AI is reshaping the field of education, from triggering educational anxiety and prompting educators to explore new tools to promoting the reconstruction of the education evaluation system, while the business collision of education and AI has also brought new opportunities.
Traditional education is being effortlessly penetrated by AI.
According to data from the 2024 Digital Education Council, more than 20% of students use AI every day, and more than half of them use AI every week. In an earlier set of data, 89% of college students in the United States are already using ChatGPT to write homework.
When AI is revolutionizing traditional learning methods, a large number of AI detection tools have begun to become popular in order to defend the original rules. As a result, AI detection tools with millions of users appeared, such as GPTzero and Copyleaks.
Using AI to defeat AI is the most real game between AI and education in the two years of collision.
Rather than the controversy over changes in education methods, the flow of funds reflects the profound impact of AI on the education industry.
Roy Lee, a foreign post-00s brother, made $2 million in 50 days using a programmer interview cheating tool “InterviewCoder”, and recently received $5.3 million in investment for this project; AI detection tool GPTZero also received a $10 million investment in June 2024.
In the capital market, from the beginning of last year to the present, education leader Duolingo has also relied on the “ALL in AI” strategy, and its stock price has soared by 132%, significantly outperforming the Nasdaq index.
Behind all this, the education industry is undergoing a deeper change – education is transforming from the traditional “indoctrination of knowledge + detection of memory” to “inspiring understanding + empowering cognition”.
And all opportunities for AI in the education industry will open up here.
01 Accidentally injured by an AI detection product! Students stage self-proof “I am me”
Two years ago, the debate over whether AI writing subverted academic integrity was still in turmoil and intensifying.
University of Houston student Leigh Burrell submitted 15 pages of screenshots and timestamps of the writing process for his self-certification English assignment, and even set up a camera to record real-time creation footage.
In June 2024, a student at San Jose State University posted on Reddit that their graduation project completed in Visual Studio Code was “too standardized” due to the code style, and half of them were accused by professors of “AI doing”, and he himself could not prove himself even if he submitted a development log.
On Reddit and TikTok, there are many cases of accidental injuries by AI detectors like this.
▲ A student’s homework at the University of California, Davis was mislabeled as AI-generated by Turnitin (Source: reddit)
The teachers’ doubts are not unreasonable, and from the popularity of question-making software at home and abroad, we can see how much students demand “copying answers”. AI makes cheating easier, and you can get a Mathematical Olympiad question, a historical essay or literary analysis in just a few seconds.
Programming is the hardest hit area, and AI tools can assist interviews and homework in real time by simply entering prompts. (Can AI teach people to interview?) With just one sentence on the resume, the offer has quadrupled)
These days, a 2023 “professor scolds AI homework” posing video suddenly exploded in China. “The professor should reflect on why he can only assign homework that AI can complete”, “My son’s homework is all done by Doubao, and I just finished beating him!” ”…… The video is full of passionate comments.
The video is fake, but the educational anxiety under the impact of AI is real.
Traditional methods of dealing with interview cheating are failing under the attack of AI models, and the new generation of AI detection technology is seriously lagging behind.
To achieve these three challenges, product managers will only continue to appreciate
Good product managers are very scarce, and product managers who understand users, business, and data are still in demand when they go out of the Internet. On the contrary, if you only do simple communication, inefficient execution, and shallow thinking, I am afraid that you will not be able to go through the torrent of the next 3-5 years.
View details >
AI detectors such as Turnitin and GPTZero frequently experience “misjudgment disasters”, frequently misjudging original content as “AI-generated”. This is because the boundaries between AI-generated content and human originality are increasingly blurred, and its text randomness and grammatical structure have converged with human writing characteristics, while existing detectors rely on pattern recognition, which is just stepping on the similarity between the two.
AI detection tools are not useful, and students have to prove that “I am me”. The new skills that these students came up with include: using historical records to track editing trajectories, using screen recording software to record the thinking process, and even deliberately adding “human error” to the code.
