With China's national college entrance examination, or Gaokao, just completed, many students are using AI to review past questions, while the next group of seniors is considering AI as an all-subject tutor. Simply photographing a problem and asking for an answer offers very little value. The answer already exists in a solution guide. What matters is understanding the problem completely and mastering every related variation. This tutorial explains how AI can genuinely support learning: photograph a problem, ask for a thorough explanation, generate fresh exercises for a weak topic, and review an essay. It also addresses the crucial limits—AI makes calculation errors, its work should not be copied, and it cannot be used during an exam.
Who This Guide Is For, and One Important Principle
This guide serves three groups: high school students preparing for the exam who want to practice more efficiently; parents who want to help but may not know every subject; and teachers who need to create problem variations and review student work quickly. Everyone should remember one principle: AI should help you understand and practice, not do the work for you. Copying an AI answer may save time tonight, but it wastes one of the most valuable opportunities to practice before the Gaokao. You will still be unable to solve the question in the exam room.
One boundary must be explicit from the start: using AI during the Gaokao, a monthly school exam, or another formal test is cheating. Every method below is for ordinary study and review. Define the tool's purpose and limits before using it.
Step 1: Photograph the Problem, but Ask for an Explanation Instead of an Answer
Most mainstream AI products now accept photos and uploaded images. Taking a clear picture of an unfamiliar problem is the first step, but the way you ask matters. “What is the answer?” returns a number that contributes almost nothing to learning. Ask the AI to explain the reasoning instead.
A useful prompt looks like this: “This is a senior-year high school math problem, attached as an image. Do not give me the answer immediately. First identify the concept being tested, explain the key insight, and show why each step works. Then provide the complete solution and answer.” The AI then produces a chain of reasoning you can study. You can identify the step where you became stuck or the formula you failed to recall—the actual gap you need to close. Use the same method in humanities subjects: ask what the question is requesting and how to organize the response rather than memorizing the AI's finished answer.
Step 2: Understand a Mistake Well Enough to Teach It
The value of an error notebook does not come from copying the question again. It comes from understanding the mistake well enough to explain it. AI is particularly useful here. Give it the problem you missed, describe your original reasoning, and ask it to focus on the exact error. For example: “I reversed the sign at this step and got the wrong result. Help me understand why that mistake happens, when this type of sign error usually appears, and how I can avoid it.”
Focusing the AI on your cause of error, rather than requesting a generic solution, is the key. It can turn one mistake into a recognizable category: an unclear concept, a missed condition while reading, or a poor calculation habit. Once the cause is clear, ask for two or three short exercises targeting it and complete them immediately. The test for genuine understanding is simple: can you explain the entire solution without looking at a hint? If you can teach it, you have mastered it.
Step 3: Ask AI to Generate Practice for Weak Topics
After understanding one problem, reinforce the concept with related exercises. This is where AI is more flexible than a workbook: it can generate new questions in a specified topic, format, and difficulty. For example: “Create five Gaokao-level multiple-choice questions on derivatives and monotonicity, increasing in difficulty. Give me only the questions first, without the answers. Explain each one after I finish.”
That creates a private question bank with unlimited variations. Several practices make it more effective. First, ask for the questions before the solutions and complete them under realistic conditions instead of looking at answers as you work. Second, after you finish, have the AI mark correct and incorrect responses, explain only the mistakes, and avoid wasting time on what you already know. Third, periodically request mixed practice that combines several concepts, which more closely resembles a real paper. Remember that AI-generated questions occasionally contain imprecise wording or a questionable answer. When uncertain, verify against the textbook or an official past exam rather than accepting the output uncritically.
Step 4: Use AI to Review Essays and Open-Ended Responses
Chinese-language essays, English writing, and open-ended humanities questions rarely receive immediate feedback during ordinary practice, making them a natural use case for AI. Submit your essay with a specific request: “This is an 800-character Gaokao essay responding to a provided source. Score it on whether the thesis addresses the prompt, structural clarity, strength of argument, and language. Identify concrete weaknesses and give examples of improved sentences, but do not rewrite the entire essay for me.”
The goal is still identify problems and offer examples, not produce a model answer to copy. Treat AI as an instructor who returns feedback immediately. If it says a paragraph lacks substance or an example does not support the argument, revise it yourself and submit the new version to see whether it improved. For English writing, ask specifically for grammar and idiomatic phrasing. Across several rounds, you practice your own expression instead of memorizing an AI-generated essay.
Step 5: Build a Review Plan and Find Knowledge Gaps
Beyond individual questions, AI can help organize a review schedule. Explain how much time remains, which subjects and units are weak, and how many hours you can study each day. Ask for a phased plan broken down by week. Treat the result as advice rather than an order. You and your teacher must decide whether it is realistic and when it needs adjustment.
An even more useful technique is a knowledge self-check. Ask the AI to create rapid questions for one chapter, answer them aloud, and let it identify what you have not mastered. Active recall is much more effective than rereading and quickly exposes topics you thought you knew but did not. Return each weak area to Steps 1 and 2, creating a loop of self-check → repair → more practice.
Common Mistakes and How to Avoid Them
The first and largest risk is that AI makes errors in mathematics and science calculations. A large model predicts text. During a complicated multistep calculation or proof, it can confidently make an arithmetic error or skip a step, producing an orderly explanation with a false conclusion. Verify every step in a science problem yourself. Use AI for hints and reasoning, never as the standard answer to memorize. The second risk is copying answers, which sacrifices the entire practice opportunity and leaves the weakness exposed in the exam room. The third is trusting every generated question and explanation. AI questions sometimes contain defects, and explanations can include factual errors; verify uncertain material against the textbook and official past papers. The fourth is using knowledge from the wrong curriculum. Syllabi and question formats vary across provinces and between the new Gaokao system and older curricula. State the province and curriculum version in the prompt or the AI may apply the wrong general pattern. The fifth point bears repeating: using AI during a formal exam is cheating. Every technique in this guide applies only to ordinary review.
Alternatives and the Best Way to Combine Them
AI is not universal, and several needs are better served by other resources. Official past papers and established study guides remain the standard for authoritative questions and answers; use AI to explain and create variations around them. A structured online course or a teacher is better than fragmented AI conversations for learning an entire chapter from the ground up. For current syllabi, admissions policies, and college-application information, rely on official announcements from the relevant education examination authority. Do not ask AI to recall policy details from memory: its knowledge has a cutoff and may be wrong.
The most practical approach combines resources. Use past papers and textbooks to establish the foundation and standard, courses or teachers for systematic instruction, and AI for immediate questions, detailed review of mistakes, problem variations, and essay feedback—the one-on-one work teachers rarely have enough time to provide at scale. Treat AI as a patient teaching assistant available around the clock and capable of generating unlimited practice, while keeping the authority to judge correctness, set direction, and do the actual work firmly in your own hands. Then it can improve your score instead of creating the illusion that you understand everything.