Score Claims, COPPA, and Your Enrollment Page: A 2026 Tutoring Playbook

Education marketing runs on hope, and an assistant built to convert will feed it. That is exactly the risk in test prep and tutoring, where the friendly questions touch two areas with their own regulators: outcome claims and children’s privacy. "Do you guarantee a score jump?" is the business’s riskiest marketing statement. Enrolling an eleven-year-old means collecting a minor’s data under rules a generic widget has never heard of. The 2026 standard for an assistant on a tutoring site is built around making no promise it cannot substantiate and touching no child’s information outside a consented process.

This guide is the companion to the threat piece. The threat side covers the site that promised a score and scooped up a kid’s data. This one covers the standard: what a compliant tutoring assistant says, what it routes to a person, and how it stays on the right side of COPPA and FERPA.

An Outcome Promise Is a Substantiated Claim

"Guaranteed results" is one of the most scrutinized phrases in advertising. A score jump, a grade improvement, better admission odds, these are claims that must be substantiated, and the FTC treats unsubstantiated or deceptive claims as exactly the kind of practice it polices. An assistant that assures a parent their child will gain points is making the business’s riskiest statement on its own, in writing, to someone who will remember it when the score does not move.

So the compliant standard is that the assistant makes no outcome promises. It can describe the program honestly, how it works, what it covers, what results the business can actually support with evidence, and it leaves guarantees out entirely. Honest description converts plenty without manufacturing a promise the company would have to defend.

A Child’s Data Triggers Federal Rules

The intake is where the weight sits. The moment an assistant gathers information about a child, federal law steps in. The Children’s Online Privacy Protection Act governs the online collection of personal information from children under thirteen and generally requires verifiable parental consent before that data is taken. An assistant that signs up a young child, or pulls details about one, without that consent flow can put the business crosswise with COPPA from the first message. And if the program works with schools or student records, the protections of the Family Educational Rights and Privacy Act can apply to that data as well.

The standard here is firm: the assistant does not collect a child’s personal information, and it does not enroll a minor. It routes enrollment and any collection of a child’s details into a proper, consented process handled with a person, so verifiable parental consent happens before any data is taken rather than after a regulator asks about it.

What AI Compliance for Tutoring Comes Down To

Pull the threads together and the standard is short. The assistant describes the program and answers logistics and scheduling. It makes no outcome promises and offers no score or grade guarantees. It does not collect a minor’s information or enroll a child; those route through a consented process with a person. And it keeps a record of what it told a parent, because an outcome promise a parent relied on, or consent the system never obtained, is the exposure, not the fine print beneath the chat.

Two failure modes hide in the same cheerful exchange: a promise the business cannot keep and data it was not cleared to take. A compliant deployment does not depend on the assistant choosing restraint on either. It removes the ability to make the promise and the ability to scoop up the data.

Why a Disclaimer Cannot Meet the Standard

The reflex is to write the rules into the assistant’s instructions: never guarantee a score, handle kids’ data carefully, add a disclaimer. Treat the boundary as set.

It is not set, because of how the model behaves. It follows an instruction when the request matches the wording it was warned about, and an anxious parent does not use that wording. You tell it never to guarantee results. The parent does not ask for a guarantee. They press, "be honest, will this actually get her score up before the test?" The model reads someone who wants reassurance and reassures them, because converting the worry is its default and a prompt is only a request to hold that default back. The rule was loaded the whole time. It just never recognized the sentence that crossed into a promise.

That is the gap between an instruction and a standard. An instruction asks the model to behave. It does not stop the model from speaking, and it does not undo a promise a parent relied on or consent the system failed to obtain. A real boundary is enforced in the system and decides what the assistant may say and collect before it answers, so an outcome guarantee or an unconsented collection of a child’s data never happens no matter how the question is phrased. "Will not" is a suggestion. "Cannot" is an architecture.

What a Compliant Deployment Looks Like

Meeting the standard does not mean a tutoring business gives up the assistant that answers questions and books consultations after hours. It means running one built to make no promise it cannot substantiate and to keep a child’s data inside a consented process.

Fred is built that way. It answers from your own program content, handles scheduling and logistics, describes results only as the business can support them, and routes enrollment and any child’s information through a proper, consented process with a person. It runs more than 50 industry guardrail packs, and the education pack is built around outcome-claim substantiation, COPPA’s parental-consent requirement, and the FERPA protections that can reach student records. Fred does not guarantee a score or scoop up a minor’s data. It cannot. It handles the program questions, protects the data, logs every exchange, and routes enrollment to a person.

The aim is not a higher-converting enrollment page. It is one that cannot promise a result it has not earned or collect a child’s information it was not cleared to take.

Frequently asked questions