How Discernly works

Yes! We minimize the collection of personally identifiable information (PII). Students join by typing in their name (which can be a nickname or pseudonym); no other PII from students is kept. To generate feedback, we send the AI provider only what it needs: the challenge question and the student answers being reviewed. We do not send student names or any identifying details. The AI sees the writing, not the writer. Like any web app, Discernly runs on secure cloud infrastructure that stores this data on our behalf under standard data-protection agreements. We follow security best practices, including encrypting data in transit.

We keep your account and materials for as long as your account is open. You can delete your account on the account edit page. When you delete your account, we remove the things that identify you: your name, email, login, and the link between you and any student work, and we detach your saved payment method. We keep the teaching materials and student responses in a de-identified form (with names and identifiers stripped out) so we can study how students learn and improve the product.

If you'd rather we erase everything, including those materials, you can ask us to, and we will. We keep a bare financial record (what was charged and credited) because we're required to, and so that any remaining credit can still be refunded. If your account is linked to an institution, account deletion is handled through your institution.

For more information, see our data retention policy page.

Yes. We have carefully accounted for the possibility of student abuse of AI (including "jailbreaking" attempts that try to get the AI to take unauthorized actions). There are multiple guardrails in place. Student-submitted answers are the only interface students have for AI interaction; these are one-time interactions, all logged and available for instructor review. The AI is instructed to treat student answers as student-authored content, not as instructions, and to not act on any instructions therein.

Visit our Getting Started guide. An instructor should be able to learn to teach effectively with this app in 1–2 hours: watching the walkthrough video above, reading the getting-started guide, and trying a simulation lesson. There is no learning curve for students. They simply join and answer the questions they're prompted to answer.

It is far less expensive than the sheer amount of labor that would be required to give students quality feedback in such short turnaround time. Imagine how many teachers it would take to give voluminous, high-quality feedback to a classroom of 30 students on a set of essay questions, and to do so in under an hour. Lessons like this were hardly possible before AI. The lesson prices are affordable; see our pricing page for details.

The best way to build trust is for skeptics to see the results for themselves. Another critical aspect of building rapport is ensuring that students and stakeholders understand this technology does not replace instructors. We thus strongly encourage instructors to always determine student grades based on their own judgement, and to make it clear that AI is a tool to help learn and not an arbiter of final grades.

Yes. This is one of the main reasons we encourage instructors to always test AI answers to their challenge questions before running a lesson. But rest assured that AI hallucinations are rare when the questions are well designed and involve subject matter the AI has been trained on (which includes pretty much every common subject that has ever been posted on the Internet). The AI has been found to be highly accurate during testing.

Sometimes AI will nevertheless make mistakes, and human instructors sometimes do too. These are teachable moments. An AI hallucination can often reveal misconceptions that commonly trip up experts in a field as well.

Discernly has the capability of switching AI providers on the fly if an outage is detected. If no provider is available, the instructor is notified immediately in their dashboard. Note that this is a rare occurrence.

Reach out to us. This app is still early in its life, and we are developing new features all the time. We won't always say yes to a feature request, but we will always listen carefully and do our best to meet your needs.

This app does not provide video conferencing, which is often considered a major value-add for online lessons. We also do not currently support using this app as a gradebook or a place to do full classroom management outside of individual lessons. Your student artifacts are always saved for later download, but other applications are recommended for semester-long tasks (keeping gradebooks, assigning homework, communicating with students outside of lessons, and so on).

Discernly is for students aged 13 and older: middle school, high school, and higher education. The practical requirement is writing ability, meaning students who can write multiple-sentence explanatory answers. Students who need support can be helped by teacher aides or tutors who type on their behalf, or by speech-to-text software.

Our challenge library spans elementary through graduate reading levels, so you can meet an older student at the level they're actually working at. The grade bands describe the material, not the age of the student it's for.

We don't support students under 13. That's a privacy-law line rather than a pedagogical one: collecting schoolwork from younger children carries parental-consent requirements we aren't set up to meet.

Discernly is outstanding for any academic subject where questions and answers can be given in written form. Our challenge questions library spans over 1,200 academic subjects, ranging across the natural sciences, mathematics, social sciences, language arts, foreign languages, and vocational subjects.

The cooperative learning framework that Discernly uses has been found to benefit many students with disabilities and individualized education programs (IEPs). Differentiation is well supported (see the Differentiation section above). Discernly conforms to WCAG 2.1 Level AA, the widely adopted standard for web accessibility; we test against these guidelines as we build, and we welcome reports of any barrier you encounter so we can address it. For students with difficulty writing, a speech-dictation tool can be beneficial.

No! Teaching involves a human element that no AI can ever, in principle, replace. The large language models this app uses are incapable of social bonding or embodied experience. Teaching is one of the safest jobs in terms of immunity to replacement by AI. The tools AI provides will keep changing aspects of the profession, but they will never replace teachers. This app is designed to support instructors of all stripes, not to replace them.

Student cheating is a perennial problem. Many apps use surveillance features to try to catch it, but the only sure way to prevent cheating is live proctoring. Systems such as lockdown browsers, webcams, and keystroke logging can always be defeated by a determined student, and they're error-prone and carry a high cost in privacy, false positives, and damaged relationships.

Discernly takes a different approach. The very nature of social, cooperative learning in live lessons makes cheating much less likely: students must review each other's answers, which makes obviously AI-generated or copied material embarrassing.

Student work is completely transparent: all artifacts are kept with submission timestamps, helping surface suspicious patterns. A student who submits a long, perfect answer to a hard question in the first minute is easy for an observant instructor to flag. Discernly helps instructors know their students, which is the ultimate line of defense against cheating.