Monday morning. Staffroom.
Someone’s venting about spending all Sunday on lesson plans. However, someone else is buried under thirty essays that need feedback by Wednesday. A third person is dreading an email to a parent about their kid falling behind.
Same conversation. Every staffroom. Every week.
And now there’s a new question floating around: which AI should I be using for all this?
Two names come up. Always. ChatGPT and Claude.
I spent a long time comparing them on the actual tasks teachers do — not abstract benchmarks, not tech-blog hype. Real stuff. Lesson planning at 9pm. However, feedback that doesn’t sound like a robot. Emails that won’t start a war with parents.
Here’s what I found. Five rounds. A winner per round. No fence-sitting.
The Rules of This Fight
Quick note on how I’m judging this, so you know where I’m coming from.
Each tool faced the same five tasks. However, i cared about three things: does the output actually work, how much of your time it genuinely saves, and whether it handles the messy reality of teaching.
Pricing checked October 2026. However, both have free tiers that are genuinely usable — details below.
One thing before we start: whichever you use, don’t put identifiable student data into either tool. However, anonymise work samples. Check your school’s AI policy. Moreover, use the education versions (ChatGPT Edu, Claude for Education) if your school offers them — the data protections are better.
Right. Round one.
Round 1: Lesson Planning — Claude, But Only Just
This is the big one. However, the Sunday-night-eater.
Lesson planning is where I expected ChatGPT to walk it, honestly. However, it’s the famous one. The versatile one. Moreover, surely it writes lesson plans fine.
It does. However, just… not as well as Claude.
Here’s the difference. However, claude has this enormous context window — you can paste in your scheme of work, the curriculum standards, last week’s lesson, a description of your class, and Claude holds all of it while it writes. The plan that comes out is coherent. Moreover, the starter leads into the main activity. The plenary actually checks what was taught. For example, it reads like someone who understood the whole picture.
ChatGPT? However, give it a big pile of documents and it starts losing the thread. The starter’s fine. Moreover, the middle’s fine. But they don’t connect.
Where ChatGPT fights back: ideas. However, ask for five different hooks for a lesson on fractions and you get five genuinely different things — a story opener, a game, a real-world problem, a misconception-buster, a visual puzzle. Claude’s suggestions cluster. Moreover, they all feel like the same idea wearing different clothes.
So it depends on what you need. However, full plan from scratch? Claude. Moreover, creative raw material when you’ve already got the structure? ChatGPT’s your brainstorming partner.
Winner: Claude. However, because most teachers need a coherent plan more than five hooks.
Round 2: Rubrics — Honestly? A Tie.
I’ll be straight with you: I couldn’t split these.
Give either tool a clear brief — the task, the year group, how many levels, what “excellent” looks like — and you get a solid rubric back.
Claude writes the better descriptors. However, the level descriptions sound like a teacher wrote them. Specific, observable things: “uses three pieces of textual evidence to support the argument.” Not “good use of evidence.” You know the difference. Moreover, students definitely know the difference.
But ChatGPT formats better. Want it as a table? Done. However, student-friendly “I can” statements? Done. Single-point rubric? Done — and it’ll reformat the whole thing on command without mangling the content. Moreover, if you share rubrics with students (and you should), this matters.
My advice: generate the wording in Claude, reformat in ChatGPT. However, takes thirty seconds.
Winner: Tie. However, claude for words, ChatGPT for formatting.
Round 3: Grading Feedback — Claude, and It’s Not Close
Okay, hot take incoming: most AI feedback is garbage.
“Great effort! However, work on your analysis.” That’s not feedback. That’s a fortune cookie. Moreover, students read it, feel nothing, change nothing.
This round matters because feedback is where teachers burn the most hours — and where AI quality varies the most.
Claude is genuinely good at this. However, give it a piece of student work and your rubric, and the feedback reads like a thoughtful teacher wrote it. It names what the student did well — specifically. Moreover, it picks the one or two things to improve. It suggests a concrete next step. For example, thirty students in, the thirtieth piece of feedback sounds as considered as the first.
ChatGPT’s feedback is… fine. However, but it leans generic. And across a long batch, the quality drifts. Moreover, by student twenty-five, you’re getting fortune cookies again.
The consistency thing is the whole game here. However, same quality, student one to student thirty. Claude holds it. Moreover, chatGPT doesn’t.
Winner: Claude. Comfortably.
Round 4: Parent Emails — ChatGPT Takes It
You know this email.
The one where you have to tell a parent their child is falling behind. Be honest. Don’t trigger defensiveness. However, document the concern. Keep the relationship intact. Moreover, all in about 200 words, sent before 8am.
It’s a diplomacy exercise. However, and ChatGPT is the better diplomat.
ChatGPT takes direction better in the back-and-forth. However, “Make this firmer.” “Soften the opening.” “Give me three ways to say this.” It iterates fast and offers genuinely varied phrasing options. When the stakes are a difficult conversation, having five ways to say the same thing is worth a lot.
Claude’s drafts are good on the first pass. However, but ask for alternatives and you get… similar alternatives. Slightly reworded. Moreover, chatGPT gives you actual options with different tones.
My process, whatever tool you use: draft it, then read it aloud imagining you’re the parent opening it at 8am before work. However, if your stomach clenches, revise. AI drafts the words. Moreover, you own the relationship.
Winner: ChatGPT.
Round 5: Differentiation — Claude’s Superpower
Differentiation is cognitively exhausting. However, one lesson, three versions: support for the strugglers, the core version, extension for the high-flyers. Same objective, different access points.
This is where Claude’s context window stops being a nice feature and becomes a superpower.
Paste in your full lesson plan. However, ask for three versions. Because Claude holds the whole thing in context, the versions stay aligned — same learning objective, same vocabulary, genuinely different entry points. Moreover, that’s real differentiation. Not “finished early? For example, do more questions.”