In order to face the impact of AI, educators have begun to accelerate the iteration of AI tools, from “detecting AI” to “defining originality”. EdTech companies begin developing “two-way verification” tools:
- The WriteSonic plug-in can automatically record the user’s modification logic with AI assistance and generate a “proof of originality report”;
- Some laboratories are testing “cognitive fingerprinting” technology to distinguish between “human creation” and “AI generation” by analyzing biometrics such as keyboard typing rhythm and thinking pause duration when writing.
If AI detection tools are backward at the technical level, then standard blurring is a greater contradiction at the human cognitive layer. Teachers and students have different opinions on whether to use AI.
On the one hand, more and more teachers see AI as a productivity tool. Last year, a national survey of more than 1,800 college teachers showed that 18% described themselves as frequent users of AI. In February 2025, California State University reached a partnership with OpenAI to provide ChatGPT Edu to approximately 500,000 teachers and students across 23 campuses.
On the other hand, the double-standard behavior of “professors disabling AI but using it themselves” has triggered people’s thinking about educational fairness. Ella Stapleton of Northeastern University in Boston sued the professor for $8,000 in tuition after discovering that she had used ChatGPT to generate courseware; A professor at Southern New Hampshire University was found to have used AI to correct assignments and not read students’ papers.
Perhaps realizing that the system cannot outperform the technological iteration, some educators have turned to encourage students to engage in “human-machine collaboration”. At the same time, the educational assessment system is also being rebuilt, examining “thinking traces” and requiring students to clearly mark AI components:
- American University Business School requires freshmen to use AI and record writing ideas in real time in Google Docs;
- Kennesaw State University requires students to submit original drafts and revised records, and to include human-robot collaboration skills in the assessment;
- some colleges and universities have returned to “on-site handwriting”, using old-school forms such as oral exams and blue book compositions;
- Some schools require students to attach writing drafts and “idea recordings” and other documents to their homework;
AI is forcing education to return to its essence – not to train “knowledge storage”, but to forge a “thinking brain”. This process requires educators and students to explore together and find a new balance between technology and humanity.
02 AI speaking education rose 43% throughout the year, and AI question search software earned 10 million yuan a year to beat competing products
There is no doubt that AI has reshaped the technical logic of educational products. The collision of education and AI has also moved towards real money business practice, and efficiency has become the mainstream of “AI + education”.
Players in the four tracks of language learning, skills training, cheating and anti-cheating, and AI problem-solving are proving that they have taken the lead in running the business model with tens of millions of users and revenue of more than 100 million yuan.
(1) AI speaking education payments are strong, with Duolingo’s annual revenue exceeding $700 million
The global AI education market will exceed 80 billion yuan in 2024 and is expected to increase to 120 billion yuan in 2025. Language learning products dominate, with AI language learning applications accounting for more than 60% of the world’s leading education AI application revenue.
Language education users are willing to pay strongly, and many AI speaking companies have taken off. AI Speaking Education Duolingo’s revenue in 2024 will reach US$748 million, an increase of 40.8%, with more than 10.3 million paying users and a payment rate of 8.9%. Looking at the longer cycle, Duolingo stock price has performed well in 2024, rising 43% for the year, surpassing the Nasdaq’s 29% gain.
Its high-end AI capabilities drive an increase in customer unit price, and the product provides GPT-4-powered role-playing and video call functions, driving ARPU to $18.5 per quarter.
In the field of AI+ education, the AI oral education track is quite mature and prominent, and other AI products are:
1) Speak, a unicorn whose product focuses on pronunciation correction, has a monthly revenue of US$4.19 million, accounting for 70% of the Korean spoken market
2) The AI speaking sparring product Praktika has an annual revenue of $20 million, ranking 12th in the U.S. education list.
3) Mondly, an AI multilingual education product, has an annual revenue of $9.8 million and is popular in the offline model.
