Advancing your literature searching skills (Science)

AI Reshapes Research

Scientific publishing produces around 3 million papers each year, overwhelming traditional search methods. In 2026, AI tools like ChatGPT and Elicit promise to cut through this deluge, analyzing vast datasets in seconds. But this shift arrives amid concerns over AI's limitations, including hallucinations—fabricated citations—and incomplete coverage of older or niche literature.

The change stems from generative AI's rise post-2023, now operational in editorial workflows. Tools such as Semantic Scholar rank papers by relevance, while others synthesize summaries across databases like PubMed and Scopus. Yet, accuracy varies: Bing AI outperforms ChatGPT in precision for systematic reviews, but neither matches human-curated searches comprehensively.

Researchers in biomedicine and climate science feel the impact most acutely. In drug development, missing key studies could delay therapies, affecting patients worldwide. Grant funders like the NIH, with budgets exceeding $40 billion annually, increasingly scrutinize AI-assisted proposals for rigor. Deadlines tighten as journals demand transparent methods, with retractions surging 20% in recent years partly due to poor reviews.

Costs mount for institutions subscribing to premium AI platforms, often $100-500 per user monthly, while free versions risk data privacy breaches. Consequences of inaction include stalled careers—postdocs face publication pressures—and broader societal harms, like infodemics in public health.

Less obvious tensions emerge between speed and equity. AI algorithms favor English-language, high-impact journals, sidelining global south research and perpetuating biases. Trade-offs pit efficiency against critical thinking: overdependence might erode skills, as seen in studies where AI users overlook contradictory evidence. Counterarguments highlight hybrid approaches, blending AI with traditional databases for robust results.

Stakeholders clash: publishers embrace AI for integrity checks, while ethicists warn of reduced human oversight. Surprising data shows AI spotting overlooked connections, like cross-disciplinary links in sustainability, but failing on gray literature—reports and theses not indexed widely.