This Technology Can Write Student Essays: Is There Any Educational Benefit?

Artificial intelligence has found a comfortable place in our lives over the last decade. Where we used to turn to physical maps for directions, Google Maps now shows us how to get to our desired destination by the quickest route possible. With one command to Alexa, we can do a math calculation, learn a random science fact, purchase an item on Amazon, and even call our best friend.

Now, an artificial intelligent model called Generative Pre-trained Transformer 3, informally known as GPT-3, aims to end writer’s block by doing the writing for us. The implications for education are enormous, in both bad and good ways.

I first heard of GPT-3 at the Learning 2025 conference hosted by AASA, The School Superintendents Association earlier this summer in Washington. In a room filled with superintendents, principals, teachers, and other education leaders, Bill Dagget, former educator and founder of the International Center for Leadership in Education, gave a keynote speech that mentioned GPT-3 and the possibility of students using this model to do their classwork and homework.

The attendees were filled with both awe and dread. What is this technology and what impact will it have on students’ ability to write in the future? Will students become overly dependent on AI to do the hard thinking and writing for them? These were just some of the questions in the room. Like any education reporter searching for a story, I was incredibly intrigued by the technology and had some of the same questions.

So what is GPT-3?

GPT-3 was created by OpenAI, an artificial intelligence company. According to OpenAI, GPT-3 is a machine-learning model that uses internet data to understand and generate natural language text. Trained on 540 billion words and 175 billion parameters, the model produces text in response to a prompt.

When you input the prompt, the model will generate a complete text that attempts to match whatever context, pattern, and directive you gave it. For example, if you input “write a tagline for a car company,” it will return a completion like “a car for every budget” or “the best way to get around.” Prompt it to “summarize a complex text for a 1st grader” and it generates a summarized text in language that a 1st grader can understand.
But it isn’t just an ask and answer technology. GPT-3 is trained on text generation, completion, summarization, and creative writing.

GPT-3 does not just understand words; it also understands context, sentence structure, and dual meanings in multiple languages, allowing it to summarize text, write outlines, essays, reports, and recommendation letters, all in human-like prose. And beyond writing, the model is trained in producing content related to math and science.

“It’s a game changer. It hasn’t really broken into mainstream yet, but it’s coming in a few years’ time. I think the education profession hasn’t really got its head around the implications of this yet,” said Alistair Van Moere, the chief product officer at MetaMetrics Inc., one of the more than 300 companies exploring the use of GPT-3 in its products.

Incorporating GPT-3 in the classroom

One of the most glaring concerns about GPT-3 and models like it is that they might have negative impacts on students’ ability to learn to write on their own. After all, feeding an essay prompt into the computer and having it produce the entire essay for you in human-like text is many students’ dream.

Longtime educator Ray McNulty, the former education commissioner of Vermont, believes educators should be proactive instead of reactive to AI advancements by finding ways to incorporate the technology into instruction.

“There’ll be a transition where this will become second nature [for teachers and students],” said McNulty. “How do we do it thoughtfully and carefully? And how do we do it to enhance the learning for our students?”

McNulty, who is now the president of the nonprofit Success Practice Networks and the National Dropout Prevention Center, gave an interactive presentation with “Future Wheels,” a tool that helps anticipate and explore future possibilities. He used the technology to explore the possibilities for artificial intelligence writing tools in education during a session at the superintendents’ conference. Participants brainstormed both potential negative and positive outcomes from using AI writing tools to begin thinking about ways that the technology could be effectively incorporated into teaching and learning.

“What we’re trying to do is have school districts know this stuff is coming and how do they begin to anticipate and prepare for [it],” McNulty said.

Right now, GPT-3 is not used in any regular K-12 classrooms, but Van Moere has some ideas about how it could be implemented. He recommends teachers get acquainted with the technology during class activities such as customizing stories with students’ names and characteristics, doing a side-by-side comparison of a GPT-3 story and one written by a human to see if students can spot the difference, or using GPT-3 to help students brainstorm their idea and then write their own essays. This way, students are being taught to use AI as a tool instead of as the answer.

Companies need to ‘bring teachers to the conversation’

GPT-3 is just one of the many natural language learning models in the technology industry that currently use applications geared towards K-12 students. Their proliferation raises questions about what the role of the teacher will be in the future, and there are concerns that AI can have its own set of biases that pose potential problems for use in K-12 education.

Ed-tech company CEO Sydney Montgomery said to be able to effectively use AI writing tools like GPT-3 in the classroom, the models need to be built with the input of teachers.

“I think ed-tech CEOs or companies that are building tools need to bring teachers to the conversation because they see things that we don’t,” she said. “If we can partner a little bit more, it will also help with the hesitancy of [educators to use] the technology.”

Montgomery is founder and CEO of Outline, an education technology company that uses an artificial intelligence writing model to help students outline admissions essays, memos, and law briefs. The AI model that Outline uses does not author the entire essay for you like GPT-3 does—rather, it creates a detailed outline to help jump start the writing process. Montgomery said AI can be a helpful writing tool for K-12 students who may not have the support of a teacher, parent, or guardian in the classroom or at home.

“We need AI to fill in the gaps where students don’t have support because otherwise students get left behind,” said Montgomery.

How well does the technology work?

Currently, a free trial for GPT-3 is accessible to the public on OpenAI’s platform. I wanted to try it out myself. I went on OpenAI, created a trial account, and input some commands.

I first asked the model to “write an essay for a 5th grade science class on the Webb telescope.” It returned an elementary-level, seven-sentence paragraph about the Webb telescope. To generate more text, I clicked submit again under the first paragraph. (This essentially tells the technology to keep generating more text about the same task.)

However, the paragraph was not without flaws. The machine-generated text seemed to not know that we are well past the year 2018, as it contained a passage indicating that the telescope would be launched in 2018. (It was actually launched in 2021.) That suggests a limitation with the technology’s ability to generate content based on recent events. A note on the free trial web page suggests a partial explanation: “Our default models’ training data cuts off in 2021, so they may not have knowledge of current events.” My prompt and the resulting essay follow:

I then directed it to “write an analysis essay about Romeo and Juliet.” It returned a simple four-paragraph essay about the story. The essay is clear and concise, and it does read like something a human—possibly a student—would produce. It returned a complete paragraph that consisted of no glaring errors.

Though experts say the technology can generate well thought-out essays, the free trial did not allow me to produce more complex and longer text. It limited the technology’s use to a maxiumum of 4,000 words.