Edtech and AI: What the Evidence Says About Substitution

The debate over whether AI substitutes for edtech products or augments them is usually argued with anecdotes. Our assessment offers something more concrete: of the 10,525 companies assessed in July 2026, several named edtech and tutoring firms sit inside the AI vertical, and their standings show which side of the substitution line they landed on.

The short version: paid, standalone tutoring and language products are the ones under visible pressure. Companies whose value was a proprietary tutoring interface on top of increasingly commodity model capability are being squeezed by free or bundled features inside larger apps. The augmentation story survives mainly where a company owns something the model layer cannot reproduce, and among edtech names in this assessment, that is rare.

The category context edtech sits inside

Edtech AI firms in our data are tracked within the AI companies vertical. Of the 263 AI companies assessed in July 2026, 170 were rated losing its protection and only 2 were judged hard to displace. Across that same July 2026 dataset, 88 companies faced high competitive pressure at the time of assessment.

The sector-level pattern is that value living above the model layer tends to be temporary, while value below or around it (chips, physical hardware, regulated or exclusive data) holds up better through model upgrades. Edtech is a near-pure test of this, because tutoring, translation, and study help are exactly the capabilities that general assistants ship natively.

The named edtech companies and how they stand

Four edtech-focused companies appear in the AI vertical of the July 2026 assessment, and none stands better than losing its protection:

Riiid, the Korean AI tutoring company, was rated Little protection left in its primary July 2026 record, one of the more exposed standings recorded in the AI vertical that month. A later assessment pass placed it losing its protection instead; neither reading found it in a comfortable position.
Yuanfudao, the Chinese online tutoring group, was rated losing its protection in July 2026.
NetEase Youdao, whose business spans dictionaries, translation, and learning hardware, was rated losing its protection in July 2026, and remained there in a later pass.
Squirrel AI Learning, the adaptive-learning company, was rated losing its protection in July 2026.

The assessment pressure map names the mechanism directly for three of these. Riiid, Yuanfudao, and NetEase Youdao face pressure from Khan Academy, Duolingo, ByteDance, and general LLM apps, because free or bundled tutoring and translation features inside larger consumer apps are eating the paid niche these companies built. That is substitution in its plainest form: the customer's job (get homework help, translate a passage, drill for an exam) is now done adequately by a product they already have, often at zero marginal price.

Adjacent learning-and-writing tools show the same shape. Grammarly, whose writing assistance overlaps heavily with what general assistants now do inline, was rated losing its protection among the 263 AI companies assessed in July 2026. Research-learning tools Elicit and Consensus were each rated losing its protection in the same assessment.

The Chegg signal

Chegg is one of the clearest real-world examples of AI substitution in education: a paid homework-help subscription whose core job, answering study questions, was largely absorbed by free general chatbots. In a historical check we ran on already-failed companies, using only information that was knowable up to 18 months before each company's collapse, Chegg turned up in an unexpected place: it was one of ten companies initially selected as a healthy survivor comparator for the study, alongside Fisker, Nikola, and Getir, and all ten turned out on closer inspection to already be struggling or gutted businesses rather than stable comparators.

That matters because it means the fundamentals of Chegg's edtech business were already pointing toward decline well before the collapse was conventional wisdom. Across that same historical check, the companies that ultimately failed were almost always correctly identified as struggling ahead of time, and none of the companies that failed within 18 months was ever described as a safe bet. We do not extend that check to declare outcomes for companies still operating today; what it shows is narrower: the kind of deterioration visible in Chegg's numbers before its decline was the same kind that reliably preceded failure elsewhere in the sample.

Where augmentation actually shows up

If substitution dominates the edtech names, where does augmentation hold? In this assessment, the companies weathering AI pressure best are those whose product is infrastructure, hardware, or exclusive data rather than a layer on top of general models. Applied Intuition sells simulation and testing software that carmakers and defense agencies depend on to build autonomy, and was rated Hard to displace. Nvidia still runs the chips and software stack behind most AI training and inference, and was rated defensible for now. Waymo already runs a paid robotaxi service in more than 10 US cities with Alphabet's backing, and was rated defensible for now. PathAI sells image analysis tools to labs and drug researchers, a specialized, high-trust niche, and was rated defensible for now. None of these four is an edtech company, and that absence is itself the finding. No edtech firm among the 263 AI companies assessed in July 2026 was rated better than losing its protection, because none in the tracked set owns a moat below the model layer, such as regulated data, physical distribution, or hardware.

The closest edtech analogue to a defensible position would be exclusive learner data, accreditation relationships, or institutional distribution that a general assistant cannot bundle. The assessment does not show a tracked edtech company that has converted those into a stronger standing yet.

Watch-items, dated, for readers to verify

These are not predictions; they are dated checkpoints from the July 2026 assessment that readers can verify themselves:

Whether more AI tooling or wrapper startups are quietly acquihired by OpenAI, Anthropic, Google, or Meta rather than raising new rounds. The assessment already logs Humanloop absorbed by Anthropic, Rewind AI acquired and shut down by Meta, Covariant hollowed after Amazon hired its founders, and Phind shut down in January 2026 after ChatGPT and Gemini absorbed its niche. Edtech tools sit in the same exposed layer.
Whether Chinese hyperscalers keep launching bundled AI platforms, since bundling is the same mechanism pressuring Yuanfudao and NetEase Youdao at home.
Whether Riiid's standing converges across future assessment passes; its July 2026 records disagree, ranging from Little protection left to losing its protection.

Questions this data answers

Is AI substituting for or augmenting edtech companies? For the edtech firms tracked in the July 2026 assessment, the evidence points to substitution. Riiid, Yuanfudao, NetEase Youdao, and Squirrel AI Learning all sit at losing its protection or worse, with free or bundled tutoring and translation features inside larger apps named as the pressure source.

Which edtech companies are under the most AI pressure? Riiid shows the most exposure, rated Little protection left in its primary July 2026 record, the most severe standing recorded among the tracked edtech names. Yuanfudao, NetEase Youdao, and Squirrel AI Learning are all rated losing its protection in the same assessment.

What happened with Chegg in the historical check of failed companies? Chegg was one of ten companies initially selected as a healthy survivor comparator that turned out, on inspection, to already be a failed or gutted business, a group that also included Fisker, Nikola, and Getir. Chegg's paid homework-help subscription had its core job absorbed by general chatbots well before its decline became widely recognized.

Do any AI-era edtech companies look durable? Not in this assessment as of July 2026. No tracked edtech firm was rated better than losing its protection, and the AI companies assessed as most durable, such as Applied Intuition and Nvidia, own infrastructure or hardware moats that no edtech name in the set has.

Why do general chatbots hurt paid tutoring apps specifically? Because tutoring, translation, and study help are core native capabilities of general assistants, the paid niche gets bundled away. The assessment pressure map names Khan Academy, Duolingo, ByteDance, and general LLM apps as the forces eating Riiid, Yuanfudao, and NetEase Youdao's paid markets.

Assessments reflect evidence available July 2026. Companies change; treat watch-items as things to verify, not conclusions.