Navigating the AI Highway: A Traffic Light Approach to Language Learning

By Maria Laura Mecias, PhD Candidate, Applied Linguistics, University of Florida

DOI: https://www.doi.org/10.69732/FGOG9332 

Introduction

As language educators, we face the challenge of preparing students for a rapidly evolving world where generative artificial intelligence (GAI) is becoming an increasingly prominent tool. In this article, I explore how GAI tools like ChatGPT, Copilot, and Gemini can be thoughtfully integrated into language courses, from beginner to advanced levels, to enhance learning outcomes while preserving academic integrity by outlining clear guidelines for GAI use with students, as well as helping them to learn to craft prompts. 

AI can be an incredibly handy tool for various language learning tasks, from providing grammar feedback to serving as a writing assistant or even helping students practice answering questions. For grammar, students can easily use AI to ask for corrections or explanations on specific structures they’re struggling with, getting instant, detailed feedback that helps them spot mistakes and better understand grammar rules. As a writing assistant, AI can suggest improvements in vocabulary, sentence structure, and overall coherence, helping students polish their work. Similarly for practicing questions, AI can simulate real-time conversations, allowing learners to ask and answer questions in context, boosting their speaking and comprehension skills. Integrating AI into these tasks gives students personalized, on-demand support that complements their classroom learning, leading to more independent and efficient language acquisition.

However, the positive sides of the technology also come along with a whole host of potential negatives. As language educators we all know how frustrating it is to provide feedback to a beginner student, pointing out that their content exceeds the expected level for their class. It’s disheartening to have to note the deduction of points for the inappropriate use of AI. Whether or not you decide to incorporate AI into your language courses, the reality is that these risks are already here. This challenge also presents an opportunity to equip learners with clear guidelines for navigating these tools responsibly. 

The Traffic Light on the AI Highway

The traffic light is a symbol that helps guide us safely and efficiently as we move forward. Its colors serve as familiar signals, allowing everyone to navigate with confidence—even on quiet roads where we rely on the system to steer us in the right direction. This analogy of driving through a city or along a winding road mirrors the experience of navigating the landscape of generative AI. While some might see it as a chaotic and unpredictable journey, there are tools, signals, and systems—like a traffic light—that can illuminate the path, offer guidance, and show us when to move, pause, or stop altogether. These markers help transform confusion into clarity, ensuring a smoother journey toward our destination. As an educator, I see it as my responsibility to set those signals—providing guidance and shedding light for learners as they navigate new technologies that once existed only in the realm of science fiction.

In my own teaching, I’ve made sure to explicitly outline these guidelines in the activity instructions, specifying when and how students may use GAI tools to complete their work, as well as when they should avoid them. To make these guidelines even clearer, I’ve included a stoplight icon in the tasks to indicate the level of AI use allowed. Inspired by Millington’s (2024) proposal for teacher professional development, I thought this system could also benefit my learners. By adopting the traffic light approach, instructors can guide students more effectively by specifying which tasks are appropriate for GAI use. First, we need to discuss in class that GAI tools refer to deep-learning models that can generate high-quality text, images, and other content based on the data they were trained on (IBM Research, 2025). For example, students may use AI to brainstorm ideas about typical food from Honduras but shouldn’t rely on it to generate complete written responses. Similarly, they may use AI to check grammar and refine drafts but not to produce original content. These task-specific guidelines help students understand the boundaries of AI use, supporting their development of critical thinking and ensuring academic integrity.

To provide further clarity, I use the traffic light system to indicate the level of AI use allowed for each task: Red means GAI tools are not allowed, and students must rely entirely on their own skills and knowledge. Yellow means GAI tools are allowed for task preparation, such as brainstorming or researching, but not for language production. Green means GAI tools are fully allowed, including for generating and refining the output. This system helps students understand the expectations and encourages appropriate use of GAI in their learning process. I was pleased to find that the University of Wisconsin Green Bay (2025) has recently proposed the same traffic light system in their course syllabus, and you can also find another article in The FLTMAG that discusses a similar idea! UW-Green Bay also emphasized the importance of academic honesty when using GAI tools, noting that students should be required to cite all sources, including AI-generated content, to ensure transparency and accountability. For example, after a request, ChatGPT can provide a list of sources it used to generate a response, enabling students to verify the original information firsthand. If the activity has a green traffic light, I encourage students to include as many citations as possible. However, citations may not be necessary if it has a yellow light, since generative AI tools were only used as part of the brainstorming process. 

