Conversational AI systems are increasingly being used in place of traditional search engines to help users complete information-seeking tasks. This has raised concerns in the political domain, where biased or hallucinated outputs could misinform voters or distort public opinion. However, in spite of these concerns, the extent to which conversational AI is used for political information-seeking, as well as the potential impact of this use on users' political knowledge, remains uncertain. Here, we address these questions: First, in a representative national survey of the UK public (n = 2,499), we find that in the week before the 2024 election, as many as 32% of chatbot users-and 13% of eligible UK voters-have used conversational AI to seek political information relevant to their electoral choice. Second, in a series of randomized controlled trials (n = 2,858 total), we find that across issues, models, and prompting strategies, task-directed conversations with AI to research-specific political topics increase political knowledge (increase belief in true information and decrease belief in misinformation) to the same extent as self-directed Google search. Taken together, our results suggest that people in the United Kingdom are increasingly turning to conversational AI for information about politics. These findings substantially extend prior work by demonstrating that conversational AI's effects on political knowledge generalize across multiple topics, political perspectives, and model families, suggesting that the shift toward AI-assisted political information-seeking may not lead to increased public belief in political misinformation.

Generative artificial intelligence (GenAI) is increasingly integrated into primary care, supporting documentation, clinical summaries, and patient-facing communication. While its potential to reduce cognitive workload is widely recognized, GenAI introduces a safety paradox: as outputs become more fluent, inaccuracies such as hallucinations and omissions may become harder to detect. This creates a largely unrecognized burden in routine care, referred to as the "verification tax," which risks redistributing rather than alleviating clinician workload. A central question is not whether GenAI is sufficiently accurate, but how its outputs are incorporated into clinical workflows and safety structures. In this commentary, we propose a human-in-the-loop (HITL) model that positions advanced practice nurses (APNs) as first-pass reviewers for high-volume, context-sensitive GenAI outputs. This role focuses on routing, contextualization, and verification, ensuring that machine-generated content is appropriately interpreted before influencing patient care. Given their roles in patient education, care coordination, and longitudinal management, APNs may be well-suited to address the gap between standardized model outputs and individualized clinical contexts. A conservative routing architecture and minimum governance framework are outlined, including explicit safety boundaries, structured escalation pathways, and the need for protected time for verification work. Potential limitations include the risk of redistributing invisible labor, model and workflow drift, and equity concerns in patient-facing communication. These considerations suggest that appropriate resourcing, audit mechanisms, and role-based governance are necessary. Positioning APNs as first-pass HITL reviewers may offer a pragmatic approach to integrating GenAI into primary care while supporting patient safety and interprofessional collaboration.

Artificial intelligence scribes use automatic speech recognition and generative language models to document clinical consultations, but official expectations for safe adoption are scattered across privacy, professional, medical device, safety and health-system guidance. We conducted a qualitative content analysis of 74 unique English-language official documents issued by regulators, government bodies, medical colleges and health-system organisations. Documents were identified through a staged grey literature search and coded using a hybrid deductive and inductive framework. We developed the analytic concept of a "drafting fiction" to describe a recurrent cross-document pattern in which artificial intelligence-generated clinical text remains provisional until clinician review and authentication. Across guidelines, transparency is common, consent is uneven, verification is treated as the central safeguard in direct AI-scribe guidance, privacy governance extends from secure storage to voice identifiability, and medical device classification shifts when transcription becomes summarisation. Organisational safety cases, vendor due diligence and data protection assessments narrow product-level uncertainty, but they do not define the encounter-level support needed for clinicians to detect hallucination, omission, automation bias or model drift. The study concludes that artificial intelligence scribe governance is taking shape as a sociotechnical architecture, although it relies on an underspecified human control assumption. Future guidance should define reasonable verification standards, allocate protected review time, resource consent refusal pathways and require monitoring of performance across accents, language and speech differences.

Charles Bonnet Syndrome (CBS) refers to visual hallucinations associated with a decline in vision. Historically, the functional effects of CBS have focused on its psychological impacts, but recent work has highlighted that CBS may also lead to physical outcomes by causing and increasing susceptibility to falls. As awareness of CBS among sight loss and falls prevention specialists is low, its role in the sight loss and falls mechanism may be missed. A falls prevention approach for CBS-related falls spanning primary, secondary, and community services is indicated to ensure this potential contributory factor is not overlooked.

