The future of education technology after the arrival of ChatGPT

Most education technologies are adopted into schools through a structured process. What happens when technologies arrive on their own?

At a Glance

• Most education technologies are adopted into schools through a process of planning and procurement.
• By contrast, ChatGPT arrived in schools without evaluation, assessment of risks and benefits, training for educators, or any steps considered essential to effective technology integration.
• The emergence of mobile technologies owned by a large percentage of students suggests that ChatGPT may be only the first in a parade of technologies to arrive instead of being adopted.
• A new theory of arrival technologies, grounded in the experience of students, teachers, and school leaders, may prove essential to the future of education technology practice, policy, and research.

Historically, most education technologies have been adopted by schools through a process of planning and procurement (International Society for Technology in Education [ISTE], 2023; Morrison, Ross, & Cheung, 2019). An ideal adoption process might include forming a committee to evaluate potential products; running a pilot evaluation in select classrooms; finalizing a district-wide purchase: training teachers about implementation; and providing ongoing maintenance, training, and upgrades. Of course, few procurement processes live up to the ideal, but even in the busy, chaotic reality of schools, most adopted technologies generally go through some structured process.

By contrast, ChatGPT and other artificial intelligence (AI) technologies simply arrived (Klopfer et al, 2024). Such “arrival technologies” bypass procurement processes. They enter schools through the unguided and spontaneous actions of students, teachers, and vendors.

How technologies arrive

During the first full academic year after the release of ChatGPT (2023-2024), we interviewed 35 teachers from across the country in different geographic regions, grade levels, subject areas, and school types to understand how educators understood this new technology. One common theme in our interviews was the surprising ways that GPT first showed up in schools. Consider the following contrasting experiences, collected from interviews between November 2023 and May 2024 with 35 classroom educators across the country during the first full year of GPT in schools.

Ms. O’Neil, a social studies teacher at an Oregon high school, took a two-year break from teaching during the 2021-22 and 2022-23 school years. When she returned to her job in the fall of 2023, she noticed that all her students were using the Bing.com search engine on their laptops. “They’re all on Chromebooks, and the default should be Google. Why are so many kids using Bing?”

Eventually, Ms. O’Neil realized that her students had switched to Bing because of Microsoft’s partnership with OpenAI, which made GPT-4 and Dall-E 3 available free through Bing’s Copilot feature. She said, “All the dots connected. [The students] aren’t just big Bing.com fans.” Without any discussion, investigation, or systematic planning, students had started using Copilot and thus, GPT-4 regularly to complete their work. Ms. O’Neil had no training, no policy guidance, and no support; she didn’t even get a warning about it from colleagues.

Unplanned student and teacher usage appear to be the two most common initial arrival pathways for GPT.

In Mr. Hunt’s high school English classroom in Washington, D.C., the teacher knew students were surreptitiously using ChatGPT. In fall 2023, Mr. Hunt surprised his students by asking them to use the tool to read and understand abstracts of academic papers for a research project. He told the students “Open up ChatGPT. Don’t hide the tab, open it up. You’re going to copy and paste the abstract into ChatGPT, and you’re going to say, ‘Hey, can you make this readable for a ninth grader?’” He directed the students to read the newly summarized abstract, which would allow them to assess whether the article was relevant to their research. Mr. Hunt, again without any guidance, training, or policy from his school, encouraged his students to try a more productive approach to using these new tools.

Unplanned student and teacher usage appear to be the two most common initial arrival pathways for GPT. However, schools need to gear up for a third entry point: the arrival of generative AI features in tools that have already undergone a procurement process. Schools that invested substantial time — perhaps many years ago — choosing Google Docs over Microsoft Teams or Infinite Campus over PowerSchool will find that their existing tools offer a whole suite of new AI features to evaluate and manage.

The early history of arrival technologies

While ChatGPT is best understood as the first widely and instantly embedded AI-powered arrival technology, it is unlikely that it will be the last. Because arrival technologies are defined by the lack of planning and organizational intention in their integration into schools, they pose serious challenges for teaching practice, school leadership, and educational research.

Historically, most prominent education technologies have been adopted (Cuban, 1986). No seventh grader has shown up in middle school dragging their SMART Board behind them. Past technologies were too large for students to carry and too expensive for families or teachers to purchase. Even during the pandemic, technologies were adopted. To students and parents, it may have appeared as though video conferencing software such as Zoom, Google Meet, and Microsoft Teams simply arrived, but decisions about which platform to use, how to quickly train teachers, and how to ensure access for students were made by school or district level technologists and administrators, often during a very busy handful of days. 

