Build an AI Study Coach in ChatGPT

Create a source-grounded study coach with ChatGPT Projects.

Introduction

30 Second Summary

A study plan can sound convincing while quietly guessing how much you already know. Those guesses make it harder to spend limited study time on the areas that matter.

In this project, you will build a reusable AI Study Coach inside Projects in ChatGPT. Your coach will turn your chosen topic into short learning sessions that adapt to your answers.

What You'll Build

You will open one study workspace that moves you from a source-backed mini-lesson to an adaptive quiz with a clear next action.

By the end of this project, you'll have:

A personalized study plan shaped around your current level, learning goal, schedule, and available time.
A set of source-grounded lessons that teach one concept at a time before asking you to recall what you learned.
An adaptive quiz and feedback loop that identifies your strongest area, weakest area, and next action.
Secret Mission: Create a reusable Progress Snapshot that lets a fresh project chat resume from your weakest area.

Are there any prerequisites?

You need a ChatGPT account with access to Projects in ChatGPT. No coding experience or software installation is required.

Before We Start

A useful study coach needs one clear topic and one specific learning outcome. This commitment gives your AI Study Coach a concrete target that reflects why the topic matters to you.

Set Up Your AI Study Coach Project

Your study brief gives the coach a clear topic. It also gives the coach a specific learning outcome.

A dedicated Projects in ChatGPT workspace keeps its chats, sources, and instructions together. This prevents your study context from being scattered across unrelated conversations.

In this step, get ready to:

Create a project named AI Study Coach.
Personalize the project with a chosen icon and color.
Configure Project-only memory where it is available.

Create the study project

A project gives your study coach one home for its future conversations. Its name makes that workspace easy to find whenever you return.

Switch back to the ChatGPT browser tab from earlier.
Select New project in the left sidebar.
Enter AI Study Coach as the project name.
Choose any icon.
Choose any color.
Complete the creation flow so the project opens.

You should see AI Study Coach in the sidebar. The project should open as your active workspace.

Your dedicated study workspace is now in place. Every conversation you start here has a shared home.

Project missing from the sidebar?

Check the left sidebar for AI Study Coach.
Retry the flow from New project if the project is missing.

Help me find my project.

Configure Project-only memory

Project-only memory keeps this project's conversations separate from chats outside the project. This gives your study workspace a clear boundary.

Open the three-dot menu next to AI Study Coach.
Select Project settings.
Check whether Project-only memory appears under Memory.

✔️ I see Project-only memory

Select Project-only memory under Memory.
Select Save.

You should see Project-only memory shown as the selected memory mode.

ⓧ I don't see Project-only memory

Your account may need its available memory controls enabled before the project-specific option appears.

Open Settings in ChatGPT.
Select Personalization.
Select Memory.
Enable Reference saved memories if your account shows that control.
Enable Reference chat history if your account shows that control.
Return to Project settings for AI Study Coach.
Select Project-only memory if it now appears.
Select Save if you changed the memory mode.

Continue with the available memory mode if Project-only memory still does not appear. The coach remains ready for the explicit context you add in later steps.

Why explicit context still matters

Project instructions define the coach's repeatable teaching rules. Saved project sources supply reusable study material across project chats.

Memory-setting changes can take a few hours. The next steps use those explicit context sources so your workflow stays dependable during the update.

Verify the project setup

Before you verify the setup, pause to predict the project name and memory mode you expect to see.

Switch back to AI Study Coach in the left sidebar.
Confirm the project opens as your active workspace.
Open the three-dot menu next to the project name.
Select Project settings.
Confirm Project-only memory is selected when that option is available.

The sidebar should show AI Study Coach with the icon and color you chose. The Memory section should show Project-only memory where the option is available.

That setup is complete. Your coach now has a dedicated workspace with an explicit plan for reusable context.

Your AI Study Coach now has one home for its future learning conversations. Next, you'll test how a generic request behaves before the coach knows your real constraints.

Test a Generic Study Prompt

Your AI Study Coach workspace is ready inside ChatGPT. Now you can test how much direction a short prompt gives it.

A complete-looking study plan can hide missing learner details. This baseline makes those hidden assumptions visible before you define stronger coaching rules.

In this step, get ready to:

Create a baseline project chat with a generic seven-day study plan.
Inspect the plan for at least three unsupported assumptions.
Capture ChatGPT's assumption list for a later comparison.

