Chinat · September 12, 2026

A Public Standard for Edtech Claims and Evidence

Summary

Schools need a reliable way to distinguish a persuasive edtech demo from a product whose learning claims have been tested. AI education products can reach schools before their claims have been tested or clearly explained. Buyers need to see what a product claims, what supports those claims, where it was tested, and what remains unknown.

Academic review, accreditation, crowdfunding, product discovery, and app stores each handle part of this job. None gives buyers one place to compare claims, evidence, costs, and limits. I propose an independent international organisation that works like GoFundMe + Product Hunt + an app store for edtech, governed by public rules for evidence and conflicts of interest. It would fund promising work, surface useful products, and certify a minimum standard of evidence and disclosure. The goal is to help good research travel further, help buyers make better decisions, and reward companies that can support their claims.

The Problem

A polished demo can make a lightly tested product look as credible as one studied carefully. Teachers, schools, parents, and learners are then left to figure out the difference.

On September 7, 2026, I met Marcos Rojas Pino at Stanford GSE and discussed an idea for an independent international organisation outside Stanford. Marcos created ClinicalMind AI in 2023, and his concern is that too many AI edtech products are promoted like marketing ploys, without enough research or pedagogy behind their claims.

A product might claim to improve learning, save teachers time, personalise instruction, or expand access. Each of those claims can mean something different depending on the learner, subject, setting, language, and method of measurement. Without a shared format for presenting the evidence, even a sincere buyer has difficulty comparing products.

The academic side has the opposite problem. Earlier this year, I spoke with a dean at Stanford about the need for an edtech-specific app store for university projects. A research team can build something useful, study it in a classroom, and publish its findings, but there is no dedicated path that helps the product secure funding, reach users, and remain discoverable after the original project ends.

Research teams are usually not set up to manage distribution, sales, support, and long-term product development. Companies are usually not set up to wait for an academic publication before every release.

This is the gap: research needs a path to adoption, and commercial products need a path to credible evidence.

What Exists Today and Why It Is Not Enough

Several existing systems already solve parts of this problem. The issue is that each one stops at a different boundary.

Accreditation bodies establish standards for institutions, programs, and credentials. They provide accountability at an important level, but they are not designed to evaluate every product update, learning tool, or early university project entering the market.

Academic peer review can examine research design, methods, and findings. It is one of the strongest tools we have for testing knowledge claims. However, a paper is not a distribution channel, a funding mechanism, or a usable catalog for a teacher trying to choose between products.

Commercial app stores make products searchable and give developers a route to users and revenue. They can enforce their own technical and commercial rules, but they are not designed to determine whether a learning claim has sound pedagogy or appropriate evidence behind it. Ratings can tell us whether users liked an experience, but satisfaction and learning are not the same thing.

Crowdfunding and product discovery platforms help early projects attract attention, feedback, and capital. They can show that people are excited about an idea. They do not tell a school whether the product was tested responsibly, whether the evidence matches the claim, or whether the design works in that school's context.

These systems do different jobs. Trust gets lost when an edtech product moves between them.

An academic project might have credible research but no path to market. A company might have distribution and revenue but no clear way to communicate the quality of its evidence. A school might have procurement rules but no common record for comparing the two. We need one institution that connects these pieces while respecting why they developed separately.

The Proposal

The proposal is an independent international standards body with a narrow first task: make edtech claims and evidence easier to compare. Think of it as GoFundMe, Product Hunt, and an app store for edtech, governed by public rules for evidence and conflicts of interest.

It would have three connected functions: fund, surface, and certify. The first version should publish evidence cards and run a small review process. Funding and product discovery should follow only if they can operate without influencing certification.

Fund

The GoFundMe function would let researchers, educators, companies, and project teams raise money for product development, classroom pilots, independent evaluation, localisation, or accessibility work.

Each campaign would explain the learning problem, intended users, current evidence, proposed milestones, and how the results will be reported. Funding would support both the product and the work needed to understand whether it performs as claimed.

This matters because evidence costs money. If we ask early teams to conduct better evaluation without helping them fund it, the standard becomes a gate that only well-capitalised companies can pass.

