I Asked ChatGPT to Tutor My Kid. Here's Everything It Got Wrong.

Read time: 11 min  ·  Last updated: June 21, 2026

I asked ChatGPT to create a study plan for my “son” based on a random score report I got from the Internet. I wanted to see how close ChatGPT could come to replicating good advice. If it could, then maybe we should find creative ways to do it more. At least, that was the charitable thought I gave to ChatGPT, even though I knew it was going to be very bad.

And the truth is, ChatGPT is not great. If you’re considering asking ChatGPT, Grok, Gemini, or Claude to create your son or daughter’s tutoring plan, please read through this in-depth article first. I’ll show you exactly what ChatGPT misses (a lot) and why it matters (it will impact your relationship with your son or daughter).

Before I say any of this: I want to point out that I have a “Responsible AI Usage Policy” on the footer of this website. It’s not that I hate AI usage, or that I don’t use it myself. I just believe there are many fundamental instances where using AI is wrong. Using it in the place of a teacher is always wrong. And to prove that to myself and to you, I wrote a whole article to try to show you.

NB: I used an incognito window in Chrome, and then did an incognito session in a logged-out ChatGPT so I wouldn’t bias the LLM with my previous chat history.

The opening read

ChatGPT

Your son is in a solid position overall. A 22 composite with a 25 Reading score means he already has strong comprehension skills.

The biggest opportunity is Math and Science reasoning consistency, plus tightening English conventions. If he improves Math from 21 → 25 and English from 21 → 24 while maintaining Reading, his composite can realistically move into the 25–27 range.

Here’s the important part: the score profile shows skill gaps, not lack of ability. The weak areas are very specific and trainable.

This doesn’t really say anything the parent doesn’t already know. The parent knows the score would increase to a 25–27 if other parts of the score increased. That’s just math. The important part for most parents isn’t that the score report shows skill gaps. The question most parents have is: what’s a realistic score for their son or daughter. And the answer to that always depends on what classes their son or daughter has taken, how long their child has before they need their score, and how motivated their kid is.

The deceptive thing about ChatGPT’s response is that there’s nothing verifiably wrong with it. But there’s nothing good about it. If you take it to your kid, and say look, you can increase your score to a 25 if you improve your subsection scores, their response is going to be: “no crap.” If they knew how to do that, they would’ve already done it. And that’s where these ChatGPT-generated study plans start to get more annoying than helpful.

The score breakdown

ChatGPT

Current Score Snapshot

  • Composite: 22
  • Math: 21
  • Science: 21
  • English: 21
  • Reading: 25
  • Writing: 7

Biggest Weaknesses by Category

Math

Strong:

  • Geometry (88%)
  • Modeling (70%)

Weak:

  • Integrating Essential Skills (38%)
  • Number & Quantity (40%)
  • Statistics & Probability (40%)
  • Functions (50%)

This usually means:

  • careless setup errors
  • weak algebra fluency
  • slower multi-step problem solving
  • inconsistency under time pressure

This isn’t necessarily wrong but it’s also not very helpful. There could be more extenuating circumstances than the above (maybe your student needs accommodations?) or it could be as simple as they got frustrated on one question, spent too long on it, then rushed through the next few questions, progressively spiraling out.

Which is to say, it might not even be a content issue at all — it might be a confidence issue, which we can work on in session. It happens. But an LLM is only trained to tell you the most likely scenario. Nuance is its worst enemy. And humans require more nuance than world-class LLMs have been capable of producing so far in the past 5 years.

ChatGPT

Science

Weakest area:

  • Scientific Investigation (38%)

Moderate:

  • Evaluation of Models & Results (56%)

This tells me he likely struggles less with science content and more with:

  • interpreting experiments
  • comparing hypotheses
  • reading graphs/tables quickly

ACT Science is mostly a data interpretation test.

This is again an okay read — it’s not wrong, but it doesn’t tell us anything. And it misses an important question. First. Yes, the ACT Science section is mostly — almost entirely — a data interpretation test. But this section score usually comes up with the rest of the section scores: after a student has seen English, Math, and Reading, then the Science section falls into place. At least it does in my sessions.

But is the science section even necessary? For most students, the answer is no — here’s when the ACT Science section is actually required. My point is that ChatGPT didn’t even ask you if Science was important. It just assumed because you had taken the science test that it would be a given your son or daughter would take it again.

ChatGPT

English

Weak:

  • Knowledge of Language (43%)
  • Conventions (59%)

This is highly improvable because ACT English is rule-based.

CSE is highly rule-based, and it accounts for ⅓ of the English section. So here, ChatGPT would be misleading you. And if you asked your kid why they aren’t improving on KLA or POW, you’d be doing them a disservice. They’re not rule-based in the way that ChatGPT represents.

If you’re confused about all these acronyms, check out this guide: Reporting Categories on the ACT.

ChatGPT

Reading

Strongest section (25)

Do not over-focus here. Maintain it.

Does the kid like reading? I would actually focus a little bit on this section. Most students are not strong at reading from the outset — so the fact that this student is leads me to wonder whether we can get even more points on this section. Some students take really well to reading strategy.