Some prior education technologies have some of the characteristics of “arrival technologies.” Handheld and wristwatch calculators might be one of the best mid-20th-century examples of arrival technologies. Students found them useful, they were cheap and small enough for both students and teachers to purchase and bring to school, and they offered a complex set of both benefits and problems for educators (Demana & Waits, 2000).

More recently, laptops, mobile phones, and the internet all have some features of arrival technologies. Starting in the 1990s, digital technologies began to shrink in size and cost, but slowly, enabling schools to adapt. Laptops were prohibitively expensive for many years, so they entered schools slowly. Similarly, mobile phones showed up in schools first as feature phones and evolved over a decade into smartphones.

As these technologies arrived with individual students, schools developed their own technological infrastructure that “competed” quite effectively with individual options. And the individual options arrived over a long enough period that schools could adopt “bring your own device” programs and other initiatives to plan, monitor, and evaluate individual device use (Cuban, 2001).

Certain web platforms also have features of arrival technologies: Wikipedia and its ready answers to simple factual questions; social media and its amplification of bullying; online paper mills for cheating; and Google Translate for muddled foreign language instruction. While schools eventually developed and adopted policies and strategies around these arrivals, they entered classrooms unbidden by school or district leaders. But, like laptops and phones, they arrived slowly.

The arrival of ChatGPT

ChatGPT had two important differences from these previous digital arrivals. First, especially during the pandemic, nearly every secondary school in the U.S. was designed around constant digital access for individual students. Students have school Chromebooks plus their own mobile phone, and any gaps in that foundation can be filled with computers in libraries, labs, and classrooms. In essence, on the day ChatGPT launched, the vast majority of U.S. secondary school students had a mobile device with online access unmediated by schools, supplemented by some other computing device with access mediated by schools (Anderson, Faverio, & Gottfried, 2023).

Second, ChatGPT was perhaps the fastest-adopted web platform in history. Instagram took 2.5 years to reach 100 million users, TikTok took nine months, and ChatGPT took 60 days (Hu, 2023). This shift was reflected in schools. Almost as soon as it was released in November 2022, teachers and students began playing with it.

The intersection of ubiquitous personal computing infrastructure and record-breaking adoption levels made ChatGPT’s arrival in schools almost instantaneous. While we can learn from the history of calculators and Wikipedia, the differences are stark enough to consider that a new arrival pathway of technology into schools has opened.

The value of adoption processes

In the history of education technology, three principles have been useful for understanding new tools.

• Teachers and students typically use new technologies to replicate existing practices.
• It takes substantial planning, time, experimentation, coaching, and practice to develop useful new practices with technologies.
• Well-resourced schools are much more likely to have the financial, social, and technical capital to support this developmental improvement process, so new technologies typically benefit the affluent (Reich, 2020).

Using technology to replicate an existing practice can be beneficial to learning, harmful to learning, or both — often in ways that are difficult to assess. For instance, five decades of research on the differences between reading on screens and on paper has not codified definitive best practices (Peras et al., 2023). The adoption process allows educators to consider questions that will help them ensure the use is beneficial, for example: “When, how often, and under what circumstances should students use calculators to perform calculations, and when should they use pencil, paper, and their own brain?” Adoption processes allow for controlled experimentation, pilot efforts, and an evaluation of risks and benefits.

New technologies are rarely useful without providing professional learning support to educators, and, ideally, adoption processes provide training, coaching, and practice. For schools operating under resource constraints, adoption processes allow communities of educators to make informed choices, deploying scarce financial resources and scarcer teacher time toward the tools that meet the most urgent student needs.

Of course, this process can be lousy or disastrous, but when compared with the alternative — that Silicon Valley simply releases new technologies whenever they want, and educators must instantly deal with them without help, training, or support — the advantages of adoption versus arrival stand out.

The challenges of arrival technology

In our interviews with teachers, the signature topic of conversation is cheating: Are students using AI to complete homework and classwork in ways that bypass productive thinking and learning? While survey research is still evolving, teachers appear to be more likely to view ChatGPT as a problem than as an opportunity (Lin, 2024). That is exactly the kind of raw deal for classrooms that procurement is designed to avoid.

At this early stage, mitigating the harms of generative AI while leveraging the benefits is a poorly understood, hugely time-consuming task that teachers are having to retroactively deal with after ChatGPT’s arrival. Not surprisingly, wealthy schools have the resources to provide support and staff time to take on these new challenges (Doss et al., 2025). Some affluent schools seem to be developing a kind of “procurement-on-the-fly” process for generative AI — writing usage policies; running pilot programs; subscribing to multiple services and testing them; surveying teachers, students, and families; and going through other steps that might typically unfold before wider adoption.