And let’s not forget about data privacy. It’s crucial to ensure that students understand how to protect their personal information when using AI tools. That’s why I’ve made sure data privacy practices are outlined in the syllabus. The syllabus also emphasizes the reflective and strategic use of AI, guiding students on how to use these tools in ways that enhance their academic performance and prepare them for future success. By teaching students to approach AI with both awareness of its strengths and limitations, we’re helping them build skills that will serve them well, whether in their studies or future careers.

Please find here a translation of what is presented in the image above of the course syllabus: 

4. Use of Artificial Intelligence: In this course, students are allowed to use artificial intelligence-based tools (such as ChatGPT) for certain tasks (this will always be specified with a sign) and will also be indicated when it is not allowed. The instructions for each task will include information on when and how these tools can be used to complete it. Activities will include an icon in the instructions indicating the level of use allowed. All sources, including artificial intelligence tools, must be properly cited. The use of artificial intelligence in a manner inconsistent with the mentioned parameters will be considered an academic violation and will be subject to investigation. Students should be aware that the results generated by artificial intelligence may be biased or inaccurate. It is their responsibility to ensure that the information they use from these tools is accurate. Additionally, we will work to ensure the privacy of personal data. Learning to use artificial intelligence-based tools thoughtfully and strategically can help students develop their skills, refine their work, and prepare for their future professional careers.

It’s a learning process, just like language learning itself, and it takes time and effort. However, we must remind students that misusing AI in ways inconsistent with the provided parameters constitutes an academic violation… but even more importantly, they have deprived themselves of an opportunity to further their language learning when they use GAI tools inconsistently with language acquisition.

Happy Prompting!

The traffic light is a clear and graphic way to remind students of the boundaries for GAI use in an assignment. However, there’s another equally important strategy to combine with the traffic light: we must teach students how to craft effective prompts for the bot. If you’re asking the bot to help with grammar feedback, be sure to include details like the student’s proficiency level—beginner, intermediate, advanced—so the feedback is pitched just right. To help students get the best responses from generative AI, I include a short text in the assignment that they must incorporate into their prompts of course, in the target language. This text mentions their proficiency level, ensuring the AI generates responses that align with their linguistic abilities. For example, a beginner should get simpler explanations without complex jargon, while an advanced student might appreciate more detailed feedback. Context matters too. If the task is to correct and provide feedback to an email in a formal tone, the AI should guide the student to use the right vocabulary and structure. Similarly, if they’re working on asking questions, the AI should prompt them with various types of questions that suit the situation (whether casual, professional, or academic). The more specific the prompt, the more personalized the AI’s help will be, which means a better experience for the learner and more progress in their language journey.

For example, when we gave a detailed prompt that included references to the Novice level (ACTFL) and A1 level (CEFR), it helped to prevent vague responses that could potentially discourage further engagement. Without this detail, the questions posed by the chatbot to the learner were lengthy and incorporated low-frequency vocabulary due to the lack of contextualization. Additionally, the conversation included indirect questions, complex syntactic structures, and multi-part commands, which may challenge the learner.

Although the bot attempts to create a natural interaction, the choice of words and syntactic complexity may make the conversation overwhelming for a learner. The lack of adaptation to the learner’s proficiency level can hinder comprehension and fluency.

In contrast, the second interaction incorporated a clear context, allowing the AI to generate more appropriate questions and responses aligned with the learner’s level. The resulting language is simpler, with more frequent grammatical structures and higher-frequency vocabulary. For instance, rather than employing “soler” (to be used to), which is less common at the beginner level, the AI uses verbs like “hacer” (to do) and “tener” (to have), as well as transparent words (cognates) and reinforcing adverbs such as “always” and “after” to help structure the sequence of routine actions. This adaptation enhances accessibility and supports language acquisition more effectively.