Target strategy

ChatGPT

Target Strategy

Goal Score Range

Realistic in 10–14 weeks:

  • 25 composite likely
  • 27 possible with strong consistency

A 25 is absolutely realistic in 6 weeks based on the math, reading, and English sub scores. A 27 in 10 weeks is the absolute floor — at these ranges, ACT points are much easier to get. It has to do with percentile points. But I would say if you’re self studying or working with a tutor — either one — you could reasonably expect more progress than ChatGPT is suggesting. As long as you don’t use ChatGPT as your guide.

The weekly schedule

ChatGPT

Weekly Tutoring Structure (Ideal)

Total Weekly Time

  • 6–8 focused hours/week

Not 20 hours. Consistency beats cramming.

Fair.

ChatGPT

Monday — Math Foundations (90 min)

Focus:

  • algebra fluency
  • equations
  • functions
  • word problem setup

Topics:

  • linear equations
  • systems
  • exponents
  • radicals
  • function notation
  • ratios/proportions

Structure:

  • 30 min concept review
  • 45 min timed ACT drills
  • 15 min error analysis

Overall, this sounds fine. But it’s not super helpful. It’s weird how ChatGPT is breaking down the structure into little sections like 15 min error analysis on Monday. I don’t think any student works well like this.

What works better is this: focus on one section at a time, earn the points you can, then move on to the next. While it’s better to have some time in between study sessions, it’s not good for learning to have so much time between sessions. I think people intuitively know this, but when they see an LLM suggest a student should work this way, they decide to go for it because it sounds fine.

ChatGPT

Tuesday — English + Grammar (60–75 min)

Focus:

  • punctuation
  • sentence structure
  • concision
  • transitions

Priority rules:

  • commas
  • semicolons
  • subject/verb agreement
  • pronouns
  • modifiers
  • redundancy

Best method:

He should explain WHY each wrong answer is wrong. That’s where improvement happens.

The priorities are all right. Notice how the Focus and Priority rules drift apart. Transitions is in the focus, and it’s a specific ACT rule, but it’s not in the priority rules. That’s because ChatGPT doesn’t know that Transitions on the ACT follow a specific pattern.

Now the “best method” truly is the best method. Here’s the thing though. Students are so bad at knowing how to explain why each answer is wrong. It’s just a rule of being a student. They are always bad at that. If they weren’t, they wouldn’t need any help. And that’s not a function of being a child or not — it’s a function of learning at any age. So that’s part of the benefit of working with a tutor — but even if you want to try to self study, you can look at my test corrections post.

Doing test corrections the right way is where you’ll see the most gains across the entire test. I won’t elaborate here as I’ve done so at great length on the linked post.

ChatGPT

Wednesday — Science (75–90 min)

Focus:

  • graph reading
  • experiment comparison
  • data interpretation speed

Drills:

  • 5–7 minute passage timing
  • identifying variables quickly
  • hypothesis comparison

Key habit:

NO rereading entire passages. ACT Science rewards scanning.

Why are we even studying science? We have never established this.

ChatGPT recommended 5–6 minutes reading each passage. There’s not much to read on most passages. And trying to read the small intro paragraph will just confuse the students more than not reading the passage. It’s different when the passage is conflicting viewpoints.

Here’s an example ACT science passage:

Example ACT Science passage with a data-representation graph and table

How is your son or daughter supposed to spend 5–6 minutes reading this? I recommend most students spend 0 minutes reading the passages. The questions are all “linear” in the sense that the questions actually build familiarity with the science passage. So the first three questions for example will establish what are the x and y axes, what are the titles, what are they keys? And from there, the next few questions will ask the students to compare results now that they’re familiar with what each sections means.

So asking the student to read the passages is awful advice for almost all students.

Here’s a very important caveat: conflicting viewpoints passages. The science section will ask one conflicting viewpoints passage, where you have to read through the different viewpoints regarding a single experiment, then answer questions that compare and contrast different perspectives.

Example ACT Science conflicting viewpoints passage with two competing perspectives

This is a passage type that I would strongly recommend most students read before answering the questions. It’s nearly impossible to answer all of the questions without knowing what each perspective means. And doing so in time without reading is impossible.

ChatGPT

Thursday — Math Timed Section (60–90 min)

One timed 60-question math section.

Then:

  • review every missed problem
  • categorize mistakes: concept gap, timing issue, careless mistake

This matters a lot. Students plateau because they review answers emotionally instead of diagnostically.

Yeah. True enough. Most students are going to look at a mistake, explain it to themselves in their head, then say they understand. That’s why making them write out detailed explanations is the number one way to ensure they understand. Try it — you’ll be amazed at the results.

ChatGPT

Friday — Reading + Writing (60 min)

Reading:

  • maintain pacing
  • inference questions
  • dual passage strategy

Writing:

  • 1 essay every other week
  • focus: structure, evidence, clarity, transitions

He does NOT need elite creative writing. ACT writing rewards organization.

Terrible advice. Why are we doing the writing? Is your son or daughter applying to one of the few colleges that require the writing section? No? Then… why are we doing this section?

