Been following zero-click search in our industry for months. A WSJ piece confirms what scholarly publishers know. 40-60% of cited domains in AI responses change monthly. Structured metadata and provenance matter more than SEO tricks. www.wsj.com/tech/ai/goog... #ASEO #ContentDiscovery

Radiologists given a mix of real and AI-generated X-rays averaged 75% accuracy. Experience made no difference. LLMs scored 57-85%. We talk about AI-generated text but image integrity needs to catch up. www.nature.com/articles/d41... #ResearchIntegrity #ImageIntegrity

ICML embedded hidden watermarks in papers to catch AI-generated reviews. 506 reviewers flagged, 497 papers rejected under their reciprocal review policy. Only catches careless use, but this type of AI policy is refreshing. www.nature.com/articles/d41... #PeerReview #ResearchIntegrity

Only 42% of US R1 universities link authorship credit to accountability. A PNAS working group says fixing this needs institutional reform and infrastructure that works across naming conventions. Journals can't solve this alone. www.pnas.org/doi/10.1073/... #ResearchIntegrity #Authorship

India's ONOS scheme crossed 113 million downloads in year one. Elsevier took 37% of all usage. ACS outperformed Wiley despite far fewer journals. Turns out content matters more than catalog breadth. www.bignewsnetwork.com/news/2789254... #OpenAccess #ONOS

Microsoft's AI CEO wrote in Nature that AI mimicking consciousness is a deliberate design choice. I work with AI tools daily and catch myself responding as if someone's there. Knowing it's engineered doesn't switch that off. www.nature.com/articles/d41... #ResponsibleAI #AIEthics

Scientists using AI tools focus on a narrower range of research fields. People's opinions shifted toward LLM outputs even when warned about bias. They didn't notice. Stylistic diversity in post-ChatGPT text is measurably down. www.nature.com/articles/d41... #AIinResearch #ResponsibleAI

STM Association published a responsible AI framework for research content. Covers attribution, Version of Record, retractions, bias. Now proposing technical pilots with publishers and AI providers. Overdue but welcome. scholarlykitchen.sspnet.org/2026/03/19/s... #ResponsibleAI #ScholarlyPublishing

Hallucinated citations passing peer review at top AI conferences. An open-source tool detects them at 97.2% accuracy. GPT-5.2 flags 48% of real citations as fake. Open-source is winning this one. the-decoder.com/hallucinated... #ResearchIntegrity #AIinResearch

I use Claude Code daily. Most of my time now goes into reviewing its output and steering direction, not doing the work myself. Ethan Mollick argues we've crossed from working with AI to managing agents. I agree. www.oneusefulthing.org/p/the-shape-... #AIinResearch #FutureOfWork

SocArXiv paused AI submissions for three months, then released a formal policy. Their approach: human judgment only, no algorithmic detection, no AI-to-detect-AI arms race. That is community governance at work. socopen.org/2026/03/09/s... #Preprints #AIinResearch

Check this essay by an academic editor in India who sent 97 reviewer invitations per manuscript and got 3 acceptances. Voluntary, unpaid work. He burned out and resigned from both boards. Surely AI can help! ese.arphahub.com/article/1820... #PeerReview #ResearchIntegrity

A Harvard psychologist's anti-publisher essay got 700 likes this week. James Butcher notes the journals market generates $7.7 billion in costs. The 700 likes are the signal publishers should be paying attention to. www.experimental-history.com/p/the-one-sc... #ScholarlyPublishing #OpenAccess

A 2025 survey found half of researchers use AI for peer review. Editors say it could help an overwhelmed system. The same tools enable fake reviews and confidentiality breaches. Detection and enforcement have not caught up. undark.org/2026/02/25/a... #PeerReview #AIinResearch

Ashutosh Ghildiyal sees a legitimacy crisis beyond pricing. James Butcher's Gist documents the same structural mistrust from the researcher side. Worth reading both. scholarlykitchen.sspnet.org/2026/03/03/g... #ScholarlyPublishing #AcademicPublishing

Ten publishers now fund the Kudos zero-click study. The sponsors put attribution, research integrity, and trust in the scholarly record on the agenda. Cross-sector alignment at this scale on one study is unusual. blog.growkudos.com/news/elsevie... #AIinResearch #ScholarlyPublishing