Generate the baseline plan

A baseline gives you a real result to evaluate. This request supplies a topic plus a beginner level.

Start a new chat inside the AI Study Coach project from earlier.
Enter your Learner Study Brief topic here: your chosen topic.

Before you send it, which parts of a useful study plan do you think ChatGPT must choose for itself?

Send the completed prompt below.

Create a seven-day beginner study plan for [[TOPIC="your chosen topic"]].

What does this prompt test?

The seven-day phrase sets the plan's duration. The beginner label suggests an experience level.

Your selected topic gives ChatGPT a subject. Every other learner detail remains open.

You'll see a seven-day beginner plan covering your selected topic. Your first visible baseline is now ready for inspection.

Plan Did Not Appear?

Confirm the conversation sits inside the AI Study Coach project.
Send the prompt again if the response stopped before all seven days appeared.

Ask for help with the baseline response.

Find the hidden assumptions

An unsupported assumption is a choice ChatGPT makes without information from you. Finding these choices reveals where the generic plan loses focus.

What Makes an Assumption Unsupported?

A detail is unsupported when your prompt did not provide it. The plan may still sound confident because ChatGPT fills gaps to produce a useful-looking response.

Check how the plan describes your existing experience.
Check which learning goal the plan appears to optimize for.
Check how much daily study time the plan expects.
Check whether the plan treats the seventh day as a firm deadline.
Check which learning style the activities favor.
Identify at least three choices that your prompt did not supply.

That is the key discovery from the baseline. You now have concrete evidence that a polished plan can rest on guessed learner details.

Ask ChatGPT to reveal its assumptions

Your review captures what you noticed in the plan. ChatGPT's self-audit gives you a second view of the same result.

Before you ask, do you expect ChatGPT's list to match the assumptions you found?

Send this follow-up prompt in the same baseline chat.

List the assumptions you made when creating that plan.

Why Ask ChatGPT Directly?

The follow-up asks ChatGPT to examine the choices behind its first response. Comparing that audit with your observations helps you separate explicit constraints from generated guesses.

The response should identify an assumption about your experience. It should identify another assumption about your goal.

It should also identify an assumption about daily study time. It should address the deadline.

It should identify a preferred learning style.

Compare ChatGPT's list with the three assumptions you recorded.
Keep both responses in this AI Study Coach project chat as your baseline.

Assumption List Missing Context?

Confirm you sent the follow-up in the chat containing the seven-day plan.
Return to the baseline chat if you started another conversation.
Send the exact follow-up prompt again.

Ask for help comparing the two responses.

Before the final check, do you expect ChatGPT's own list to overlap with all three assumptions you identified?

Confirm AI Study Coach remains selected in the ChatGPT sidebar.
Scroll to the beginning of the baseline chat.
Confirm the first message requests a seven-day beginner study plan for your topic.
Scroll to the follow-up response.
Confirm the response documents the assumptions behind the plan.

You'll see the generated seven-day plan in the chat. The follow-up response names assumptions about your experience plus your goal.

It also covers daily study time. It covers the deadline.

It identifies a preferred learning style.

Your baseline now shows exactly where a generic study prompt falls short. Next, you'll replace those guesses with a repeatable coaching contract.

Add Personal Coaching Instructions

The baseline plan exposed how quickly ChatGPT fills missing details with guesses. Those unsupported assumptions can push a study session away from your real goal.

Project instructions replace those guesses with a repeatable coaching contract. Every new project chat receives the same learner profile plus the same teaching rules.

In this step, get ready to:

Add a concrete learner profile to the project instructions.
Define a repeatable teaching contract for the coach.
Test the contract in a fresh project chat.

Add the coaching contract

Your learner profile gives the coach five concrete constraints. Your chosen topic plus learning outcome come from the study brief you created earlier.

Think of the instructions as two layers. The learner profile controls what fits your situation.

The coaching rules control the response sequence. They stop the coach from revealing answers before retrieval practice has tested your understanding.

Select the three-dot menu beside AI Study Coach.
Select Project settings.
Complete the learner profile by entering concrete values in the coaching contract below:

You are my practical study coach.