Effective management of pediatric ophthalmic conditions such as myopia, amblyopia, and strabismus depend on caregivers understanding patient education materials (PEMs). However, many existing resources exceed recommended grade 8 reading levels. This review maps AI-generated PEMs in pediatric ophthalmology, summarizes reported outcomes across key educational domains, and identifies barriers and opportunities for clinical integration. Following PRISMA-ScR guidelines, a search of Ovid MEDLINE, Embase, and Scopus (January 2015-July 2025) identified studies evaluating LLM-generated PEMs in pediatric ophthalmology. Eligible studies evaluated artificial intelligence for pediatric ophthalmology patient education and reported at least one educational outcome. Data were extracted independently by two reviewers and synthesized descriptively. Twenty studies (2023-2025) evaluated 3286 AI-generated materials across 14 LLMs. Readability was assessed in 17 studies; 59% (37/63) of study-level LLM evaluations exceeded recommended 8th-grade reading levels. ChatGPT-4o evaluations were more often acceptable (52%) than ChatGPT-3.5 evaluations (18%), although these were not head-to-head comparisons. Median understandability (PEMAT-U) exceeded adequacy thresholds (≥70%). Professional society materials were more accurate and higher quality but significantly less readable. No included study explicitly classified outputs as hallucinations under its applied assessment framework; however, incomplete treatment-related content was common. Only two studies assessed patient or caregiver perspectives. LLM-generated PEMs in pediatric ophthalmology are generally understandable and accessible but often exceed readability standards and lack actionable guidance. While professional society materials remain more accurate, LLMs offer scalable tools for generating patient-centered resources. Integrating LLMs into clinician-supervised workflows and emphasizing co-designed, multimodal education strategies may enhance accessibility and clinical safety.

Writing discharge notes (DNs) is a time-intensive task in inpatient care that contributes substantially to physician workload and burnout. Large language models (LLMs) offer potential opportunities to automate clinical documentation. However, concerns regarding accuracy, reliability, data privacy protection laws and data governance, and compliance with healthcare privacy remain. This study evaluated the performance of locally deployed, domain-adapted LLMs for generating German-language DN in internal medicine. In this retrospective study, electronic health record data from a Swiss tertiary care hospital were used. Clinical information from 3 inpatient scenarios served as input for the 3 open source LLMs (Mixtral 8×7B, Mixtral 8×22B, and LLaMA 3.1 70B). Zero-shot prompting, 4-shot in-context learning, and supervised fine-tuning using QLoRA were compared. Generated DNs were evaluated using automated metrics, blinded physician assessment with direct comparison of all relevant clinical information using the modified Physician Documentation Quality Instrument, and additionally an LLM-as-a-judge framework and a Turing Test. Across all evaluation methods, the fine-tuned Mixtral 8×7B model achieved the best overall performance compared with prompting-based approaches and larger nonspecialized models. Simpler cases consistently received higher-quality ratings. Physician inter-rater agreement was low, highlighting the challenges of objective quality assessment for clinical documentation. In a Turing Test, physicians were unable to reliably distinguish artificial intelligence-generated from clinician-authored DNs, performing at chance level, demonstrating that high-quality DN generation is feasible using locally deployed LLMs within secure hospital infrastructures. Domain-specific fine-tuning substantially improves performance and may enable practical deployment. Future work should address further hallucination mitigation, robustness across medical specialties, and real-world integration.

To describe suspected acute acrylamide toxicosis in dogs following ingestion of consumer polyacrylamide-based hydrogel migraine caps, eye masks, and wraps, and to quantify acrylamide monomer concentrations in associated products.