Surface

The Product Hunt function would give new and existing projects a place to launch, receive structured feedback, and become discoverable.

Each product would have a common evidence page. A teacher or school could see the intended learner, pedagogical basis, evidence status, implementation requirements, known limitations, languages, pricing model, and history of meaningful product changes.

Discovery should not depend only on which company can spend the most on marketing. Products could be filtered by learning goal, age group, subject, language, evidence stage, accessibility needs, and deployment setting. University projects would appear beside commercial products under the same disclosure format.

Certify

The catalog would admit products that meet a published disclosure standard. Any claim of effectiveness would require a separate review of the evidence for that claim, product version, and setting.

Certification would not mean that a product is universally effective or that the organisation endorses every claim it makes. It would mean that a particular product version has met a published standard, disclosed its evidence honestly, and provided enough information for a buyer to make an informed decision.

Certification must be tied to the product, version, intended setting, and specific claims reviewed. A major change to the product or its claims should trigger an updated review.

Governance

No university, company, or funder should control it. Its governing structure should include researchers, educators, learners and families, product builders, funders, procurement leaders, and representatives from public education systems.

International representation matters because a standard written only for well-funded English-speaking schools will fail as soon as it reaches a different language, curriculum, budget, or regulatory setting. The core standard can be shared while local reviewers explain what adoption requires in their context.

The organisation should publish its review criteria, conflicts of interest, decisions, and reasons for changing those decisions. Reviewers should disclose financial or research relationships and recuse themselves when needed. Funding and sponsorship should be separated from certification decisions, with a clear rule that nobody can pay to pass.

There should also be an appeal process and a public record of significant certification changes. Trust will come from the organisation showing its work, especially when it changes its mind.

Making Academic and Company Incentives Coexist

Companies need speed, revenue, intellectual property protection, and a route to customers. Academic teams need methodological honesty, independence, time to study results, and the ability to publish credible work.

The organisation should accommodate both without pretending that their incentives are identical.

A company could choose which learning claims it wants reviewed and submit evidence for those exact claims. Researchers could evaluate the methods and boundaries of that evidence without demanding access to unrelated company information. The public record could summarise the basis for the decision while protecting student information and legitimate confidential material.

Companies that meet the standard would gain discoverability, credibility, and a clearer path into schools. Researchers would gain a route for turning useful work into adoption, along with potential support for evaluation. Educators would gain a common language for comparing products.

The organisation needs a funding model that does not reward favourable decisions. Membership dues, review fees, grants, and sponsorship are all models worth testing, but the source of the money cannot determine the outcome of a review. The final structure needs to make that separation visible.

What the Minimum Standard Would Certify

At a minimum, the standard should certify that:

• The product clearly identifies its intended learners, learning problem, use case, and setting.
• Its public claims are mapped to available evidence, with no stronger conclusion than the evidence supports.
• The team explains the pedagogical reasoning behind the product, including where that reasoning remains uncertain.
• Relevant learner safety, privacy, accessibility, and human oversight questions have been addressed and disclosed.
• Outcomes can be measured, and the product identifies known limitations, failed assumptions, and missing evidence.
• The certified product version, review date, conflicts of interest, and future material changes are recorded.

This standard would not replace local laws, institutional review, accreditation, procurement, or the judgment of educators. It would create a common starting point so that each buyer does not have to reconstruct the same basic questions alone.

The organisation should also allow evidence to mature. An early university prototype should not be held to the same evidence expectations as a product making broad claims to school systems. The label should tell the buyer what has actually been demonstrated, not turn every promising experiment into a pass or fail judgment.

Why Now

AI education products can change faster than schools can evaluate them. A team can change a model, interface, or workflow quickly. The evidence attached to the old version may no longer describe what learners are using.

Fast product development can also create useful learning tools. A 2024 Stanford Learning Design and Technology report on AI Playground, a project by Jerry Shan, studied a 2-hour class of 29 students in grades 8 to 12. All but 2 had never learned about neural networks before, and by the end of the class they had built and tested their own. Of the 22 students who answered the post-survey, 11 described the tool as intuitive or easy to use without being asked.