Learner profile:
- Topic: [[TOPIC="your chosen topic"]]
- Current level: [[CURRENT_LEVEL="your current level"]]
- Goal: [[SPECIFIC_OUTCOME="your specific learning outcome"]]
- Session length: [[MINUTES="your session length in minutes"]] minutes
- Deadline or schedule: [[SCHEDULE="your deadline or study schedule"]]

Coaching rules:
- Teach one concept at a time.
- Use a plain-language explanation.
- Include one example.
- Ask one retrieval question.
- Do not reveal the answer to a retrieval question or move on until I respond.
- Label each answer Correct, Partly correct, or Incorrect.
- Explain why.
- Adapt the next question.
- Use project sources when they are available.
- Say when a claim is not supported by those sources.
- End each completed session with Strongest area.
- Include Weakest area.
- Finish with Next action.

How Does This Contract Work?

The role line gives every project chat the same coaching purpose.
The learner profile replaces the assumptions exposed by your baseline plan.
The teaching rules create a consistent lesson structure.
The feedback rules make the coach respond to evidence from your answers.
The source rules require the coach to disclose unsupported claims.

Paste the completed coaching contract into Project instructions.
Select Save.
Confirm that Project instructions shows concrete values for all five learner-profile fields.

That is the coaching contract locked in. Your project now has a concrete profile plus a repeatable teaching sequence.

Instructions Missing or Incomplete?

Confirm that every learner-profile field contains your own concrete value.
Select Save again after your final edit.

Help me check my project instructions

✔️ Awesome, I've got everything!

Your saved project instructions include five concrete profile values. They also include the complete coaching contract.

ⓧ I'd like to double check the full code

Compare your saved instructions with this complete coaching contract.

You are my practical study coach.

Learner profile:
- Topic: [[TOPIC="your chosen topic"]]
- Current level: [[CURRENT_LEVEL="your current level"]]
- Goal: [[SPECIFIC_OUTCOME="your specific learning outcome"]]
- Session length: [[MINUTES="your session length in minutes"]] minutes
- Deadline or schedule: [[SCHEDULE="your deadline or study schedule"]]

Coaching rules:
- Teach one concept at a time.
- Use a plain-language explanation.
- Include one example.
- Ask one retrieval question.
- Do not reveal the answer to a retrieval question or move on until I respond.
- Label each answer Correct, Partly correct, or Incorrect.
- Explain why.
- Adapt the next question.
- Use project sources when they are available.
- Say when a claim is not supported by those sources.
- End each completed session with Strongest area.
- Include Weakest area.
- Finish with Next action.

Run and compare the first session

A fresh project chat tests whether the instructions apply beyond the baseline conversation. The response should follow your teaching contract without needing the full profile in the request.

Start a new chat inside AI Study Coach.

Before you send the request, do you expect the coach to deliver a complete lesson or pause after one retrieval question?

Request the first session by sending:

Create my first study session. Stay within my session length and stop after the first retrieval question.

What Does This Request Test?

The first sentence asks the coach to use your saved learner profile.
The second sentence tests the session-length constraint.
The final phrase tests whether the coach waits for your answer.

Read the response until ChatGPT pauses.

You should see one plain-language explanation followed by one example. The response should end with one retrieval question.

ChatGPT should wait for your answer at that point. You now have a coach that teaches within your chosen constraints.

Switch back to the baseline chat from earlier.
Compare its guessed constraints with the concrete profile used in your first session.
Return to the first-session chat.

The baseline plan spread advice across seven days. Your new session follows one focused teaching loop based on the profile you supplied.

Coach Keeps Going Past the Question?

Check that the request ends with stop after the first retrieval question.
Return to Project settings from earlier.
Confirm that the waiting rule appears in the saved project instructions.

Help me fix the first study session

Your personal coaching contract now turns a broad plan into a focused study session. Next, you will give the coach reliable sources for grounded lessons.

Build a Grounded Source Pack

Your AI Study Coach already follows your learner profile and teaching rules. However, its lessons can still make unsupported claims when the coach has no evidence base.

In this step, you will use Web search in ChatGPT to build a small source pack. Saving that pack gives every new project chat evidence it can cite.

In this step, get ready to:

Generate a source pack from two authoritative sources.
Verify each source against clear quality criteria.
Save the source pack for a grounded lesson.

Generate the source pack

A reliable source pack gives the coach a defined evidence base. Each entry needs enough detail to support a lesson without forcing you to research the topic again.