We present a case of disseminated Nocardia otitidiscaviarum infection in a 72-year-old male with a history of renal transplantation on chronic immunosuppressive therapy with prednisone, cyclosporine, and mycophenolate mofetil. He presented with pneumonia and was initially treated with ceftriaxone and azithromycin. Bronchoalveolar lavage (BAL) and cultures from a biopsied thigh lesion both grew N. otitidiscaviarum. His hospital course was complicated by visual hallucinations, and brain magnetic resonance imaging (MRI) revealed multiple intracranial abscesses. He initially received meropenem, trimethoprim-sulfamethoxazole (TMP-SMX), and linezolid. His antimicrobial regimen was subsequently modified during the treatment course. Linezolid was discontinued after hematologic abnormalities developed, TMP-SMX was temporarily held because of worsening renal function and later reintroduced, and meropenem was also discontinued. He was discharged on TMP-SMX and minocycline and completed an initial six-week antimicrobial treatment course from the time of diagnosis. He was subsequently maintained on minocycline and moxifloxacin; moxifloxacin was later discontinued because of QT-interval prolongation, while minocycline was continued for slightly longer than one year, with clinical and radiologic resolution.

Artificial intelligence (AI) is entering orthopaedic education through image interpretation, simulation, technical skill assessment, examination support and content generation. Most evidence, however, comes from undergraduate and postgraduate settings, whereas practising surgeons require continuing medical education and continuing professional development (CME/CPD) that supports safe adoption, maintenance of competence and practice improvement. This matters because AI may improve task performance while it is present without producing durable learning or safer independent practice. This critical narrative review asks: what learner benefit does the current orthopaedic education literature demonstrate, and how should that evidence be translated into AI-enabled CME/CPD for practising orthopaedic clinicians? Orthopaedic-specific reviews and primary studies were synthesised and supplemented by a targeted PubMed and Google Scholar update through 4 August 2026. Across modalities, a consistent pattern emerges: AI is strongest in tightly bounded tasks with objective outputs, whereas evidence for retention, transfer, workplace performance and patient benefit is sparse. Diagnostic support can improve accuracy, speed or confidence while available; machine-learning models can distinguish expertise levels in selected simulated procedures; and large language models perform strongly on some text-based examinations but remain vulnerable to explanation errors, hallucinated reasoning, model and version dependence, and image-rich tasks. One recent historical cohort study suggests that structured AI-assisted peer teaching may improve knowledge, clinical reasoning and three-month retention. We therefore propose an AI-specific outcomes framework that separates model capability and assisted performance from learning, transfer, workplace performance and patient or system outcomes, alongside four design principles for CPD. For CME/CPD, the priority is not simply to use AI more often, but to ensure that AI-enabled education produces demonstrable learning, safe transfer and improved practice.

The ability to access specific operational guidelines in unstructured technical documentation is a crucial and unsolved problem in safety sensitive areas like aerospace, clinical care, and regulatory compliance, where approximate retrieval and hallucination are unacceptable. Current systems of Retrieval Augmented Generation (RAG) relying on dense vector embeddings are unable to offer the deterministic accuracy such environments require. In this paper, a neurosymbolic framework is provided to overcome this limitation through the translation of natural language documents into three complementary representations of symbolic knowledge: a SQLite (Lightweight SQL Database Engine) database of SAT (Boolean Satisfiability) - based propositional reasoning Conjunctive Normal Form (CNF) clauses, a directed property graph relational traversal and a geometric spatial index which represents the quantitative conditional rules as axis aligned hyper rectangles in an eight dimensional parameter space. Each of the three representations is built on the same LLM (Large Language Model) - based rule extraction pipeline at the same time during offline stage, where there is no embedding of vectors of any type. Online three path reasoning architecture combines SAT at query time entailment checking, knowledge graph traversal, and exact geometric containment testing - translating the operational state of a user to a parameter space point and recalling all rules with hyper rectangular regions containing the point. The outputs of all the three paths are combined before constrained answer generation, basing each answer on logical, relational, and numerical evidence. Because retrieval is performed using mathematically precise symbolic procedures rather than similarity approximation, the SAT, graph, and geometric reasoning components are deterministic and fully explainable; large language model calls are still used for rule extraction and answer synthesis, so the system as a whole is hybrid rather than fully deterministic, and a faithfulness filter together with context-confined generation substantially mitigate, rather than formally guarantee, hallucination. The experimental findings show that this tri-representation strategy markedly enhances retrieval accuracy and reliability compared to embedding-based baselines in complicated operational query conditions, developing a scalable base of rule-based decision support in high stakes areas.