Best AI Oral Assessment Platforms for Higher Education

How to read this guide#

This guide compares platforms that support oral assessment in higher education. It is meant to help a faculty member, an instructional designer or a committee build a shortlist of three or four products to pilot. It does not rank products on a single scale, because the products are built on different ideas about how an oral assessment should work.

Everything in the comparison table comes from each vendor's own public pages, read on 9 October 2026. The sources are listed at the end of the guide. We did not test the products, and we did not use any other vendor's or reseller's description of a product as evidence. Prices change often, so the table records whether prices are published, not what they are.

A cell marked "Not verified" means we did not find the information in the pages we reviewed. It does not mean the product lacks the capability. If a capability matters to you, ask the vendor.

The first decision: who conducts the session#

Before comparing features, decide who should be the examiner. The platforms fall into four groups.

1. A person conducts, AI supports. The assessor asks the questions and decides the grade. AI helps prepare questions and review the recording. FeedbackFruits describes this model.
2. Students record, people mark. Students answer prompts on video on their own time, and markers score the recordings. Cadmus describes this model, with AI assistance for preparing prompts planned.
3. An AI examiner questions the student. The AI reads the student's work or the course content, asks questions, follows up and drafts a score. Rocketproof.ai, OralExam.ai and Eduface describe this model.
4. A platform for building voice agents. You or the vendor configure an agent for a scenario. Tough Tongue AI describes this model, and its pages focus on role-play, coaching and screening, with oral exams described in its blog.

None of these is wrong. A department that wants a person in the room for every high-stakes viva will choose differently from a department that wants every student in a section of 200 to have a conversation.

The comparison table#

Notes on each platform#

Rocketproof.ai#

Rocketproof.ai is published by the same company that publishes this guide. It runs oral defense of submitted papers, slide decks, readings and cumulative work, and it also runs instructor-defined AI role-play. The AI asks questions drawn from the submission, students respond by speaking, in a live conversation or by typing, and the instructor reviews the transcript and draft rubric scores before releasing a grade. It connects to an LMS through LTI 1.3 and has been verified with Canvas. Blackboard, Moodle and Brightspace are supported but not yet observed on a vendor test system. It has no VR option, no public API and no MCP server.

Best for: oral defense of submitted work and instructor-defined role-play in one platform. Check before you buy: your LMS, if it is not Canvas.

FeedbackFruits Oral Assessment#

FeedbackFruits describes a live oral assessment in which a person conducts the session. Its page describes shared rubrics and question sets, recording and transcription, a second assessor who can join, an assistant that suggests follow-up questions only the assessor sees, and grades that return to the LMS gradebook. It names Canvas, Moodle, Blackboard and Brightspace.

Best for: programs that want an instructor or assessor to stay the examiner while AI supports preparation and review. Check before you buy: it is not designed for an AI to question every student without a person present.

OralExam.ai#

OralExam.ai describes an AI examiner that reads a student's upload and discusses it for about six minutes, then reports Understanding and Authorship as separate results with evidence and a transcript. Its page names Canvas, states that pricing is not set yet and says the first university courses go live this fall. It is early in its life.

Best for: short AI-led conversations about a student's upload. Check before you buy: how the Authorship result is produced and what evidence supports it, rubric options, and support at your scale.

Cadmus Oral Assessment#

Cadmus describes asynchronous video responses linked to a student's earlier written submission, with rubric-aligned marking by people and recorded evidence. Its page states that its AI will not mark, grade or give feedback, and that AI help with preparing prompts is planned for later.

Best for: departments that want people to mark recorded oral responses without scheduling live sessions. Check before you buy: it is not a live adaptive examiner.

Eduface Oral Examination#

Eduface describes three AI-powered oral exam formats, including one that generates questions from a student's submitted work and a practice mode with unlimited runs. Sessions are delivered through an LMS link, take 15 to 20 minutes, and grades return to the gradebook after the lecturer approves them. It says it works with Canvas, Moodle, Blackboard or any LTI 1.3 platform.

Best for: programs that want AI-led exams delivered through LTI 1.3 with a practice mode. Its pricing page lists a free tier and says institutional pricing is custom. Check before you buy: evidence retention and security documentation, which we did not find on the pages we reviewed.

Tough Tongue AI#

Tough Tongue AI describes a platform for building live voice agents that run on phone, Google Meet, Zoom and the web, with free trial minutes and published tiers. Its product page focuses on sales, hiring and training. Its blog describes university oral exams and role-play built on it, and says a scenario can carry rubrics. Its blog also describes a path through MCP servers, with a Canvas integration built from third-party components.

Best for: teams that want to build custom voice-agent experiences and deploy them where students already meet. Check before you buy: LTI, grade return and accommodations, which its blog says to confirm for your workflow.

Bongo Question & Answer#

Bongo describes timed video responses to a prompt, with AI feedback scored against a rubric on accuracy, completeness and composure, inside an LMS. It names Canvas, Blackboard and Moodle, and says student data is never used to train AI models. It also describes a separate AI role-play product, covered in the role-play guide.

Best for: structured video responses under time pressure. Check before you buy: a timed prompt is not an open conversation.

Where another product is the stronger choice#

• If you want a person to remain the examiner, FeedbackFruits or Cadmus fits better than any AI-led product.
• If your priority is documented LTI 1.3 delivery to any compatible LMS, Eduface's page states that more broadly than Rocketproof.ai, which has verified only Canvas.
• If you want an agent that joins Zoom or Google Meet, or you want your own AI assistant to operate the platform, Tough Tongue AI documents that and Rocketproof.ai does not offer it.
• If you need a short AI conversation about an upload and are comfortable with an early-stage product, OralExam.ai fits that description.
• If your need is communication practice rather than graded oral assessment, see the guide to AI role-play and simulation platforms.

A seven-step evaluation#

1. Decide whether the goal is practice or assessment. Practice can be private and repeated. Assessment needs consistency, evidence, accommodations and an appeal path.
2. Choose who conducts the session. Use the four groups above.
3. Test with real course material. Use an actual assignment and rubric, not a vendor demo.
4. Inspect the evidence behind the score. Ask to see a session record, how a score links to it and how you change it.
5. Walk the LMS path. "Integrates with Canvas" can mean a link or a full LTI launch with grade return. Ask for each step. Our guide to using AI with Canvas for assessment explains the terms.
6. Review privacy and accessibility. Ask where recordings are stored, how long, whether models are trained on student data, and how accommodations work.
7. Pilot one recurring moment. One course, one activity, a few repetitions. Compare the evidence with what the existing assignment gave you.

Other comparisons#

Other guides cover similar ground. SnapSDR and Tough Tongue AI have each published comparisons of oral assessment and role-play platforms in 2026. Both are published by companies with an interest: SnapSDR states that it runs on Tough Tongue AI's agent infrastructure, and Tough Tongue AI's guide is written by a person who builds Tough Tongue AI. Read several guides and check claims against vendor pages, including this one.

What we could not verify#

On 10 October 2026 we read additional pages for Cadmus, Eduface and Bongo and updated those rows. The other rows are as of 9 October.

Professr.io appears in other guides. Its domain did not resolve when we checked on 9 October 2026, so we did not include it. We also reviewed fewer pages for some vendors than others. If you spot an error or a missing capability, tell us through the contact page with a link to the vendor page and we will review it.