Daniel Gomez teaches humanities at Ideal School in Tampa. He speaks to bilingual students. He sees AI as a friend.
“Teachers can’t know everything.” Not in every subject. Definitely not in multiple languages. AI fills those holes. It grabs a topic the kid loves. It handles it. “Reasonably well.”
Harvard researchers agree. Ying Xu, an assistant professor there, sees opportunities everywhere. Adapt materials into new languages? Yes. Expand access for bilingual learners? Absolutely.
Some projects try something interesting. Keep the core lesson in English. Give explanations in the kid’s home language. Scaffolding, but multilingual.
It has potential. Big potential. If done right.
Chatbots Do the Heavy Lifting
Krystle Salas runs Second Mile charter schools. At-risk kids. At-risk futures.
“Bilingual learners learn the culture and language of the classroom,” Salas says. They miss the conversations. The instructional materials slip through the cracks. Teachers are tired. They don’t have the tools to bridge the gap constantly.
Enter the bot.
It hands out pieces of content in Spanish. Simple sentences. Broken down.
Then there’s Melissa Henning from The Source for Learning. She talks about confidence.
“Maybe I’m shy,” a student thinks. “I won’t speak English in front of everyone.” The bot doesn’t care. It doesn’t judge the accent. The grammar? Maybe messy. But the bot lets you try.
“You rephrase. You practice.” Henning calls it empowering. Gomez agrees. The kid gets the concept first. The vocabulary in English comes later.
Why wait?
AI is always on. 24/7. Practice requires activity. The bot provides it.
Stephanie Howell uses SchoolAI in Ohio. Second grade. Third grade. Intervention work.
Students treat AI as a “guide on the side.” They read. They hit a word. They ask, “Explain that differently.” Self-awareness builds. Howell steps back.
She programs the bot, too.
“Miss it once? Make them struggle. Miss it twice? Give them a sentence starter.” Productive struggle matters.
Howell also asks AI to analyze texts for her. Which vocabulary will trip them up? What patterns emerge? It gives her time. Time to teach. One on one.
Home Is Not a Monolingual Zone
The classroom ends. The homework starts.
Parents speak Spanish. Kids bring home English assignments. The silence grows.
Henning sees this often. Kids speak better English than their moms or dads. It’s a barrier. Or it could be an asset.
Chatbots translate simultaneously.
“Ask your child about the character’s brave move,” the teacher suggests via the bot. The parent hears the question. The kid answers. Dialogue happens.
The home language becomes strength, not a weakness.
Be Careful Here
Don’t get reckless.
Xu warns of widening gaps. Many AI systems are bad at accented speech. Dialects confuse them. Code-switching breaks them.
“Bilingual learners benefit less if the system doesn’t hear them.”
Assessments fail, too. Accented speech gets marked wrong. Inaccurate evaluation.
Then there’s the hardware problem. Do they have internet? A device? Henning asks if we are just leaving people behind again.
Howell manages the languages carefully. Students type in Spanish unless forced otherwise. But she knows the goal.
“They need to learn English,” she says. Back and forth doesn’t help if the destination is fluency in English. Intent matters.
Training counts, Gomez notes. The model needs technical words? Fine. Everyday language? Easy. Customize it.
But remember this.
Teachers lead. Always.
“AI is valuable.” Yes. But it misses emotions. Context. Salas at Second Mile stresses the human need. A bot sees data. A teacher sees a child.
AI makes mistakes. Humans adjust in real time.
Who holds the room? The teacher. The bot is just a tool.
