Artificial intelligence (AI) is everywhere. From platforms such as ChatGPT, Claude, and MagicSchool AI to everyday conveniences like smart appliances and customized entertainment streaming options, AI is clearly here to stay and changing how we live. Educators in particular encounter information about AI from the media, their school districts, and educational vendors, as well as in professional conferences and workshops. But because information about AI is seemingly everywhere, sometimes it is difficult to know what has been rigorously researched and what is just commonly accepted or supported.

Sifting through the “research” on AI can be confusing. Duke and Martin (2011) point out that research can refer to an idea, product, intervention, or other material that has been studied and tested for a particular effect (sometimes referred to as “research-tested”), or it can mean that one of those items was informed by existing research but has not been tested (sometimes referred to as “research-based”). They also caution that something written by a researcher is not necessarily considered research. For instance, they point out that if a novelist writes a shopping list, the list is not automatically a novel.

The NSTA Research Committee defines research as the systematic study of how people learn science to improve science education practices and student outcomes across various levels of learning. For our purposes, the key phrase is systematic study. For something to be considered research (and therefore have conclusions, ideas, and practices that are likely to be more trustworthy), a systematic study must have been undertaken.

So, how can teachers make sense of the information available on AI and find out what has actually been studied and what the research shows? First, we’ll walk through two examples of questions that teachers can ask themselves when reading an article about research on AI. Then, we discuss the takeaways from this exercise and how teachers can use research to inform their practice.

Questions to Ask When Reading Articles About Research on AI

Article 1: Vilcarino, Jennifer, and Lauraine Langreo. 2025. "Rising Use of AI in Schools Comes with Big Downsides for Students." Education Week, October 8.

What are the research questions and purpose? There are no research questions or research purpose; instead, the article summarizes a few reports. There may be questions in the reports the authors cite, but they did not link to these reports, so the reader can’t be sure what they include.

What methods did the researchers use? Do the methods seem rigorous and appropriate for the question? The article does not have a methodology. The reports that the authors cite may include a methodology, but they did not link to the reports, so the reader doesn’t know what they include.

What are the limitations of the study? What might cause me to think twice about using this information? Again, this article is not a research study. A limitation of the article is that the sources are not linked, so the reader doesn’t know if the statistics and opinions are grounded in research.

Who supported the research? Do they have any biases? This article discusses findings and opinions from the Center for Democracy and Technology, ISTE + ASCD, and Common Sense Media. The Center for Democracy and Technology is described as center-left politically, but its website states that it is a “nonpartisan, nonprofit organization fighting to advance civil rights and civil liberties in the digital age.” The center receives support from Google, Meta, Microsoft, Apple, and Amazon. ISTE + ASCD is a nonpartisan nonprofit as well and described as apolitical, focusing on supporting learning “through impactful pedagogy and meaningful technology use.” Finally, Common Sense Media is also a nonpartisan, nonprofit organization that is described as progressive or left of center. It provides education to families and advocates for legislation to keep children safe when they engage with media and technology.

These organizations appear to be relatively unbiased sources, although it would be wise for a reader to look at each research report closely to see who sponsored reports and research and whether the research was conducted without bias.

Are the authors discussing something that is research-tested or research-based? Once again, this article is not a research study, but rather a description of generalities of other studies, so it is neither research-tested nor research-based.

What should I take away from this study, and how confident should I be in the conclusions? The article cites a number of statistics that point to AI as useful to teachers and students alike, but the authors indicate that teachers need more professional development and students should have more guardrails. However, we would feel more confident about these takeaways after reading the original reports the author cite.

Article 2: Schwartz, Sarah. 2025. "Brain Activity Is Lower for Writers Who Use AI. What That Means for Students." Education Week, June 26.

What are the research questions and purpose? This research focuses on finding out the “cognitive cost” of using AI in the educational context of writing an essay. 

What methods did the researchers use? Do the methods seem rigorous and appropriate for the question? The researchers assigned participants to one of three groups—LLM group, Search Engine group, or Brain-only group—where each participant used a designated tool (or no tool in the Brain-only group) to write an essay. The researchers used electroencephalography (EEG) to record participants’ brain activity to assess their cognitive engagement and cognitive load and gain a deeper understanding of neural activations during the writing task.

What are the limitations of the study? What might cause me to think twice about using this information?  The authors themselves note that the paper “has not yet been peer-reviewed, thus all the conclusions are to be treated with caution and as preliminary” (Kosmyna and Hauptmann 2025). They also acknowledge as a limitation the fact that the study only included undergraduate and graduate students from a particular region and focused on essay writing in an educational setting, so the information may not be generalizable across other tasks.