Procurement-on-the-fly will have to contend with steep competition from consumer products that lack the guardrails of educational products. At Byram Hills School District in Westchester County, New York, the technology coordinators started a procurement process to adopt an education-specific AI platform (such as Playlab, Magic School, or SchoolAI). But at a 2024 district professional development event, a panel of five current students all said they would use ChatGPT whenever possible, rather than the education-specific AI tools. Although these platforms have additional capacity to filter and block inappropriate content, tools to protect student privacy, and other education-specific functionalities, students appear to have already decided which product they prefer.

One silver lining for under-resourced schools is that they have been sufficiently resource-constrained to avoid some harmful trends. For example, affluent schools suffered terrible opportunity costs during the surge of interest in interactive whiteboards. While these tools probably didn’t harm student learning, they cost enormous amounts of money and time that could have been used to help children. Schools that couldn’t afford interactive whiteboards bypassed those opportunity costs.

But no secondary school in the networked world is spared the cost of dealing with ChatGPT. As one urban teacher in southern California told us: “AI hasn’t provided ANY benefits to me as an English teacher. And it hasn’t forced me to change much. It’s just another thing I have to deal with. One of too many . . . I’d love to take a professional learning day on ChatGPT, but our district has zero subs.”

Rethinking research paradigms for arrival technologies

While arrival technologies pose challenges for classroom teachers and school leaders, they also require rethinking educational research. To understand whether a new technology benefits learning, it is immensely helpful to know 1) what the technology is, 2) how it is supposed to be used to benefit learning, and 3) when a school adopted a technology, especially in contrast to similar schools that did not adopt a technology. Arrival technologies confuse the answers to all three questions.

ChatGPT may be a unique event in education technology history. Or ChatGPT could be the first generation of a new cadre of arrival technologies

Existing educational theory has surprisingly little to say about how technology should find its way into schools. The technological pedagogical content knowledge (TPACK) model describes conditions for effective learning with technology but says little about the selection of technology (Koehler & Mishra, 2009). The substitution, augmentation, modification, and redefinition (SAMR) model suggests that schools use technology for transformative purposes rather than simple task substitution. But it says little about how to evaluate potential new tools (Blundell, Mukherjee, & Nykvist, 2022). The Community of Inquiry model describes effective online learning designed in postsecondary institutions, but it doesn’t substantially address how to select a learning management system (Garrison, Anderson, & Archer, 2010). Issues of adoption are certainly addressed in education technology discourse: For instance, the 2024 National Education Technology Plan mentions adoption over 30 times (Office of Education Technology, 2024). But these references to adoption aren’t linked to any explicit theory. It’s assumed the reader has a sense of what a typical procurement process might be like.

ChatGPT may be a unique event in education technology history. Or ChatGPT could be the first generation of a new cadre of arrival technologies. If that is so, the field needs a new theory of arrival technologies to guide the development of new research paradigms and methods, new school-based practices (like “procurement-on-the-fly” or strategic ignoring), new pedagogy, and new policy.

Historically, school technology budgets have been almost entirely devoted to procuring and maintaining education technology systems. If arrival becomes more common, perhaps budgets will need to be held in reserve as rainy-day funds for when new technologies fall from the sky. State departments of education or other consortia may need to play a larger role in the rapid evaluation, piloting, and training around new tools, rather than our current practice of letting 130,000 U.S. schools individually figure out what to do. Perhaps government regulation can ensure the makers of new technologies provide professional development or guidelines for teachers, just as new pharmaceuticals will not ship without guidelines for medical practitioners.

Four suggestions for developing a theory of arrival technologies

Developing a theory of arrival technologies will require a range of scholars looking at different countries and cultures, subjects and disciplines, and school types to develop new approaches that help teachers, researchers, and policy makers. If the effort succeeds, educators of the future will have a strong foundation to react to new arrival technologies. From these early days, it is difficult to know what useful theory for arrival technologies will look like, but we offer four suggestions for its development.

Arrival technologies should not be seen as inevitably integrated in schools.

First, anchor theory development in the lived experiences of students and teachers in classrooms. There are all kinds of ways we might hope ChatGPT could be used, but the most important starting place is how people in the instructional core experience it.

