{"id":7111,"date":"2026-11-06T09:13:38","date_gmt":"2026-11-06T09:13:38","guid":{"rendered":"https:\/\/blog.tchop.io\/?p=7111"},"modified":"2026-11-06T11:50:39","modified_gmt":"2026-11-06T11:50:39","slug":"tiktok-knows-when-your-reader-fell-asleep","status":"publish","type":"post","link":"https:\/\/blog.tchop.io\/en\/tiktok-knows-when-your-reader-fell-asleep\/","title":{"rendered":"TikTok knows when your reader fell asleep. You don&#8217;t know their name."},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">A video called &#8220;<a href=\"https:\/\/x.com\/ayoships\/status\/2117650684774580249\">I reverse engineered tiktok<\/a>&#8221; went up on X on Tuesday, and within three days it had been seen more than three million times. The developer behind it says the TikTok app runs more than 25 AI models on your phone, and that all of them work toward a single goal: keeping you from putting the phone down.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">If the claim holds, TikTok knows whether your reader is walking, lying in bed or sitting in a car. It has a guess at when they fell asleep with the phone in their hand, and a prediction of when they will open the app again.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Most newsrooms don&#8217;t know that reader&#8217;s name.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">What the teardown claims<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The developer is Ayo, co-founder of a Y Combinator startup, posting as @ayoships. He showed code from the app and walked through what he says he found:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>A model that reads the phone&#8217;s motion sensor to tell walking from lying down from driving.<\/li>\n\n\n\n<li>A model that guesses you dozed off, based on how long the screen has gone untouched.<\/li>\n\n\n\n<li>A model that predicts when you will come back.<\/li>\n\n\n\n<li>An &#8220;emergency batch&#8221; of videos when the app expects you to skip what&#8217;s next, and a livestream if you skip again.<\/li>\n\n\n\n<li>Battery level, screen brightness and mute status sent along with each batch request.<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">One caveat before anything else. He showed snippets of code, and no one outside his team has checked them yet. Treat the specifics as claims until someone does. We are writing about this because the bigger picture doesn&#8217;t depend on any single detail being right.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">What we already know<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The teardown is new. The pattern behind it is not, and parts of it are on the record.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">In 2020, a privacy warning in an iOS beta caught TikTok reading whatever users had copied to their clipboard. TikTok confirmed it and said it would stop. In 2024, researchers from the University of Wisconsin-Madison showed at the IEEE Symposium on Security and Privacy that the app runs vision models on the phone that estimate a user&#8217;s age and gender from their images and videos. Later that year, a faulty redaction in Kentucky&#8217;s lawsuit against TikTok made internal documents readable. NPR reported one of TikTok&#8217;s own findings from them: 260 videos are enough for a viewer to form a habit. TikTok said the documents were outdated and quoted out of context. And in February 2026, the European Commission found in preliminary findings that TikTok&#8217;s addictive design breaches the Digital Services Act. TikTok called the findings &#8220;categorically false&#8221;.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Put together, the picture is consistent. A platform with this much money and this many engineers can study attention at a level of detail no publisher will ever match.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">The asymmetry publishers live with<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Now think about where your journalism meets your audience.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">For most publishers a big part of it happens inside feeds like this one, where your clip plays next to whatever a sensor model decided suits the reader&#8217;s posture, and your headline has to survive an emergency batch designed to stop them leaving. The platform keeps everything it learns in that moment. What comes back to your newsroom is a view count.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">That is the trade most publishers made over the last decade, often without anyone deciding it. The platform gets the relationship and the data. The publisher gets reach, on terms that change whenever the platform changes them. When search and social traffic drop, as they have for most newsrooms we talk to, there is nothing left underneath.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Seen next to the teardown, the publisher side of the ledger looks thin: an anonymous session ID, a bounce rate, maybe an email address if the paywall caught one. TikTok, if Ayo is right, has a model of the reader&#8217;s whole evening.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Don&#8217;t try to win this race<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The tempting conclusion is that publishers need their own version: more tracking, more models, more signals. That would be a mistake, for two reasons.