But, like with all tools, AI isn’t perfect. For instance, take a look at the image it generated for my syllabus (the traffic lights in the classroom, Picture 1)—it contains a misspelled word! A learner might not even notice these small errors, which can easily be overlooked. But students should keep in mind that AI-generated content can sometimes be biased or inaccurate, especially when it comes to cultural aspects. AI tends to rely on generalized assumptions about countries, traditions, and populations and may miss the rich cultural diversity within communities (Tao et al., 2024). This lack of nuance can lead to oversimplifications, which is not ideal for learners who need to appreciate the full spectrum of the language’s cultural context. One of our challenges is to frequently remind them to analyze GAI as another media resource and critically evaluate the information it provides. Encouraging them to critically evaluate AI output promotes deeper engagement with the material and a more comprehensive understanding of the language and culture. For example, as an individual from Argentina, I demonstrate in class how two different GAI tools—ChatGPT and Copilot—provide distinct responses to the same prompt about the typical dance of my country. While both tools offer similar primary information, their responses differ in key details. Given that many students may not be familiar with the topic, I use this opportunity to foster a discussion about which tool to trust as a user. This debate also serves as a way to highlight the importance of critically evaluating and validating the information provided by these tools.

What´s the typical Argentine dance? For Copilot (above) it is tango, followed by milonga, chacarera and malambo, while for ChatGPT (below) it is tango, followed by zamba and chacarera.  Determining which response is more accurate is challenging because dance traditions are deeply tied to personal experiences and regional contexts. In some regions of Argentina, malambo is more popular, while in others, chacarera is the main folk dance. This variability highlights a crucial limitation of GAI tools—they provide general answers based on available data but lack the cultural perspective understanding that a human facilitator can offer. Educators play a vital role in broadening students’ perspectives, encouraging them to analyze and reflect on cultural practices and products beyond what AI-generated responses can capture. While AI tools can provide useful starting points, they are not yet capable of offering the depth of insight, regional specificity, and critical thinking that human guidance brings to cultural discussions.

AI-Powered Assignments

I’d like to share specific examples of activities that utilize the traffic light method, along with practical strategies to guide learners in creating effective prompts and some details that justify the selection of those particular lights. While these activities are designed for intermediate-level students, they are easily adaptable for both novice and advanced proficiency levels.

Course: Spanish for the Healthcare Professionals
Level: Intermediate High – (Upper Level Division)
Students: L2 and Heritage Language Learners

Example of Green Light: 

Context: In Module 4, we explore topics related to National Linguistic Policies in Healthcare, ensuring patients receive treatment in a language that aligns with their needs. The Spanish course aims to motivate learners to continue developing their speaking skills within the healthcare context, preparing them to become future healthcare professionals and interpreters taking the national exams for accreditation. As a first stage before interpreting, we work on some translation materials so they can start practicing and progressing with short written texts.

Estaciones con Casos: Assignment for Healthcare Prevention Campaign

In this assignment, students are preparing materials for a prevention campaign aimed at the local Hispanic community. The focus is on heart conditions, which are covered in the current module. Each group will work on translating real patient stories into Spanish to create subtitles for the video.

Step 1: Case Listening and Analysis

• Each group will receive a case (YouTube video) involving a patient with a heart condition. The patient will describe their symptoms, treatment, and the evolution of their condition.
• The group will listen to the story in English and analyze the key details, focusing on medical terminology and the patient’s experience.

Step 2: Translation for Subtitles

After listening, the group will translate the patient’s narrative into clear, accessible Spanish to create subtitles for the video. The translation should be written clearly and in a way that is easily understandable to the general public. Students may use a dictionary and some AI tools to assist with the translation. Recommended Resource Linguee for reference.

Step 3: Word Marking and Learning Integration

• Highlight and underline in your translation the words and phrases that you studied in Quizlet, using your group’s designated color for these terms.
• Bold any new vocabulary you are incorporating in your translation for the first time. This will help reinforce your learning and showcase your expanded vocabulary.

Step 4: Feedback Using a GAI Tool

• After completing your own translation, paste your group’s production into a Generative AI (GAI) tool, such as ChatGPT, Copilot, or another one you are familiar with.
• Ask the AI tool to provide feedback on your text, including grammar corrections, accentuation, and any misspelled words. This step will help refine your translation and ensure accuracy. Copy and paste the GAI feedback.

Prompt in Spanish:
“Por favor, revisa mi traducción y corrige cualquier error gramatical, de acentuación o de ortografía. Además, si encuentras que la estructura de las frases no es adecuada, por favor sugiéreme mejoras pero explícame la sugerencia. La traducción es sobre un caso médico relacionado con las enfermedades cardíacas. Asegúrate de que el texto sea claro y adecuado para un público general que podría no tener conocimientos médicos.”

