Hidden to Whom? Invisible Text in Ordinary HTML Email and the Rules Against It

Abstract Mail filters and AI assistants are being given rules that remove or flag text a reader cannot see. We measure what such rules meet in ordinary mail. From the mail delivered to one personal mailbox in 2005 to 2026 we sample 53,914 messages and determine, per client profile and per surface, which text is shown, using a static analyser and a headless browser that agree on 99.998% of text nodes; definitions and stopping criteria were fixed in advance. Hidden text went from rare to common: at most 0.2% of HTML messages a year until 2013, a third now, rising within the same sender domains. Most of it precedes the first visible text, where the provider’s preview string shows it: hidden in the opened message, visible in the message list. Machine readers differ again: tag-stripping extractors return nearly all hidden text, rendering-aware extractors a fifth to a quarter. Presence-based criteria therefore have a large ordinary-mail denominator. Hiding declarations occur in 76% of recent HTML messages and criteria specified on spam are met by 12%, whereas a deployed rule with a length threshold fires on 2%. Post hoc, a second mailbox and 7,422 web-archived newsletters show the same placement, the newsletters also a rise within one email service provider; in a spam archive sample, hidden text is about as frequent but sits elsewhere in the body.