Chegg pivots into AI model training, framing its expert network as a data asset

Chegg said this week it is expanding into AI model training, pitching its network of subject-matter experts as a source of reasoning data for foundation model developers. The move repositions a business that has spent the past two years bleeding subscribers to ChatGPT into a supplier for the companies that displaced it.

The framing matters. Chegg is not announcing a new tutoring product. It is announcing that the same expert network used to answer student homework questions will now be sold as a data pipeline to AI labs training models on step-by-step problem solving.

Chegg's underlying claim is that long-form, human-written explanations of how to solve a problem, the format its experts have produced for years, are harder to synthesize than the raw question-answer pairs that earlier model generations trained on. Frontier labs are paying for chain-of-thought traces, graded solutions, and verified reasoning steps, and Chegg argues its archive and ongoing expert capacity are tuned for exactly that output.

Whether the archive is actually differentiated is the question buyers will press on. Scale AI, Surge AI, and Invisible Technologies have spent years building specialist-contributor networks producing the same kind of reasoning data, and OpenAI, Anthropic, and Google have direct contracts with several of them. Chegg's claim of being "uniquely positioned" is the marketing line. The defensible version is narrower: it has a standing pool of vetted experts in math, sciences, and engineering who have produced graded explanations at scale, and the cost of standing up that supply from scratch is non-trivial.

What the data labor market already looks like

The reasoning-data segment Chegg is entering is not greenfield. Scale AI was valued at roughly USD 14 billion in its May 2024 funding round on the strength of exactly this thesis: that the bottleneck in frontier model training has shifted from raw web text to curated, expert-generated reasoning traces. Surge AI, Invisible, and a handful of smaller specialists have followed. The labor model in each case is similar to what Chegg has been running, contractors with subject-matter credentials, paid per task, reviewed for quality.

What Chegg has that the data-labeling specialists do not is an existing archive. What it lacks is the AI-lab account coverage and the data-pipeline tooling those specialists have spent years building. Closing that gap is the actual business question, and the announcement does not address it.

Why the pivot is happening now

Chegg's subscription business has been under sustained pressure since ChatGPT launched. The company has warned in successive earnings calls that students are substituting free general-purpose chatbots for paid homework help. Selling the expert network as a B2B data product is a way to monetize an asset whose B2C value has collapsed.

It is also, implicitly, an admission. The same expert-written content that was supposed to be Chegg's moat against AI is now being repositioned as fuel for AI. The strategic logic is sound. The execution risk is that AI labs are sophisticated buyers who already have established suppliers and internal data teams, and they will price Chegg against those benchmarks, not against Chegg's prior subscription economics.