Prompt in English:
“Please review my translation and correct any grammatical errors, accentuation, or misspelled words. Additionally, if you find any sentence structure that isn’t suitable, kindly suggest improvements but give me feedback about your suggestions. The translation is about a medical case related to heart conditions. Make sure the text is clear and appropriate for a general audience who may not have medical knowledge.”

Step 5: Finalizing the Translation and Addressing GAI Feedback

1. Put the group video / participants names. Paste the original translation in the classroom-shared document. This is the text before any revisions based on the feedback from the GAI tool.
2. Paste the GAI-reviewed translation after the feedback from the GAI tool (ChatGPT, Copilot, or another you used). This version should include grammar corrections, accentuation improvements, and any changes suggested by the AI.
3. Final Version with Addressed Suggestions. Paste the final version of your translation, where you’ve incorporated the GAI suggestions that were valid for the group. Include a comment explaining how you addressed the GAI feedback.

This is one of the messages from the mini-group in the shared document, reflecting on their process while using AI (Step 5, C): “We used ChatGPT b/c it gave us more appropriate vocabulary based on the context. It also elevated our language and consolidated the information to make our ideas more complete and sharp. We did not take all of the edits b/c some were not appropriate (like shifting it to informal language). It also undid some grammar. ChatGPT can also be prone to following English grammatical rules, so we took this into consideration when deciding which edits to take. Overall, we took the strengths from our original text and the AI version, and created a stronger paragraph.”

Why Green?

The “green” light emphasizes the free use of GAI. Once the draft is complete, learners can use GAI tools as writing assistants to identify grammatical errors, refine structural coherence, and receive immediate feedback. However, an essential aspect of this step is comparing their original production with the AI feedback to critically evaluate the suggestions provided. This comparison enables learners to make informed decisions about which changes to incorporate, guiding them through a reflective process that enhances their understanding of the tool’s strengths and limitations. As highlighted by studies such as Rahmi et al. (2024), GAI tools enhance grammatical accuracy and coherence, but they do not seem to have a significant impact on other areas, such as students’ lexical density. The facilitator role of the educator in this final stage is crucial, as students can bring questions, concerns, and doubts about specific words or reflect on fossilized errors.

Example of Yellow Light:

Pre Natal Nights: Patient Questionnaire Activity

Context:
In this activity, students will assist patients arriving at the healthcare center by taking vitals and reviewing their clinic history before their doctor’s appointment. The goal is to prepare a comprehensive questionnaire to gather essential information from patients.

Step 1: Create a Preliminary Questionnaire (10-12 Questions)

In pairs, prepare a list of 10-12 questions that you would ask the patient upon arrival at the clinic. You can use online resources (e.g., Mayo Clinic guidelines) or GAI tools (like ChatGPT and Copilot) to brainstorm ideas and ensure you’re covering relevant topics for a prenatal checkup.

Step 2: Choose the Most Relevant Questions

After preparing your list, discuss with your partner which questions are most relevant for the prenatal consultation. Write a comment explaining why you chose those specific questions, considering their importance in evaluating the patient’s condition and preparing for the doctor’s appointment.

Step 3: Compare with Another Pair

Find your closest pair of students and compare your lists of questions. Discuss what’s missing from each list and what could be added. Decide which questions should be discarded based on their relevance and clarity. This collaborative step helps ensure your questions are comprehensive and well-suited for the context.

Step 4: Create the Final Questionnaire as a Group of 4 (15 Questions Max.)

Now that you’ve compared and refined your questions, join with the other pair to create a final questionnaire with 15 well-structured questions. This combined list will serve as the final version for patient intake during the Pre-Natal Nights.

Why Yellow?

The “yellow” light signifies a controlled and deliberate use of AI, where its integration is limited to a specific stage of the activity but is not permitted for the final production. In the Patient Questionnaire Activity, students start by brainstorming and developing their preliminary questions in pairs, allowing for creativity and critical thinking. AI tools are introduced during this stage with other resources to inspire ideas or provide additional input, ensuring that students actively engage while benefiting from technological assistance. However, as the activity progresses to refinement and final production, the use of AI is restricted. This ensures that the final questionnaire reflects the learners’ collaborative efforts, critical evaluations, and authentic understanding of the task. By limiting AI usage to one stage, the “yellow” light emphasizes the importance of balancing technology with human input, fostering independent thinking, and preparing students for professional contexts where they must rely on their own expertise.