That study should not be treated as proof that every student, classroom, or AI learning product will produce the same result. It does show what may become possible when the product, instruction, and learning design work together. We need a better system for helping work like this move from a study into responsible use.

MentorMates has run hackathons in the Bay Area, Hong Kong, and Kuching. These events have shown me that people across regions are ready to build. The missing piece is a trusted route from an early project to funding, evidence, and adoption after the event ends.

Brazil is one possible test market, subject to local partners confirming the need, costs, and route to adoption. Lucas, a Brazilian at Stanford GSB, is leaving Toma Educação at the end of December 2026 and has offered introductions to its CEO and the Paraná education secretariat. Toma Educação runs 36 public schools serving about 20,000 students, while the Paraná path could reach a system of about 2,000 schools.

Lucas also named the blockers directly. Products need to work in Portuguese, and school budgets are thin, at about an 11% margin. These are offered introductions, not commitments or a signed pilot.

A Brazil pilot would force the organisation to address language, cost, evidence, and implementation together. If the model only works for wealthy schools using English-language products, then it is not yet an international model.

Who We Have Talked to and Who Should Be in the Room

The current proposal comes from several conversations and operating experiences, but the founding group has not been formed.

Marcos Rojas Pino, Stanford GSE. Marcos created ClinicalMind AI in 2023 and raised the concern about marketing-first AI edtech without enough research or pedagogy behind it. On September 7, 2026, he pitched the idea of an independent international organisation outside Stanford, with a target around June or July 2027.

A dean at Stanford. An earlier conversation focused on an edtech-specific app store for university projects. That perspective is important for understanding why good university work struggles to find a route beyond research and how an institution could support it without claiming ownership.

Alvin Liu, Johns Hopkins. Alvin co-organised the Kuching Healthcare AI Hackathon on August 1, 2026, which I helped run. His academic and healthcare hackathon experience would be useful when figuring out how standards should work in settings where educational and domain-specific claims overlap.

Jacob Hsu, Just Ventures. Jacob is the former CEO of Catalyte and now works through Just Ventures. His workforce and apprenticeship experience could help connect product standards with real pathways into learning and work.

Lucas, Stanford GSB. Lucas has already identified Portuguese localisation and school economics as real constraints, which makes him the right person to test whether the Brazil path is practical.

The first working group should include people beyond my existing network. Educators, students, parents, researchers, founders, funders, public-system leaders, procurement teams, and people responsible for privacy and accessibility all experience different parts of the decision.

They should be involved while the standard is being written, not asked to react after it is finished.

Open Questions

• Should this begin as a consortium, nonprofit standards body, research collaborative, or another structure?
• What gives the organisation legitimacy before it has a long operating history?
• What evidence is enough for an early prototype, a growing company, or a product selling into school systems?
• Which requirements should be universal, and which should be adapted by country, language, age group, or subject?
• How can the organisation charge fees without creating a pay-to-pass certification system?
• How should it protect student data, unpublished research, and legitimate company information?
• Who hears appeals, and when should a certification be suspended or withdrawn?
• How do we stop popularity on the discovery platform from being mistaken for educational quality?
• What would count as enough evidence to expand beyond an initial Brazil pilot?
• How do we keep the standard useful when products can change faster than a traditional review cycle?

Next Steps

1. This month: Circulate this paper to Marcos, the Stanford dean, Alvin, Jacob, and Lucas. Ask each person to challenge the problem definition, proposed functions, and missing voices, then correct the factual record before wider distribution.

2. Next month: Draft the first evidence card and minimum certification standard. Test it on a small set of volunteered university, company, and hackathon projects, and publish what the process gets wrong.

3. The following month: Convene a working discussion, decide whether Brazil should become the first discovery and pilot market, and identify the organisational path that could support the target around June or July 2027.

If you are an educator, researcher, founder, funder, or public-system leader who wants to help define this institution, send me the product, evidence, or unanswered question that should shape the first standard.

← Back to all posts