Who supported the research? Do they have any biases? The study is reported by EdWeek and was conducted by researchers at the Massachusetts Institute of Technology, Wellesley College, and the Massachusetts College of Art and Design. The work doesn’t acknowledge any external funding or conflicts of interest.

Are the authors discussing something that is research-tested or research-based? This article reports on an actual research study of the impacts on brain activity related to using ChatGPT during a writing exercise. The researchers found that participants who constructed essays with the assistance of ChatGPT exhibited less brain activity during the task than participants who were asked to write an essay on their own. The AI users were much less likely to be able to recall what they had written and felt less ownership over their work. Independent evaluators of the essays found that the AI-supported essays lacked individuality and creativity.

What should I take away from this study, and how confident should I be in the conclusions? Given the preliminary nature of the findings, the limitations of the study, and the lack of generalizability, the study raises more questions than it provides answers. The lead researcher, Nataliya Kosmyna, offers only a modest takeaway in this article: “What it could potentially tell us is that timing could be very important for when you integrate these tools.”

Takeaways for Teachers

While many are quick to embrace AI and cite the unprecedented opportunities for enhancing learning and access to information afforded by these tools, we echo caution that research is needed to understand the potential impacts (both positive and negative) on learning. That evidence base is still being built. In the meantime, we put forward a few things to keep in mind:

1. Don’t confuse reporting on research with research itself. 
   The two examples above illustrate that articles about AI often vary dramatically in how directly they engage with evidence. Be sure to distinguish between journalism, opinion pieces, and peer-reviewed research studies. Whenever possible, follow citations back to the original source rather than relying solely on summaries or headlines.
2. Be cautious about broad claims—positive or negative.
   Public conversation about AI often swings between the extremes of enthusiasm and alarm. However, the research suggests a bit more nuance: the effects of AI depend on how, when, and why the tools are used. Teachers should be skeptical of claims that AI will either revolutionize learning or irreparably harm it, especially when those claims are not supported by a great deal of evidence.
3. Use professional judgment alongside research evidence.
   Research can inform practice, but it rarely provides simple answers. Teachers must balance research findings with their own professional knowledge of their students, learning goals, curriculum, and local context. Evidence should inform decision-making – not dictate it.
4. Stay informed as the evidence on AI evolves.
   AI technologies are changing rapidly, and the research is developing just as quickly. Conclusions that seem well-supported and reasonable today may be refined or challenged in the future. Maintaining a critical and curious stance toward new research on AI will help educators make informed decisions as the field evolves.
5. Participate in Discussions About Responsible AI Use
   Beyond considering how research on AI can inform teaching practice, we encourage educators to join the discussion about the range of social and ethical concerns related to AI. A good place to start is the recent STEM Teaching Tool, “What Concerns SHOULD We Be Discussing About Using AI in Education?” (Bell 2026).

Ultimately, teachers do not need to become AI researchers to make informed decisions about AI in education. However, developing a habit of critically examining claims, locating original sources, considering study limitations, and looking for patterns across multiple studies can help educators distinguish evidence-based recommendations from ideas that “seem right.” 

References

Bell, Philip. 2026. “STEM Teaching Tools #109: What Concerns SHOULD We Be Discussing About Using AI in Education?” STEM Teaching Tools. https://stemteachingtools.org/brief/109.

Duke, Nell K., and Nicole M. Martin. 2011. “10 Things Every Literacy Educator Should Know About Research.” The Reading Teacher 65 (1): 9–22.

Kosmyna, Nataliya, and Eugene Hauptmann. 2025. “Your Brain on ChatGPT” Accumulation of Cognitive Debt When Using an AI Assistant for Essay Writing Task.” Brain on LLM. https://www.brainonllm.com/.

Julianne A. Wenner, PhD, is a Professor of Science Education at Clemson University and the new Division Leader for Research in Science Education for NSTA. Her areas of research coalesce around ensuring that all students feel as though they can participate in and/or pursue science careers or hobbies. Thinking of this from a systems perspective, JWenner investigates how teachers take on leadership to advocate for science education; explores how to best prepare elementary teachers to teach high-quality science; and examines the ways in which families support their children in participating in and enjoying science.

Debi Hanuscin, PhD, is a Professor of Science, Math, & Technology Education at Western Washington University. Her work focuses on supporting teachers’ development of science teaching knowledge and practices across a range of settings, including university coursework, methods classes, and field experiences. She is the outgoing NSTA Division Leader for Research in Science Education.

Note: This blog post is part of the blog series Research Worth Reading, which features the latest highlights of recent research in science education with practical applications for your classroom.

The mission of NSTA is to transform science education to benefit all through professional learning, partnerships, and advocacy.