Second, we should bring a great deal of humility to initial suggestions. For instance, early suggestions for how to search and evaluate online information were grounded in reasonable hypotheses. However, those suggestions proved to be disastrously wrong, and a generation of young people — if they got any guidance on web search at all — were likely to have received demonstrably bad instruction (Wineburg & McGrew, 2019). Educators in the field should understand that any advice on ChatGPT or future arrival technologies could easily be similarly flawed, and initial guidance should be seen as tentative.

Third, in the American context, the yawning inequalities between schools need to play a central role in theory development. Strategies for managing arrival technologies in well-resourced schools and neighborhoods may not be as effective in more resource-constrained environments.

Finally, perhaps most importantly, arrival technologies should not be seen as inevitably integrated in schools. ChatGPT has arrived in schools, but if it harms student learning more than it helps, educators should make concerted efforts to push it back out. Over the last decade, education technology researchers have observed that mobile phones provide instant access to vast stores of human knowledge, and they are designed to drive human eyeballs as unconsciously and infinitely as possible toward advertisements. A growing body of research suggests that despite their theoretical value, mobile phones in schools harm adolescent well-being and learning (Abhramsson, 2024). Schools that opened the door to phones are now locking them up in pouches and seeing substantial benefits in student socialization and learning.

As of today, we’re agnostic and uncertain about the best possible future for GPTs in schools, but we strongly discourage the conflation of arrival and inevitability. If new tools enhance learning and well-being, let’s use them to the best of our ability, guided by robust theory on maximizing the benefits of arrival technologies. And if innovations entering our schools harm our students, let’s fight.

 

References

Abrahamsson, S. (2024). Smartphone bans, student outcomes and mental health [Discussion paper]. NHH Department of Economics.

Anderson, M., Faverio, M., & Gottfried, J. (2023). Teens, Social Media and Technology 2023. Pew Research Center.

Blundell, C.N., Mukherjee, M., & Nykvist, S. (2022). A scoping review of the application of the SAMR model in research. Computers and Education Open, 3, 100093.

Cuban, L. (1986). Teachers and machines: The classroom use of technology since 1920. Teachers College Press.

Cuban, L. (2001). Oversold and underused: Computers in the classroom. Harvard University Press.

Demana, F. & Waits, B.K. (2000). Calculators in mathematics teaching and learning. Past, present, and future. In M.J. Burke & F.R. Curcio (Eds), Learning mathematics for a new century (pp. 51-66). National Council for Teachers for Mathematics.

Doss, C.J., Boziak, R., Schwartz, H., Chu, L., Rainey, L.R., Woo, A., Reich, J., & Dukes, J. (2025). AI use in schools is quickly increasing but guidance lags behind: Findings from the RAND survey panels. RAND Research Report RRA4180.

Garrison, D.R., Anderson, T., & Archer, W. (2010). The first decade of the community of inquiry framework: A retrospective. The Internet and Higher Education, 13 (1-2), 5-9.

Hu, K. (2023). ChatGPT sets record for fastest-growing user base — analyst note. Reuters.

International Society for Technology in Education. (2023). Better edtech buying: A practical guide. ISTE.

Klopfer, E., Reich, J., Abelson, H., & Breazeal, C. (2024). Generative AI and K-12 education: An MIT perspective. An MIT exploration of generative AI.

Koehler, M. & Mishra, P. (2009). What is technological pedagogical content knowledge (TPACK)? Contemporary Issues in Technology and Teacher Education, 9 (1), 60-70.

Lin, L. (2024). A quarter of U.S. teachers say AI tools do more harm than good in K-12 education. Pew Research Center.

Morrison, J.R., Ross, S.M., & Cheung, A.C. (2019). From the market to the classroom: How ed-tech products are procured by school districts interacting with vendors. Educational Technology Research and Development, 67, 389-421.

Office of Education Technology. (2024). National education technology plan. U.S. Department of Education.

Peras, I., Klemenčič Mirazchiyski, E., Japelj Pavešić, B., & Mekiš Recek, Ž. (2023). Digital versus paper reading: A systematic literature review on contemporary gaps according to gender, socioeconomic status, and rurality. European Journal of Investigation in Health, Psychology and Education, 13 (10), 1986-2005.

Reich, J. (2020). Failure to disrupt: Why technology alone can’t transform education. Harvard University Press.

Sandholtz, J.H. (1997). Teaching with technology: Creating student-centered classrooms. Teachers College Press.

Wineburg, S. & McGrew, S. (2019) Lateral reading and the nature of expertise: Reading less and learning more when evaluating digital information. Teachers College Record, 121 (11), 1-40.

---

This article appears in the Winter 2025 issue of Kappan, Vol. 107, No. 3-4.