Students benefit less from strict rules than from guided practice and space to reflect.
In classrooms where AI tools have quietly become fixtures of student life, one English instructor at Amarillo College chose neither prohibition nor indifference — she chose education. Sara Welshimer observed that her students possessed familiarity with AI without the critical judgment to use it well, and so she designed assignments that slowed them down, made them choose, and asked them to explain why. Her experiment suggests that the deeper challenge of this technological moment is not access, but intentionality — and that teaching students to think alongside AI may matter more than teaching them to avoid it.
- Half of Welshimer's students were submitting AI-polished essays they hadn't truly written, while the other half were struggling through drafts that a well-used tool could have meaningfully improved.
- The real crisis wasn't cheating — it was that students couldn't distinguish between using AI as a crutch and using it as a craft partner, because no one had ever taught them the difference.
- Welshimer designed a structured revision assignment using the Hemingway Editor, forcing students to accept, rewrite, or reject each suggestion and then defend their choices in writing.
- Students arrived with genuine anxieties — fear of misinformation, loss of voice, ethical ambiguity — and the assignment gave those fears a productive place to land rather than fester.
- By spring 2025, refined timing and in-class modeling had transformed how students engaged: graded AI use made them more selective, more skeptical, and more deliberate than casual use ever had.
In fall 2023, Sara Welshimer, an English instructor at Amarillo College, noticed a troubling divide in her students' work. Some essays arrived suspiciously polished — and private conversations revealed that many of those students had asked AI to generate their drafts entirely, without much sense that this was a problem. Others turned in rough, unfocused papers that might have genuinely benefited from the very tools their peers were misusing. The contradiction raised a question Welshimer couldn't ignore: what if students learned to use AI deliberately, as a revision aid rather than a ghostwriter?
Rather than ban the technology, she built an assignment around it. Using the Hemingway Editor — a tool that flags clarity and grammar issues without generating new content — students revisited a previously submitted paper, reviewed a set number of suggestions, and made conscious choices: accept, rewrite in their own words, or reject. A written reflection required them to explain each decision. The assignment imposed a productive slowness that ordinary revision rarely demands.
What she found was uneven but revealing. Some students could already distinguish between generative AI like ChatGPT and assistive tools like Grammarly; others had barely thought about the difference. Many brought real concerns — fear of factual errors, worry about losing their voice, uncertainty about where the ethical line fell. The assignment didn't dismiss those anxieties; it gave students a structured way to work through them. Even skeptics who found the tool's suggestions too mechanical often ended up making meaningful revisions, selecting only what served their intentions and discarding the rest. That selective engagement pointed toward something Welshimer was hoping to cultivate: judgment.
When she taught the assignment again in spring 2025, she refined the structure — demonstrating her own decision-making aloud, moving AI work earlier in the drafting process, and giving students in-class time to engage with the tool. A key observation emerged: when AI use was optional and ungraded, students treated it casually. When it became part of the graded writing process, they became noticeably more deliberate, approaching the tool as something to consult and question rather than simply follow.
Welshimer's broader takeaway is one that many instructors may find useful: students gain little from blanket prohibitions or vague warnings about AI. They gain from clear expectations, guided practice, and space to reflect. When AI is framed as a partner in learning rather than a threat to be managed, students engage more honestly with their own writing — and leave better prepared to use these tools responsibly in whatever comes next.
In the fall of 2023, Sara Welshimer, an English instructor at Amarillo College, began noticing a troubling split in her students' work. Half the essays gleamed with perfect grammar and polish, yet when she sat down with those students one-on-one, many confessed they had leaned heavily on AI tools—sometimes asking the software to generate entire drafts from scratch. What struck her most was their lack of concern. They didn't think they'd done anything wrong. The other half of the class produced the opposite: rough, meandering papers that seemed to cry out for a quick pass through some kind of writing assistant. Welshimer found herself thinking these students could have benefited from exactly the tool their peers were misusing.
That contradiction sparked a question that she suspects many instructors are now asking themselves: what if students learned to use AI deliberately and critically, not as a shortcut but as a genuine aid to revision? Rather than forbid the technology or pretend it didn't exist, Welshimer decided to experiment. She would teach students to treat AI as an assistant, not a replacement, and see what happened.