Enter the topic from your saved learner profile here: your chosen topic.
Start a new chat inside AI Study Coach for building the source pack.
Create the source pack by sending this request:

Search the web for two beginner-friendly, authoritative sources about [[TOPIC="your chosen topic"]]. Create a short source pack with each link, its key ideas, and what it is useful for. Cite every source.

How does this prompt shape the pack?

This request defines the source count and the audience. It also defines the information that every source entry must contain.

The two-source limit keeps the evidence base focused.
The beginner-friendly requirement filters for material that matches your current level.
The authoritative requirement prioritizes publishers with direct expertise in the topic.
The citation requirement makes each supporting source visible.

You should see two source entries with working links. Each entry should include a citation.

You should also see key ideas for each source. Each entry should state what that source is useful for.

How do you check each source?

A beginner-friendly source explains the topic without assuming advanced knowledge. An authoritative source comes from a publisher with direct subject expertise.

A relevant source clearly addresses your chosen topic. These checks keep weak evidence out of future lessons.

Open the first source link.
Evaluate the first source against all three quality criteria.
Return to the source-pack response.
Open the second source link.
Evaluate the second source against all three quality criteria.
Check that both entries contain all four requested fields.

Both pages should cover your chosen topic. Their publishers should also have clear expertise in the subject.

Missing citations or useful sources?

Ask ChatGPT to replace any source that fails the quality criteria.
Ask ChatGPT to complete every missing field in the source pack.

Help me improve this source pack.

Save the response as a project source

Saving the response turns it into reusable project context. This explicit handoff keeps future lessons independent from delayed memory updates.

Open the source-pack response's message menu.
Select Save to project or Add to project sources, whichever label appears.
Confirm the source-pack response appears in the AI Study Coach project's sources.

You have now turned a chat response into reusable evidence for the whole project.

Can't save the response?

Confirm the response was created inside the AI Study Coach project.
Reopen the response's message menu.
Look for either documented save label again.

Help me save this response as a project source.

Test a grounded lesson

A fresh chat tests whether the source pack is genuinely shared across the project. The next request also gives the coach a strict evidence boundary.

Before you send the request, do you think the coach can stay within the saved evidence while still following its teaching format?

Start a fresh chat inside AI Study Coach.
Test the saved source pack by sending this request:

Using only the saved source pack, teach me the first concept and identify which source supports it.

Why does this request work?

The saved-source boundary acts as a prompt constraint. It limits the lesson to evidence you have already checked.

The first-concept instruction keeps the lesson focused.
The source-identification instruction makes source grounding visible in the response.
Your saved coaching rules still control the lesson structure.

You should see one plain-language explanation. A single example should follow.

The lesson should end with one retrieval question. The coach should wait for your answer.

The lesson should name or link the source that supports its first concept.

Lesson missing its source?

Confirm the source pack remains listed in the project's sources.
Remind the coach to use only the saved source pack.
Request the supporting source for the first concept again.

Help me fix this ungrounded lesson.

The grounding loop is working: your coach can now teach from evidence you selected. Next, you will test whether it adapts to strong and weak answers.

Run an Adaptive Quiz

Your grounded lesson now links its teaching to material you checked. That gives the coach a reliable base for retrieval practice.

Next, you'll test whether ChatGPT can react to evidence of understanding. A genuine best-effort answer and an intentionally incomplete answer reveal whether the coach adjusts its feedback.

In this step, get ready to:

Start a five-question quiz grounded only in the saved source pack.
Compare feedback from a best-effort answer with feedback from an incomplete answer.
Close the session with evidence-based learning findings.

Start the source-grounded quiz

A one-question-at-a-time quiz gives the coach a chance to evaluate each response before choosing what comes next. Limiting the quiz to your saved source pack keeps every question tied to the material you checked.

Switch back to the Grounded Lesson Chat from the previous step.
Start the quiz by sending this prompt:

Quiz me with five questions based only on the saved source pack. Ask one question at a time and wait for each answer.

Why One Question at a Time?

The prompt limits the quiz to the saved source pack. The one-question-at-a-time format creates a clear feedback cycle.

Waiting for each answer lets the coach use evidence from your response before choosing the next question. Each response can shape what follows.

You should see one quiz question based on the saved source pack. ChatGPT should wait for your response.

Seeing the Whole Quiz at Once?

Tell ChatGPT to return to one-question-at-a-time delivery if it shows all five questions.
Remind ChatGPT to use only the saved source pack if it introduces outside material.