Pop-up journals that gather evidence on a policy question, deliver the answer, and close. Sounds great. But we live in a world where research is curated around bibliometrics incentives. I would love to see how this plays out. www.nature.com/articles/d41... #SciencePolicy #ScholarlyPublishing

CCC launched internal-only AI re-use rights for higher education. Voluntary collective licensing as a way for institutions to use copyrighted content in AI tools properly. One more way responsible AI shows up in education. www.researchinformation.info/news/ccc-lau... #AIinEducation #ResponsibleAI

Academic knowledge is shifting from scarcity to abundance. AI generates, AI filters, humans govern. And publishers sitting on authoritative content should be enriching it semantically, not just licensing text. Worth reading both. papers.ssrn.com/sol3/papers.... #ScholarlyPublishing #AIinResearch

Clarivate's Highly Cited Researchers 2021-2025: CAS at 3.6%, Harvard 3.2%. The US has 34 institutions in the top 70. China has seven. And NIH funding is at 20-30% of normal right now. How does that math work out in ten years? clarivate.com/academia-gov... #SciencePolicy #Bibliometrics

An EMBO Reports study asked what authors actually want from AI in peer review. The top answer was pre-submission self-check. Basically AI-assisted screening with LLMs. Surely people are already doing this with Claude Code and NotebookLM? link.springer.com/article/10.1... #PeerReview #AIinResearch

Wiley licensed Cochrane to a platform that runs on advertising and has a history of pay-to-play misinformation. Think Google, but for doctors. If your licensing partner makes money from ads, what happens to the content? www.the-geyser.com/openevidence... #MedicalAI #ResearchIntegrity

At R2R last week, the consensus was sobering. US science funding at 20-30% of normal despite Congress approving the budgets. Add AI displacing junior roles. Why would a young scientist stay? www.nature.com/articles/d41... #SciencePolicy #ResearchFunding

Three stories, one thread. AI is already displacing junior roles. But a rather wild 2028 thought experiment asks what happens when it extends to all of us. This should be part of responsible AI thinking. Also, read more sci-fi. www.nature.com/articles/d41... #AI #FutureOfWork

1.1 million papers cite 12,000 retracted studies. Cabanac says citing retracted work should itself qualify for retraction. Two conferences, same conversation: everyone wants ironclad signals. Will we ever have them? www.timeshighereducation.com/news/retract... #ResearchIntegrity #Retraction

Excited about LeapSpace? 18M paywalled articles via AI. Sounds great. But if it's trained on a select set of publishers, would the outputs be biased? Would open-access-based tools do better? Are we fragmenting the system further? www.science.org/content/arti... #ScholarlyPublishing #AI

CAS cutting APC funding for journals charging $5,000+. Many of us think these costs are unacceptable. The industry is already shifting to China at multiple levels. What would a move to cheaper domestic journals do to the balance? www.researchinformation.info/news/chinese... #OpenAccess #APC

AI that coaches peer reviewers to be more constructive. Not writing reviews. Helping humans write better ones. Tested on 20,000 reviews before ICLR 2025. Problem with this? Is it AI or inertia? www.nature.com/articles/d41... #PeerReview #AIinResearch

Science is trust based and AI has eroded it. Punitive AI detection forces researchers to homogenise their writing to avoid false flags. Worse in the Global South. AI literacy and transparency, or do we just penalise human ingenuity? www.nature.com/articles/s41... #AIinResearch #ResponsibleAI

$250M for one AI researcher. Early-career scientists are 100x more likely to leave for industry. There will always be AlphaFolds that pay big. The community needs to accept that and keep working at what research institutions stand for. www.nature.com/articles/d41... #AI #SciencePolicy

Read Mollick's latest. He breaks AI into Models, Apps, and Harnesses. Most of us (publishing included) are stuck in the Apps era. Chatbots, summaries, rewrites. Harnesses is where the next two years get interesting. www.oneusefulthing.org/p/a-guide-to... #AI #AgenticAI

Interesting paper testing AI as peer reviewer. It rewarded prestige, penalised cautious language, and caved to confident pushback without evidence. Every known human bias, but at scale and without self-awareness. www.the-innovation.org/article/doi/... #PeerReview #AIinResearch