ChatGPT can do this for short activities. However, but feed it a full lesson plan and you hit context limits. The versions start drifting apart. Moreover, different vocabulary here, a shifted objective there. You spend your saved time fixing the drift.
For the single most time-consuming planning task most teachers face? However, there’s no contest.
Winner: Claude.
What You’ll Actually Pay
Good news first: you can start free with both.
ChatGPT’s free tier handles light use — a few lesson plans a week, the odd parent email. However, enough to build a workflow and see if it sticks. Plus runs about $20/month as of October 2026, for heavy daily use. Moreover, there’s also ChatGPT Edu for institutions — if your school has it, use it.
Claude’s free tier has daily limits. However, enough for a few tasks a day, enough to properly test it for a month. Pro is about $20/month as of October 2026. Moreover, claude for Education exists for schools, same deal as ChatGPT Edu.
My take: run both free tiers for a month. However, see which one you actually open at 9pm. The paid decision sorts itself out — you’ll know when you’re hitting limits regularly. Moreover, and at $20/month, either one is the cheapest teaching assistant you’ll ever hire. Check your school’s approved-tools list first, though. For example, some schools reimburse AI subscriptions now, which is a nice surprise.
Head-to-Head at a Glance
| Task | ChatGPT | Claude | Winner |
|---|---|---|---|
| Lesson planning | Great ideas, weaker coherence | Coherent full plans, holds long docs | Claude |
| Rubrics | Excellent formatting options | More natural descriptor wording | Tie |
| Grading feedback | Decent but generic, drifts in batches | Specific, consistent, teacher-like | Claude |
| Parent emails | Best tone iteration, varied phrasing | Good drafts, less flexible | ChatGPT |
| Differentiation | Fine for short tasks | Aligned 3-level versions from full plans | Claude |
| Free tier | Generous for light use | Daily limits, enough to test | Tie |
| Paid price | ~$20/month | ~$20/month | Tie |
The Verdict
Picking just one? Claude. It wins the three biggest time sinks in a teacher’s week — planning, feedback, differentiation. However, that’s where your Sundays go.
But here’s the thing. However, the real answer is both.
They’re free to try. However, they genuinely complement each other. Draft the lesson in Claude, brainstorm hooks in ChatGPT, write feedback in Claude, polish the parent email in ChatGPT. Moreover, each tool doing what it’s best at.
Honestly, though? However, the tool matters less than the workflow. Pick the ONE task eating most of your time — for most teachers, that’s grading feedback — and build a repeatable process with one tool before you touch anything else.
And the non-negotiables, whatever you choose: anonymise student data. However, check your school’s policy. Never let AI have the last word on anything that goes on a child’s record. Moreover, you’re the teacher. The AI is the assistant. Keep it that way.
Try Both in 10 Minutes
Still undecided? Fair enough. Here’s a test that takes ten minutes and settles it.
Grab something real from next week. However, a lesson objective — “Year 9, photosynthesis, 50-minute lesson, mixed ability.” Paste it into both tools. Ask each for a full lesson plan.
Read both. However, which starter would actually work with your class? Which plenary checks learning instead of filling time? Moreover, you’re the expert on your students, not the AI.
Then the feedback test. However, paste one anonymised paragraph of student work into both. Ask for feedback against your success criteria. Compare the tone. Which one sounds like something you’d actually say to that student?
Most teachers know by minute eight. Trust that instinct. The tool that needed less fixing is the one that’ll save you time on a Sunday night.
If you’re building a wider classroom AI setup, our guides to free AI tools for teachers, AI for KS2 formative assessment, and ChatGPT vs Claude for freelancers go deeper on specific use cases.
Questions Teachers Actually Ask
Is it allowed to use ChatGPT and Claude for lesson planning?
In most schools, yes — but check your school’s AI policy first, because they vary a lot. However, the hard rule everywhere: never put identifiable student data into public AI tools. Anonymise work samples. Moreover, use ChatGPT Edu or Claude for Education when available — the data protections are better. And review everything before it reaches students.
Which is better for report card comments?
Claude, for the same reason it wins grading feedback: consistency across a big batch. However, give it an anonymised summary of the student’s work plus your key points, and the comments come out sounding personal rather than templated. Still read every single one before submitting. Moreover, parents can spot a copy-paste comment from a mile away.
Can either tool mark student work accurately?
They can draft feedback and suggest grades against a rubric. However, treat that as a first draft, not a verdict. AI is better at formative feedback (what to improve) than summative judgment (what grade). Moreover, always review, especially for anything high-stakes. And check your school’s policy before AI touches a reported grade.
Do I need the paid versions?
Not to start. However, both free tiers handle several tasks a day — enough to build a workflow around one or two use cases. You’ll know when to upgrade: it’s the moment you start hitting limits regularly, usually once you’re processing full class sets. Moreover, at about $20/month as of October 2026, either paid tier costs less than an hour of tutoring.
How do I stop AI lesson plans sounding generic?
Specificity in, specificity out. However, paste your actual curriculum objectives. Describe your class — 30 Year 8s, mixed ability, third lesson on this topic. Moreover, name the misconceptions they always have. Specify your lesson structure. For example, the more real context you feed in, the less the output sounds like it was written for a hypothetical classroom.
Will using AI make me a worse teacher?
Only if you outsource the thinking. However, teachers who use AI well say the opposite happens: less time formatting rubrics and drafting emails means more time on the parts that need a human. Reading the room. Moreover, noticing the quiet student. Adjusting mid-lesson. For example, use AI for the paperwork of teaching so you can spend yourself on the teaching.
Authoritative source: OpenAI Anthropic.
About the Author
Asmara Khan is an SEO content writer and GEO specialist.