Help me keep the quiz grounded and one question at a time.

Test the adaptive response

Adaptive feedback depends on differences between your answers. Your first response shows what you understand without changing your normal approach.

Read the first quiz question.
Type a genuine best-effort answer in the message box.
Submit your answer.

You should see the answer evaluated as Correct, Partly correct, or Incorrect. You should also see an explanation based on your response.

How the Evidence Loop Works

The evaluation label gives you an immediate judgment of your answer.
The explanation identifies the evidence behind that judgment.
The next question gives the coach a chance to respond to what your answer revealed.

Read the next quiz question.

Before you submit the next answer, what do you expect the coach to change? Hold that prediction in mind.

Type an answer that deliberately omits one important detail.
Submit your intentionally incomplete answer.

You should see the incomplete answer evaluated with one of the required labels. The coach should adapt its explanation or next question to address the missing detail.

That contrast is intentional. It shows whether the coach responds to evidence instead of delivering a fixed sequence.

Feedback Did Not Change?

Point ChatGPT back to the coaching rules if it skips the evaluation label.
Ask ChatGPT to explain how the next question responds to your previous answer.

Help me test whether my coach is adapting to my answers.

End the session with evidence

The closing summary turns individual quiz responses into a study decision. Its next action should target the evidence collected from your answers.

Before you finish, which answer do you expect to shape the next action most strongly? Keep your prediction in mind.

End the quiz session by sending this prompt:

End this session now and give me my Strongest area, Weakest area, and Next action.

What Does This Prompt Do?

The request closes the quiz before all five questions are complete. It asks the coach to convert your available answer evidence into three practical findings.

The strongest area captures demonstrated understanding. The weakest area directs the next action.

You should see Strongest area, Weakest area, and Next action findings. Each finding should refer to evidence from your answers.

Seeing Generic Findings?

Ask ChatGPT to revise each finding with evidence from your two answers.
Check that the next action focuses on the area exposed by the incomplete answer.

Help me turn generic quiz feedback into evidence-based findings.

You've closed the learning loop. Your coach now uses answer evidence to identify a weak area and recommend a concrete next action.

Secret mission

Build a Reusable Progress Snapshot

Your adaptive quiz identified where your understanding is strongest and where it needs more work. Turn that evidence into a saved Progress Snapshot so a fresh project chat can resume from your weakest area.

Clean Up Your Resources

Your AI Study Coach lives inside ChatGPT without an extra project charge, subject to your plan's normal usage limits. Decide whether to keep it available, pause your use, or delete it entirely.

Resources you used:

One AI Study Coach project containing its memory choice and coaching instructions.
Five project chats covering the baseline plan, first session, grounded lesson, adaptive quiz, and fresh resume test.
Two saved project sources: the Saved Source Pack plus the Saved Progress Snapshot.

Keep everything running

No action needed. Choose this if you want to continue using the coach for future study sessions.

Leave AI Study Coach in the ChatGPT sidebar.
Use the saved Progress Snapshot when you return to continue from your weakest area.

Pause - I'll come back to this later

Stop using the project for now. The project creates no separate metered service or API charge.

Close the ChatGPT browser tab when you finish this session.
Return to AI Study Coach from the ChatGPT sidebar when you are ready to study again.

Delete - I don't want to use this again

Remove the entire study workspace from your ChatGPT account. This deletes every project file, chat, and instruction.

Permanent Deletion

Deleting the project cannot be undone, so choose this only when you no longer need its files, chats, or instructions.

Find AI Study Coach in the ChatGPT sidebar.
Open the menu next to AI Study Coach.
Select Delete project.
Confirm the deletion when ChatGPT asks.

Your cleanup is complete when AI Study Coach no longer appears in the sidebar. ChatGPT has permanently removed the project's files, chats, and instructions.

Nice Work!

You did it! Your reusable AI Study Coach inside ChatGPT now teaches from selected sources within your learning constraints.

You've learned how to:

Expose the guessed assumptions in a generic study plan. Replace them with project instructions grounded in your learner profile.
Build a cited source pack from two authoritative web sources. Reuse it for source grounding in future lessons.
Run adaptive retrieval practice that evaluates each answer. Use the resulting adaptive feedback to identify your strongest area, weakest area, and next action.
Secret Mission: Save a Progress Snapshot that lets a fresh project chat resume from your weakest area.

Ready to quiz yourself?