The gap she identified wasn't really about access. Most of her students already used AI tools. The problem was intentionality. They had familiarity without critical judgment. They knew these tools existed but couldn't reliably say when or how to use them well. This distinction—between knowing a tool exists and knowing how to use it thoughtfully—became the foundation for what researchers call AI literacy: the ability to evaluate AI technologies critically, work alongside them effectively, and deploy them ethically in different contexts.
For her first experiment, Welshimer designed an assignment around the Hemingway Editor, a tool that flags issues with sentence clarity, grammar, and readability but doesn't generate new content. Students took a previously submitted character analysis paper, ran it through the tool, and then made deliberate choices: accept the suggestion, revise it in their own words, or reject it entirely. They had to review between three and eight changes and then write a reflection explaining what they'd done and why. The assignment forced a kind of slowness. It made students think.
What emerged surprised her. Before the project, students held uneven ideas about AI writing tools. Some could distinguish between generative systems like ChatGPT, which produce content, and assistive tools like Grammarly, which support revision. Others had barely encountered either. Those with prior experience mentioned a handful of familiar names, suggesting that while AI wasn't new to students, knowing when and how to use it effectively remained murky. Students also brought real anxieties into the work: fear that AI would introduce false information, worry that it would flatten their individual voice, concern that using it crossed some ethical line into cheating. Rather than dismiss these worries, the assignment gave students space to confront them directly through hands-on practice and reflection.
Many students found the assistive tool genuinely helpful during revision. They reported clearer grammar, sharper sentences, and patterns in their own writing they hadn't noticed before. Some resisted, calling the suggestions overly simplistic or robotic. But even those skeptics often made meaningful revisions, picking only the suggestions that fit their intentions and discarding the rest. That selective engagement—the willingness to question and choose—suggested something important was happening. Students were developing judgment.
When Welshimer taught the assignment again in spring 2025, she made structural changes based on what she'd learned. She gave students in-class time to work with the tool, demonstrated her own thinking aloud as she decided which suggestions to accept or reject, and moved the AI work earlier in the writing process so it happened before final submission, not after. She also added a survey question asking whether students' concerns about using AI in academic work had shifted. The timing mattered. When AI use was optional and didn't affect grades, students experimented casually. When it became part of the graded writing process, they grew noticeably more selective, treating the tool as something to consult, question, and sometimes refuse rather than blindly follow.
Welshimer's experience points to a broader lesson for instructors navigating AI in their classrooms. Students don't benefit much from strict prohibitions or vague warnings. They benefit from clear expectations, guided practice, and time to reflect on what they've done. When AI is positioned as a partner in learning rather than a threat to be managed, students engage more actively with their own writing and leave better equipped to use these tools responsibly long after they leave the classroom.
Citas Notables
Students were using AI frequently, but not always thoughtfully. The gap was between familiarity and intentional, ethical use.— Sara Welshimer, English instructor at Amarillo College
La Conversación del Hearth Otra perspectiva de la historia
You noticed students falling into two camps—those over-relying on AI and those not using it at all. What made you think the answer was teaching them to use it better rather than just saying no?
Because saying no doesn't actually work. The students who were using AI heavily didn't think they were doing anything wrong. They weren't hiding it. They just didn't understand the difference between using a tool and letting a tool do your thinking. And the other group—they could have genuinely benefited from revision help. Banning it felt like throwing away something useful because some students were misusing it.
So you're saying the problem wasn't the tool itself, but how students thought about it?
Exactly. They had familiarity without judgment. They knew these tools existed, but they couldn't tell you when to use them or why. That gap between knowing something exists and knowing how to use it well—that's what I was trying to close.
Why did you choose Hemingway Editor specifically instead of something like ChatGPT or Grammarly?
Because it doesn't generate content. It presents options. Students have to evaluate each suggestion and decide whether to accept it, revise it, or reject it. That forced the slowness I needed. With a generative tool, students can just accept a full draft and move on. With Hemingway, they have to think.
What surprised you most about how students actually used it?
That even the students who were skeptical—who said the suggestions felt robotic or oversimplified—still made meaningful revisions. They didn't blindly accept everything. They picked what aligned with what they were trying to do. That selective engagement told me something was working. They were developing judgment, not just following orders.
Did positioning it earlier in the writing process and making it graded change how they approached it?
Dramatically. When it was optional and didn't affect their grade, they experimented more freely but also more casually. When it became part of the graded work, they became noticeably more selective. They started treating the tool as something to consult and question rather than something that would fix their problems for them.