraud investigation and for the geography lookup above. We don't currently store names, emails or other identifying account information alongside this data, but the IP address itself is retained, not just a hash of it.
Where this goes next
None of the data described here changes what a visitor sees today. It's being collected because a future delivery layer will need exactly this kind of information, page context and genuine engagement, to make a relevant match. Collecting it accurately now, on top of traffic we've already filtered for bots, is the groundwork for that.