AI’s Growing Footprint in Brazilian Research
The rapid adoption of generative artificial intelligence tools has transformed how researchers in Brazil and around the world approach scientific work. At the Universidade Federal de Goiás (UFG), Professor Sávio Teles of the Centro de Excelência em Inteligência Artificial (CEIA) has issued a clear warning about the dual risks of AI shaping both the creation and the assessment of scientific output in the country.
Teles points to a 2026 study published in Organization Science that documented a 42 percent rise in scientific submissions since the 2022 launch of ChatGPT. Roughly 10 percent of the 4,715 papers analyzed showed strong AI influence, with measurable declines in readability scores using the Flesch Reading Ease metric. These trends, he argues, signal more than efficiency gains; they point to structural vulnerabilities in Brazil’s research ecosystem.
The Publish-or-Perish Pressure and Paper Factories
Brazil’s academic environment remains heavily influenced by the “publish or perish” culture. Researchers face intense pressure to produce volume for career advancement, funding from agencies such as the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) and the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES), and institutional rankings. Teles notes that this environment creates fertile ground for “paper factories”—automated operations that handle data analysis, code generation, and even manuscript drafting with minimal human intervention.
While AI can accelerate routine tasks, Teles emphasizes that skipping the intellectual labor of hypothesis formulation, deep analysis, and contextual interpretation leaves researchers ill-equipped to judge the quality of outputs. “If the researcher does not go through the generation of that content, he cannot evaluate and judge whether the result is good or not, because he never lived that process,” he told Jornal Opção.
AI in Peer Review: A Closed Loop Risk
Perhaps most concerning is the expanding role of AI in the evaluation process itself. The Organization Science study found that more than 30 percent of journal reviews already incorporate some level of AI assistance. Teles warns that without robust human oversight, science risks entering a self-referential cycle: “If we write and judge the texts with AI, then, in the end, everything we will consume of science will be produced and maintained by the AIs.”
This scenario threatens the integrity of Brazil’s Qualis system, which classifies journals and influences everything from graduate program evaluations to researcher productivity metrics. Automated reviews could amplify biases, reduce nuance, and erode the critical dialogue that has long defined peer review in Brazilian universities.
Ethical Guidelines Emerging from Brazilian Researchers
In response to these challenges, Brazilian scholars have begun developing practical frameworks. Ricardo Limongi, also at UFG and editor-in-chief of the Brazilian Administration Review, co-authored “Diretrizes para o uso ético e responsável da Inteligência Artificial Generativa: um guia prático para pesquisadores,” released in late 2024 by Intercom. The guidelines stress mandatory disclosure of AI use, prohibition of presenting AI-generated content as human-authored work, and retention of full researcher responsibility for final outputs.
Similar efforts appear in SciELO discussions and workshops at institutions nationwide, where faculty debate whether training should focus on prompt engineering or on preserving the full scientific process. Limongi’s work highlights that transparency and human accountability remain non-negotiable.
Regulatory Developments at National Level
Brazil’s regulatory landscape is evolving. Recent CNPq directives and debates at the Conselho Nacional de Educação (CNE) require declaration of generative AI use at every research stage, ban the presentation of AI content as original authorship, and discourage AI-generated peer reviews due to risks to quality and impartiality. These measures align with broader national strategies for responsible AI, yet enforcement and cultural adoption across thousands of graduate programs remain works in progress.
Impacts on Training the Next Generation
The stakes extend beyond current publications to the formation of future researchers. Graduate students and early-career academics risk developing shallow critical skills if AI handles core intellectual tasks. Teles and colleagues advocate for curricula that treat AI as a powerful assistant while requiring students to demonstrate independent reasoning, hypothesis development, and rigorous evaluation.
Universities such as UFG are already piloting workshops that combine tool training with process-oriented reflection, aiming to prevent the “shortcut culture” from taking root in doctoral programs.
Stakeholder Perspectives Across the Sector
University administrators express concern that unchecked AI adoption could distort performance metrics used in CAPES evaluations and funding allocations. Journal editors worry about maintaining standards amid rising submission volumes. Funding agencies like CNPq emphasize the need for auditable, transparent algorithms. Meanwhile, researchers themselves report both excitement about productivity gains and anxiety over maintaining authenticity in an increasingly automated environment.
Practical Solutions and Best Practices
Experts recommend several concrete steps:
- Mandatory disclosure of AI tools and their specific contributions in every manuscript and review.
- Retention of human responsibility for all intellectual claims and conclusions.
- Training programs that teach critical evaluation of AI outputs rather than mere prompting.
- Development of institutional policies that reward quality and reproducibility over sheer volume.
- Investment in open-science practices that make AI-assisted work verifiable.
These approaches echo recommendations from the Intercom guidelines and ongoing SciELO dialogues.
Photo by Markus Spiske on Unsplash
Future Outlook for Brazilian Science
If addressed proactively, AI can accelerate discovery, improve accessibility of literature, and support researchers in resource-constrained settings. However, without deliberate safeguards, Brazil risks a homogenized, lower-quality scientific output that undermines its growing international presence. The warning from UFG’s Sávio Teles serves as a timely call for the entire higher-education community—from individual labs to national agencies—to reaffirm human judgment at the center of the research enterprise.
Looking Ahead: Actionable Steps for Institutions
Brazilian universities and research bodies can lead by example. Updating graduate program criteria to include AI literacy alongside traditional research skills, piloting AI-audit protocols for journals, and fostering cross-institutional working groups on responsible use are immediate priorities. International collaboration with bodies such as UNESCO and peer networks in Latin America can further strengthen these efforts.