(2) Mindstone uses “AI to learn AI” and earns 200,000 yuan per month as an education SaaS
Mindstone is a company that provides AI education SaaS services for skills training institutions, with monthly revenue exceeding $200,000 in 2025, with customers covering more than 200 enterprises and institutions. Mindstone enters the market by providing practical AI training for non-technical personnel, and cooperates with large organizations such as Hyatt and Pearson to achieve large-scale monetization through an enterprise payment model.
In terms of product logic, Mindstone focuses on the 10-hour “AI Capability Plan”, which adopts the “real-time + asynchronous” hybrid teaching mode to disassemble complex AI skills into scenario-based practical content, such as custom GPT, prompt engineering, etc., to help users master AI tool applications in a short period of time. Technically, the platform is built based on Google Cloud, and AI tools such as Replit are gradually introduced to improve the teaching experience.
(3) The AI cheating artifact exploded, and the “Columbia University dropout” raised $5.3 million
At the end of April, the programmer’s interview cheating tool “InterviewCoder” became popular on the Internet, earning $2.2 million in 50 days. The company behind it, Clauly, also recently received $5.3 million in financing, and the company’s annual revenue has exceeded $3 million, and founder Roy Lee has also become the top of Generation Z by relying on his experience of going to court with Columbia University and being expelled from Columbia University.
The guy directly admitted that almost every interview would use AI tools such as ChatGPT to cheat through “InterviewCoder”, and he had successfully obtained offers from TikTok, Amazon, Meta and other companies. (This post-00s brother, selling AI cheating software, earning tens of millions in 50 days)
(4) 3 million new users in 6 months, and AI detection products such as GPTZero exploded
GPTZero is one of the first AI detection products to become popular. It was launched in January 2023, with more than 30,000 visits in the first week, causing the server to be paralyzed, and by July 2024, the number of users has exceeded 4 million. The product uses a large number of human writing and AI-generated text corpus training to comprehensively analyze language characteristics and subtle differences in writing style, and accurately judge the source of the text.
GPTZero completed a $10 million Series A funding round in June 2024. Today, the customer base has expanded from the initial teachers to government procurement agencies, charities, hiring managers, data annotators, etc.
Other AI Detection Products:
1) Turnitin is a long-established plagiarism checking product, which is used by more than 15,000 schools and institutions in nearly 170 countries around the world, with nearly 30 million active teachers and students. Turnitin mainly compares user-submitted manuscripts with massive global databases and web content, and now uses AI technology to quickly generate “originality reports” and target non-original content.
2) Copyleaks is a product that detects plagiarism and AI-generated content, and it can detect AI-generated source code even if it is obfuscated or modified. The company claims to have millions of users. The company has raised more than $8.4 million in total financing.
(5) The AI topic product Photomath has an annual income of $35 million and a monthly active income of 50 million
Photomath is a phenomenal AI question shooting product under Google, with annual revenue of $35 million and more than 50 million monthly active users. It allows users to take pictures of math problems with their mobile phones to get answers and detailed problem-solving steps, and can solve homework in fields from basic arithmetic to advanced calculus.
At a time when online education is hot, many AI search products that have gone overseas have also achieved remarkable results:
1) ByteDance’s Gauth focuses on mathematics tutors, with a user base of 200 million, and has topped Apple’s free education list in the United States;
2) Question under Homework Help focuses on general practice coverage, with 596,000 dual-end DAUs and 26 million downloads worldwide;
3) Ape Tutoring’s Uknow.AI has an overseas layout of more than 100 countries, with nearly 10 million active users throughout the year;
Students pursue “instant answers”, educators rely on “batch processing”, and deep thinking is squeezed by efficiency.
Looking back at the node of 2025, those products that break through in terms of revenue list and user scale may no longer be simply “AI education tools”, but industry pioneers who are defining the next generation of educational infrastructure.