Researchers get answers from AI without ever visiting the publisher platform. COUNTER sees nothing. The Scholarly Kitchen calls it the "zero-click" problem. Silverchair's MCP is a first attempt at fixing this. I'm sold. scholarlykitchen.sspnet.org/2026/02/12/t... #ScholarlyPublishing #AI

Frontiers coins "fast-churn science" and upgrades AIRA, arguably the first publisher integrity screening tool of its kind at this scale. Layered AI detection, not just one method. www.frontiersin.org/news/2026/02... #ResearchIntegrity

AI-assisted grant proposals win ~4% more often at NIH. But they also look more like previously funded work. Clearer, more polished, more... conventional. So is it AI that's the problem, or what review panels reward? www.nature.com/articles/d41... #AIinResearch #ResearchFunding

24,000 submissions to ICML. ArXiv rejections up fivefold since ChatGPT. A preprint from social-science editors maps four scenarios for peer review by 2027. They suggest more AI in production, but with more human evaluation. osf.io/preprints/so... #PeerReview #AIinResearch

Found the AIR matrix via Chris Leonard's Scalene blog - maps responsible AI use across the research process. Exactly the shared language we need. We're running a workshop on this at R2R next week. Come say hi. figshare.com/articles/onl... #ResponsibleAI #R2R2026

PDFs are "lossy" for AI. Proposal: "Knowledge Objects" with Asset + Context + Provenance. Also: add "Understandable" to FAIR → FAIRU. www.researchinformation.info/analysis-opi... #ScholarlyPublishing #AIinResearch

Survey: JIFs still dominate hiring and grants despite being poor proxies for paper quality. Researchers are dissatisfied but keep using them. Lack of alternatives or reluctance to accept them? www.timeshighereducation.com/news/journal... #ResearchImpact #JournalMetrics

Aletheia-Probe aggregates databases to help researchers assess journal credibility. Won't solve predatory publishing, but makes due diligence faster. www.nature.com/articles/d41... #ResearchIntegrity #PredatoryJournals

ArXiv's English-only requirement is a poor decision. AI translation as a workaround creates disadvantages for Global South researchers. Question the policy, not just build bandages. www.nature.com/articles/d41... #Preprints #ResearchEquity

RentAHuman.ai: AI agents hire humans for physical tasks. Payment in crypto. AI as employer, humans as API endpoints. futurism.com/artificial-i... #AI #FutureOfWork #AgenticAI

Science communication needs to be a core publisher competency. The opportunity: communication ecosystems around research - summaries, podcasts, explainers. Shift from gatekeeper to amplifier. scholarlykitchen.sspnet.org/2026/02/06/g... #ScienceCommunication

Three SubStracks worth following for scholarly publishing: James Butcher's Journalology journalology.substack.com Helen King's PubTech Radar pubtechradar.substack.com And mine - STM Innovation Brief - inspired by the two above. chachoch.substack.com

AI training may not be the big revenue source for publishers. The opportunity is "inference" - charging when AI uses content to answer queries. scholarlykitchen.sspnet.org/2026/01/29/g... #AI #ScholarlyPublishing

Anthropic's Super Bowl ads mocking ChatGPT's advertising decision are worth watching. "Deception," "Betrayal," "Treachery" - "Treachery" hits closest to academia. Tagline: "Ads are coming to AI. But not to Claude." youtu.be/3sVD3aG_azw?... #AI #SuperBowl

Prism (OpenAI) and Papers (Opennote) aren't just writing tools - they capture "cognitive signal": how researchers think, not just what they write. Publisher workflows must adapt. robotscooking.substack.com/p/prism-open... #AIinResearch #OpenAI

5% AI disclosure rate across 49,000 JAMA submissions. Maybe we're asking the wrong question. We should perhaps ask how AI was used and how accuracy was evaluated. Writing is thinking - outsourcing it entirely means something gets lost. blogs.lse.ac.uk/impactofsoci... #AIinResearch #ResponsibleAI

The gravity of global science has been moving East for 20 years. Now accelerating. China has challenges, but most researchers are doing legitimate work. Support them, don't isolate them. www.nature.com/articles/d41... #SciencePolicy #GlobalScience

A Science paper found that 13% of special issues had guest editors authoring over a third of papers. Good to make the case that publishers aren't enforcing their own rules. Isn't it time for automated editor integrity checks as well? www.science.org/content/arti... #ResearchIntegrity