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">You can&#8217;t win it. A newsroom will not out-engineer a platform that studies attention full time at this scale. Anyone who promises you TikTok-grade behavioural prediction for your app is selling you the losing side of an arms race.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">You don&#8217;t want to win it either. Readers notice when they are being farmed for time, and what&#8217;s left of their trust in news is the one advantage you hold. Regulators have started to notice too.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">What a publisher can have, and a feed can&#8217;t, is a relationship the reader chose. That runs on different data. A name, because they signed up. Topics they follow because they picked them. Comments, questions and replies. Knowing who showed up to last month&#8217;s reader event. All of it is first-party data given with consent, and readers know you have it, because they gave it to you.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">That kind of knowing is smaller than what TikTok has. It is also yours. Nobody can change the terms on you.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">What this looks like in practice<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Publishers who move in this direction tend to make three changes.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">They build a place readers come back to on purpose: their own app and web community, in their own brand, where journalism and conversation sit together. Social stays a channel for discovery. The relationship lives at home.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Their scorecard changes from who saw it to who stayed. Returning members, participation and the share of readers they know by name tell you more about your future than reach does.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">They use AI for the reader, not on the reader. The same technology that predicts when someone will relapse into scrolling can help a small newsroom curate, personalise and moderate for its community. The difference is who controls it and who sees the rules. On tchop, publishers connect their own AI agents to the platform, and those agents curate, tag and personalise under rules the newsroom sets. Editors approve what ships. Readers get a feed that someone is accountable for.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Funke Digital and Ruhr24 already run their reader apps this way on tchop.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">The question to ask your team<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">TikTok will keep learning more about your readers. That part is out of your hands.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">What you control is whether you learn anything about them at all. Ask your team one question this week: of the people who read us every week, how many could we reach tomorrow without a platform in between?<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">If the honest answer is &#8220;very few,&#8221; that is the gap worth closing.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/tchop.io\/solutions\/use-cases\/news-app\" target=\"_blank\" rel=\"noopener\">See how publishers build their own reader community on tchop<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>A TikTok teardown claims 25+ AI models run on your readers&#8217; phones. What that asymmetry means for publishers, and what to build instead.<\/p>\n","protected":false},"author":11,"featured_media":5622,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[174],"tags":[712,450,490,713,711,40,710],"coauthors":[191],"class_list":["post-7111","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news-and-media","tag-audience-data","tag-belonging","tag-first-party-data","tag-owned-community","tag-platform-dependence","tag-publishing","tag-tiktok"],"_links":{"self":[{"href":"https:\/\/blog.tchop.io\/en\/wp-json\/wp\/v2\/posts\/7111","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/blog.tchop.io\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/blog.tchop.io\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/blog.tchop.io\/en\/wp-json\/wp\/v2\/users\/11"}],"replies":[{"embeddable":true,"href":"https:\/\/blog.tchop.io\/en\/wp-json\/wp\/v2\/comments?post=7111"}],"version-history":[{"count":2,"href":"https:\/\/blog.tchop.io\/en\/wp-json\/wp\/v2\/posts\/7111\/revisions"}],"predecessor-version":[{"id":7128,"href":"https:\/\/blog.tchop.io\/en\/wp-json\/wp\/v2\/posts\/7111\/revisions\/7128"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/blog.tchop.io\/en\/wp-json\/wp\/v2\/media\/5622"}],"wp:attachment":[{"href":"https:\/\/blog.tchop.io\/en\/wp-json\/wp\/v2\/media?parent=7111"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/blog.tchop.io\/en\/wp-json\/wp\/v2\/categories?post=7111"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/blog.tchop.io\/en\/wp-json\/wp\/v2\/tags?post=7111"},{"taxonomy":"author","embeddable":true,"href":"https:\/\/blog.tchop.io\/en\/wp-json\/wp\/v2\/coauthors?post=7111"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}