Next, yes, these are seemingly good places to focus on for reading improvement. But really, the reading section has a key strategy that helps almost every student improve their score.

ChatGPT

Saturday — Full ACT Rotation (2.5–3 hrs)

Alternate:

  • full practice test
  • OR 2 timed sections back-to-back

Goal: build endurance and pacing.

Terrible advice. If a student doesn’t know the material, they won’t be able to produce it in time. So don’t waste your kid’s time doing practice tests until they know the material.

A study by the College Board found that students given extra time only answered about 1 question more correctly. The purpose of that study was to understand whether extra time was an “unfair” advantage. And the answer is a conclusive: no.

The reason why that’s important here isn’t because your son or daughter necessarily needs accommodations. It’s because the point is a good one. If your son or daughter doesn’t know the material, even extra time won’t help them get a higher score. Nothing improves their score other than knowing the material.

Now, a note on the above program. On the low end, this adds up to 8.25 hours — not 6–8 focused hours that the LLM recommended above.

Out of full transparency, I’m going to include the next section of ChatGPT’s response. I feel like I’ve responded to all the important points above, so I’m not going to break down the following subsections. There is some information below this, however, I think is good.

Section-specific recommendations

ChatGPT

Section-Specific Recommendations

Math Priority Order

  1. Algebra Fluency — Most important. He should be able to solve basic equations almost automatically. Daily drill examples: solving equations, simplifying expressions, factoring, function evaluation.
  2. Statistics & Probability — Easy points once practiced. Topics: mean/median, probability, combinations/permutations, interpreting tables.
  3. Integrating Essential Skills — This category mixes concepts. Best fix: mixed timed drills.

Science Strategy

This section can improve FAST. Most students incorrectly study science facts.

Instead train:

  • graph interpretation
  • trend spotting
  • experiment structure

Key shortcuts:

  • read questions first
  • identify axes immediately
  • compare variables systematically

A jump from 21 → 25 here is realistic.

English Strategy

High ROI section.

Memorize:

  • punctuation rules
  • sentence boundaries
  • transition logic

The ACT repeats patterns constantly.

He should maintain an “error notebook” with:

  • missed question
  • rule violated
  • corrected version

That accelerates retention dramatically.

ChatGPT

Recommended Resources

Best Overall ACT Prep

  • Official ACT Prep
  • Khan Academy SAT/Grammar Resources (great for foundational math/grammar even though SAT-focused)

Best Math Resource

  • The Organic Chemistry Tutor YouTube Channel

Best Science Strategy Resource

  • Magoosh ACT Science Guide

Best Grammar Resource

  • Purdue OWL Grammar Resources

These are pretty terrible recommendations except for the first one. The Official ACT Prep book is great — buy it or get it from your local library.

Khan Academy has in the past given incorrect information about the SAT — misteaching a rule, saying the equivalent of 1+2 = 16235 when in fact it does not. The grammar on the SAT is also structured quite differently. Use my website instead.

The math resource isn’t as terrible as it seems at first. The Organic Chemistry Tutor YouTube does teach math — but honestly, Khan Academy is a better resource for math. The Organic Chemistry Tutor is overkill.

Magoosh science is fine — you can use that as a free website if you don’t want to use mine. But the ACT Science prep book is actually great. If — you need your son or daughter to study the science section.

Grammar, I’m going to firmly disagree with Purdue OWL as a good resource. It’s good for grammar concepts, but it’s not structured around what kids need to know for the ACT English section. So their studying is going to be inefficient, somewhat ineffective, and overwhelming. Use my website or any other website first.

Tutor guidance

ChatGPT

Tutor Guidance

If hiring a tutor:

  • prioritize ACT-specific experience
  • avoid general academic tutoring only
  • insist on timed strategy work

The right tutor should:

  • track error trends weekly
  • assign targeted drills
  • teach pacing explicitly

Not just “help with homework.”

This isn’t bad. But timed strategy work is separate from any of the content. Learning the content is the first part of timed strategy and pacing. Then strategy (which varies from section to section) helps with pacing. Expecting your tutor to focus on “timing” because ChatGPT told you to is a recipe for disaster.

The “mindset shift”

ChatGPT

Most Important Mindset Shift

Right now, his scores suggest: he understands more than he consistently executes under ACT conditions.

That’s actually encouraging.

The fastest gains usually come from:

  • timing strategy
  • error analysis
  • repetition of predictable question types

—not from trying to “get smarter.”

If you’d like, I can also help you with:

  • a detailed 12-week calendar
  • daily homework assignments
  • recommended score goals by month
  • custom drill packets by weak area
  • a parent tracking spreadsheet
  • a plan for raising Math specifically into the 26–28 range.

It’s true your son understands more than he executes under ACT conditions. And that is encouraging. So the rest of section might seem tempting to believe, given that ChatGPT has understood your concern.

But timing strategy IS learning. Error analysis is the area that most teens — and people — struggle to execute alone. And repetition of predictable question types is kind of a nothing-burger. ALL ACT questions are predictable.

So if you want a tutor who won’t use ChatGPT to run sessions, let’s work together.


We use cookies on our site. Learn more.
Chat on WhatsApp