Chegg
verticalized AI assistant built as a wrapper on a third-party LLM
After ChatGPT launched, Chegg saw its subscribers fall from about 7.8 million to 3.2 million and its Q1 2025 revenue drop 30 percent; its CheggMate assistant built on GPT-4 created no barrier against the free model.
Objective
Defend a homework-help subscription model (step-by-step answers, tutoring) against the arrival of free generative AI, by adding an in-house AI assistant meant to retain students.
The deployment
Chegg sold subscription access to a database of exercise answers and to tutoring. After ChatGPT launched in late 2022, students could get the same kind of help for free. In April 2023, Chegg responded with CheggMate, an assistant built on GPT-4 in partnership with OpenAI, to retain its subscribers. The wrapper created no barrier: students who accessed ChatGPT directly no longer had a reason to pay. Subscriber count fell from a peak of about 7.8 million to 3.2 million in the first quarter of 2025. In February 2025, Chegg sued Google, accusing AI Overviews of destroying its traffic, and entered a strategic review. In the first quarter of 2025, revenue fell 30 percent and the company cut about 22 percent of its workforce.
Results Proof A
Financial results filed with the SEC (quarterly results quantifying revenue, subscribers, and layoffs), corroborated by major financial press (CNBC, Bloomberg) reporting the strategic review and the lawsuit against Google.
How it works
Inferred typical approachThe internal detail is not public. Here is a proven approach that leads to the same result, to adapt to your stack.
The stack in detail
- llm GPT-4 (OpenAI) underlying model of CheggMate, provided through the OpenAI partnership announced in April 2023; the same type of model as the free ChatGPT that was disrupting the market
- outil CheggMate in-house conversational assistant built as a layer on GPT-4, backed by Chegg's question-answer database
- infra Base de contenus Chegg Study proprietary database of exercise answers and tutoring that constituted the paid value of the subscription
Post-mortem
GraveyardWhat happened sourced
Chegg sold subscription access to a database of exercise answers and tutoring. ChatGPT launched in late 2022 and offered equivalent help for free. In April 2023, Chegg launched CheggMate, an assistant built on GPT-4 with OpenAI, to retain its subscribers. Subscriber count fell from a peak of about 7.8 million toward 3.2 million in Q1 2025. In February 2025, Chegg sued Google, accusing AI Overviews of destroying its traffic, and announced a strategic review. In Q1 2025, revenue fell about 30 percent and Chegg cut about 22 percent of its workforce (248 positions).
Reason for failure sourced
Chegg's paid content (step-by-step answers) became available for free through ChatGPT, and its own CheggMate assistant, a simple layer on GPT-4, added no barrier since the underlying model was directly available at no cost. On top of that came the search traffic decline tied to Google's AI Overviews, which Chegg is contesting in court.
Cost sourced
Market capitalization dropped from about 14.7 billion dollars (February 2021) to a few hundred million dollars; subscribers cut by more than half; about 22 percent of the workforce cut in Q1 2025.
Warning signs inferred
Inferred: when the value proposition is content the user can get for free elsewhere, an in-house assistant built on the same model as the free competitor protects nothing. The fact that Chegg's response relied on GPT-4, the very technology threatening its market, was a signal in itself. Dependence on Google's organic search traffic was a second known fragility.
Lessons in hindsight inferred
Inferred: a wrapper on a third-party LLM creates no defensibility if the base model is directly accessible. The barrier must come from elsewhere: proprietary data, workflow integration, captive distribution, or an experience the raw model does not provide. Responding to a disruption with the technology that causes it, without a proprietary advantage, amounts to funding your competitor.
Inferred: nuanced. Generative AI in education remains a valid pattern, and Chegg has real assets (a question-answer database, a brand). What is doomed is the moatless wrapper strategy: adding an AI layer built on the free competitor's model, with no data or distribution moat. Players that own proprietary data and a usage lock-in hold up better.
How your customers perceive this type of use
Sourced studiesLes consommateurs n'acceptent pas les chatbots par defaut : 64% prefereraient que les entreprises n'utilisent pas d'IA dans leur service client (Gartner, 2024) et pres d'un utilisateur sur cinq du service client par IA n'en retire aucun benefice (Qualtrics, 2025). L'acceptation se construit sur trois conditions mesurees par Salesforce : savoir qu'on parle a une IA, pouvoir escalader vers un humain, comprendre la logique de l'agent.
Acceptance conditions
- Etre informe qu'on parle a une IA et non a un humain (pres de 75% le demandent, Salesforce 2024)
- Un chemin d'escalade clair vers un agent humain (45% plus enclins a utiliser l'agent IA, Salesforce 2024)
- Une logique de l'agent clairement expliquee (44% plus enclins, Salesforce 2024)
Red lines
- Rendre l'humain injoignable : c'est la premiere inquietude des consommateurs sur l'IA dans le service client (Gartner 2024) et 50% craignent que l'IA les coupe du contact humain (Qualtrics 2025)
- Remplacer le service client par l'IA sans alternative : 53% envisageraient de partir chez un concurrent (Gartner 2024)
Sources: Salesforce 2024 · Gartner 2024 · Qualtrics 2025
How to replicate
Inference, not sourcedData prerequisites
- proprietary data not replicable by a generic LLM
- a usage signal or captive workflow
Org prerequisites
- ability to measure real added value vs the free raw model
- diversification of acquisition channels beyond organic search
Possible stack
- third-party LLM only as a complement to a proprietary moat
- exclusive first-party data
- deep product integration
First step: Before adding an AI layer, test whether a user already gets the same result for free with the base model; if so, the advantage must come from data or distribution, not the model.
Sources
- S1 CHEGG, INC - Form 8-K, résultats financiers T1 2025 Primary archive pending
- S2 Chegg sues Google for hurting traffic with AI as it considers strategic alternatives - CNBC Established press archive pending
- S3 Chegg Q1 2025 revenue down 30% as company restructures and explores strategic alternatives - EdTech Innovation Hub Secondary archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.