Example of Red Light: 

Context: Students in the Spanish for Healthcare Professionals course participate in five virtual meetings with native Spanish speakers to enhance conversational skills and explore healthcare practices in Spanish-speaking countries. These sessions align with course topics and provide real-world insights into cultural and medical practices. After each meeting, students complete a reflection.

English Version:

After the meeting (6 points) 

1. Write a Comparison Paragraph:
   Write a paragraph (250-300 words) comparing what you learned about healthcare in this country with the healthcare system in your own country. As the traffic light image indicates, this is a moment to demonstrate your spontaneous production without assistants, translators, or text generators. This is a learning moment, and as your professor, I’m interested in seeing your genuine effort.
2. Choose 3 New Words:
   Select three new words you learned during the meeting and use them in complete sentences.
   • Example: “I learned that healers used medicinal herbs to treat illnesses.”
3. Cultural Conclusion:
   In your text, try to address what you learned about the connection between culture and healthcare in the country discussed during the meeting.

Why Red?

The “red” light represents a complete restriction of AI use, emphasizing independent, spontaneous language production. In this activity, students engage in a deeply personal and reflective process where they compare healthcare practices between their own country and a Spanish-speaking country. This restriction encourages learners to rely on their own linguistic knowledge and critical thinking, fostering a sense of ownership and confidence in their language skills.

By emphasizing independent production, the “red” sign helps students to remember that learning languages have to do with self-reliance. Ultimately, this approach prepares students for real-world scenarios where they must navigate conversations and professional tasks in the L2 language without external assistance.

Empowering Future-Ready Students

By thoughtfully integrating AI tools into language education, I aim to empower students to thrive in a world where AI is everywhere. Teaching them how to use these tools responsibly not only sharpens their language skills but also prepares them for future careers as innovative and adaptable professionals.  

As language educators, we can use these strategies to create dynamic, engaging learning environments that uphold the core values of academic integrity and critical thinking from novice learners to advanced. I encourage you to consider adopting the traffic light system and sharing its benefits in promoting responsible use of GAI tools. This approach will help raise awareness and guide students in using AI tools effectively. By incorporating it and preparing our students in multiple ways for how to best interact with these tools, we give students the skills and knowledge they need to succeed not only in language learning but also in navigating the ever-changing landscape of their academic and professional lives.

For another perspective on setting AI guidelines using a stoplight, we invite you to read the article Red Means Stop and Green Means Go: Creating AI Guidelines with Students.

References

American Council on the Teaching of Foreign Languages (ACTFL). (n.d.). Assigning CEFR ratings to ACTFL assessments. Retrieved February 1, 2025, from https://www.actfl.org/uploads/files/general/Assigning_CEFR_Ratings_To_ACTFL_Assessments.pdf

IBM Research. (n.d.). What is generative AI? IBM Research. Retrieved February 1, 2025, from https://research.ibm.com/blog/what-is-generative-AI 

Millington, M. (2024) TechLearning. (n.d.). How to integrate AI policies into professional development programs. Published: December,10; 2024. Retrieved January 14, 2025, from https://www.techlearning.com/news/how-to-integrate-ai-policies-into-professional-development-programs

University of Wisconsin-Green Bay. (n.d.). Syllabus snippets. Center for Advancement of Teaching and Learning Blog. Published: January 13, 2025. Retrieved January 14, 2025, from https://blog.uwgb.edu/catl/syllabus-snippets/

Rahmi, R.; Amalina, Z.; Andriansyah, A. and Rodgers, A. (2024). Does it really help? Exploring the impact of Al-Generated writing assistant on the students’ English writing. Studies in English Language and Education. DOI: 10.24815/siele.v11i2.35875

Tao, Y.;  Viberg, O.;  Baker, R.S.;  Kizilcec, R. (2024) Cultural bias and cultural alignment of large language models, PNAS Nexus, Volume 3, Issue 9, September 2024, https://doi.org/10.1093/pnasnexus/pgae346