How to Turn Any Article or Video into a Language Lesson with AI

Finding great authentic content may be the easy part. A news article about automation lands in your inbox. A viral YouTube video is exactly what your B2 adult learner would connect with. A pop song your teenage student already knows is sitting right there on Spotify. The bottleneck isn't the material, it's the 30 to 45 minutes of extraction work that turns a great text into a lesson with vocabulary, comprehension questions, and a grammar focus. AI removes that bottleneck, and this post walks you through exactly how.

Why Do Teachers Keep Reaching for Authentic Materials?

Authentic materials work because they give learners language in the context it actually lives in. A textbook unit on "business travel" gives you clean, controlled sentences. A real New York Times article about business travel gives you hedged language, idiomatic expressions, subordinate clauses that no textbook writer would ever produce, and a topic the learner probably has an opinion about. That opinion matters. When a learner has a stake in the content, they read more carefully, look up words they actually want to know, and show up to the next session ready to talk.

The underlying principle here goes back to what Stephen Krashen described as comprehensible input; language acquisition happens when learners encounter real language just above their current level, in a context where meaning is clear enough to follow. Authentic texts do this in a way that synthetic textbook material rarely can, because the language patterns are real rather than engineered.

None of this is new knowledge to experienced teachers. The practical problem is time. You find a genuinely great article on Monday afternoon, and by the time you've extracted vocabulary, written comprehension questions, identified the grammar point, and formatted everything, you've spent longer on the worksheet than the lesson itself.

What Happens When You Add AI to That Workflow?

Adding AI to the process changes the math. Instead of spending 40 minutes building exercises from scratch, you paste the text into a prompt, specify the level and exercise types you want, and get a first draft back in under a minute. You still review it, adjust the vocabulary selection, maybe rephrase a comprehension question, but that takes 5 minutes, not 45.

The key is the prompt structure. Vague prompts produce vague output. "Turn this into an ESL lesson" will give you something generic. A structured prompt that specifies level, target vocabulary count, exercise types, and output format gives you something you can actually use. Try something like the below

You're an English language teacher and must use the below text with a learner.
Please create the following exercises from this text:

- A vocabulary list of 10 words or phrases the learner should know before reading (target level: [B1/B2/C1]). For each word, include: the word, its part of speech, a short definition in simple English, and one example sentence drawn from or inspired by the text.
- Five comprehension questions that test understanding of the main ideas. Mix factual and inferential questions.
- One grammar focus task: identify a specific structure that appears in the text (e.g., passive voice, present perfect, reported speech), explain it briefly, and give the learner 3 practice sentences to complete using that structure.

Here is the text:
[PASTE TEXT HERE]

Say your learner is a project manager at a tech company, working at B2 level, and you've found a short article about how companies are using AI to automate internal reporting. You paste the article into the prompt, replace [B2] and [PASTE TEXT HERE], and hit send. 

You can add additional instruction for the vocabulary, for instance to also show translations or explanations in the learners native language if you find it relevant. You can adjust and give additional instructions on the comprehension exercises or grammar task. You can also add instructions on other specific exercises you'd like. The output should be ready to use, but you may swap out a vocabulary item that's too obscure or simplify a comprehension question, but the heavy lifting is done. 

Does This Work for Videos and Songs Too?

It works just as well, because the prompt doesn't care what the source was only what text you give it. The method is the same, get the text, paste it into the prompt, and let the AI generate the exercises.

For YouTube videos, the built-in transcript is the fastest route. It typically involves expanding the video description, scrolling down and looking for a "Show transcript" button or link. On desktop after clicking the link it appears in the right side. There is an option to remove the timestamps by clicking a three-dot menu and toggle the timestamps, which you should use before copying the text. If you can't make it work or Youtube changes it, try searching a search engine for "youtube transcription". Besides a hopefully updated description, it will also show a lot of services that offer the same by just pasting the Youtube URL for the video.

For song lyrics, the text is usually a few clicks away on a lyrics site. Songs tend to be shorter than articles, so you may want to adjust the vocabulary count down and keep the comprehension questions focused on interpretation rather than plot. Grammar focus tasks may work well with some song lyrics because artists make interesting, deliberate structural choices that may give learners something real to examine.

Are There Things to Watch For?

AI vocabulary selection can run academic. The model sometimes gravitates toward formal or literary vocabulary, even when your learner needs professional or conversational register. Review the vocabulary list with your learner's context in mind and swap out anything that feels off. You can also add instruction for the domain and register of the vocabulary to extract.

Comprehension questions can be too literal. "What did the author say about X?" is a weaker question than "What does the author imply about Y, and do you agree?" If the questions all feel like a reading quiz and you want something more like a discussion springboard, add one inferential or opinion-based question of your own, or adjust the instruction further.

Grammar tasks occasionally identify a structure that barely appears in the text. If the prompt returns a grammar focus that feels like a stretch, ask AI to try again: "The grammar task you suggested (conditionals) only appears once in the text. Please suggest a different grammar structure that appears at least 3 times." Or instruct on the specific you wanted it to use.

None of the things to watch for are reasons to avoid the method. They're reasons to spend 3-5 minutes reviewing the output rather than trusting it blindly. The review is still a fraction of the time it would take to build the exercises from scratch.

Quick Reference: Prompt Templates to Get Started Today

The below table sketches some variations for inspiration to use different sources and adjust the prompt to get different types of exercises.

For a wider set of prompts built specifically for language teachers, the post 7 ChatGPT Prompts Every Language Teacher Should Know covers related territory with copy-paste prompts for vocabulary, grammar, and dialog generation.

Turn Good Material into a Good Lesson

The gap between "I found a great text" and "I have a lesson" is almost entirely a time problem, not a creativity problem. Most teachers already know authentic content works better than textbook material with their learners. The question has always been whether the preparation time is worth it. AI makes it worth it, because the time cost drops significantly.

The prompt in this post is one you can use today, with any article, video transcript, or song. Try it once with a text you've already been meaning to use. The kind that's been sitting in a browser tab for two weeks because you haven't had time to build exercises around it.