What are the benefits and drawbacks in using different AI-powered tools for assessment? In part three of this four-part series, Philip Seyfried, Suet Cheah, Alok Sharma, and Dana Bassynbekova highlight the differences in the key features and level of teacher oversight and control that several prominent AI platforms, tools, and chatbots offer for assessment and development of student learning. Their analysis was produced as part of a project working with District 79 of the New York City Public Schools. Part one of this series provided an overview of some uses of AI in both large-scale standardized tests and classroom-based assessments. Part two described some of the new platforms, apps, and tools that teachers can use to create, analyze, and score assessments, particularly those that support more student-centered learning. Part three took a deeper look at and compared the strengths and weaknesses of the assessment capabilities of selected teacher-directed AI platforms, EdTech tools, and AI “assistants” with multimodal capabilities. In the final post in this three-part series, Adelaida Kim and Thomas Hatch provide examples of “micro-innovations” in assessment that leverage AI and new technologies to support the development of specific skills and abilities across subjects and levels. For related stories on AI and education, see: Can AI “ignite the mind and heart”? Stability & change in the education system in China (Part 3); Scanning the global headlines for recent news on AI, schools, and education; AI, Cellphones, Literacy, Students’ Mental Health, Political Turmoil and More: Scanning the Headlines for the Top Education Stories for 2025.
+++++++++++++++
Artificial intelligence (AI) is rapidly reshaping what is possible in educational settings. AI enables new pathways not only in how students learn, but in how their understanding can be demonstrated, assessed, and developed over time. But the benefits and drawbacks of AI for assessment differ substantially, depending on the affordances of different platforms and tools. Broadly, three categories of AI-powered tools differ in their key features and the degree of teacher oversight, which has important implications for assessment, particularly assessments that move beyond traditional text-based responses toward multimodal, creative, and conversational demonstrations of understanding.
- Teacher-Directed AI Platforms: Purpose-built EdTech tools designed to put curriculum and guardrails in teachers’ hands, with visibility into student interactions
- Creative AI Tools: Platforms that empower students to demonstrate learning through the creation of videos, presentations, podcasts, and images
- Multimodal General AI: Large-scale AI assistants that can generate text, images, and interactive experiences, and which are increasingly finding educational applications
Comparison at a Glance
| Tool | Teacher Setup | Student Creation | Quiz / Feedback | Multimodal Output | Transcript Access |
| Magic School AI | ✓ | ✓ | ✓ | ✓ | ✓ |
| School AI | ✓ | – | ✓ | ✓ | ✓ |
| Flint AI | ✓ | ✓ | ✓ | ✓ | ✓ |
| NotebookLM | ✓ | ✓ | ✓ | ✓ | – |
| Canva (AI) | ✓ | ✓ | ✓ | ✓ | – |
| Google Gemini | ✓ | ✓ | ✓ | ✓ | – |
| Google Gems | ✓ | ✓ | ✓ | ✓ | – |
Note: Current capabilities as of July, 2026. “Teacher Setup” refers to the teacher’s ability to configure the AI experience before students use it. “Student Creation” refers to whether students can produce their own multimodal artifacts (images, video, slides, audio) within the tool. “Quiz / Feedback” refers to the ability to quiz or provide feedback on student work. “Transcript Access” refers specifically to the teacher’s ability to read the student’s conversation with the AI.
Teacher-Directed AI Platforms
The tools in this category have been built specifically for K–12 and higher education contexts. Their defining features are teacher control, transparency, and safety. Educators can design AI-powered activities and chatbot experiences, set parameters for student interaction, and critically review what students have said and done on the platform. These features help make these tools useful for formative assessment, differentiated practice, and the ethical deployment of AI in school settings.
Magic School AI
| Magic School AI | Teacher-Directed AI Platform | |
| Overview | A comprehensive AI platform for educators offering 60+ AI-powered tools, including a “Magic Student” suite that lets teachers build structured AI activities for their classrooms. |
| Key Features | Teacher-designed AI chatbots and activities; student guardrails; access to conversation transcripts; learning analytics dashboard; differentiation tools. |
| Assessment Use | Teachers can assign AI-facilitated activities and review student interactions to assess comprehension, reasoning, and writing development. |
| Access Model | Free tier available; paid plans for full feature access. Widely adopted in US K–12 schools. |
School AI
| School AI | Teacher-Directed AI Platform | |
| Overview | A student-facing AI learning environment built around “Spaces,” which are teacher-configured AI experiences with defined purposes, personas, and guardrails. |
| Key Features | Custom AI Spaces for each assignment or learning context; teacher visibility into all student-AI conversations; real-time monitoring; safety filters. |
| Assessment Use | Teachers can use conversation data as evidence of student thinking; Spaces can be configured to ask Socratic questions rather than simply provide answers, generating richer assessment evidence. |
| Access Model | School and district licensing model. Designed to integrate into existing classroom workflows. |
School AI’s “Spaces” architecture can support alternative assessment approaches. Because teachers define what the AI does and does not do within each Space, they can design experiences that elicit student thinking rather than simply complete tasks for students. The transcript visibility feature is central to its use as an assessment tool.
Flint AI
| Flint AI | Teacher-Directed AI Platform | |
| Overview | An AI tutoring and classroom engagement platform that enables teachers to create custom AI tutors with specific personalities, knowledge bases, and behavioral constraints. |
| Key Features | Custom AI tutor creation; teacher monitoring of all student interactions; analytics on student engagement and learning; content guardrails; assignment integration. |
| Assessment Use | AI tutors can be configured to probe student understanding through questioning. Teacher-facing analytics provide formative data. Conversation logs serve as qualitative assessment artifacts. |
| Access Model | School and district licensing. Offers LMS integrations. |
Flint emphasizes the tutoring relationship and students receive personalized support while teachers maintain visibility and control. For assessment purposes, the combination of analytics and conversation transcripts offers a window into student thinking that traditional written assessments may not capture.
What Teacher-Directed Platforms Share
Across Magic School AI, School AI, and Flint AI, three features stand out as essential to their value for learning and assessment:
- Teacher-designed activities and guardrails: the teacher shapes the learning experience before students ever interact with the AI
- Conversation visibility: teachers can read what students said and how the AI responded, creating a rich record of student thinking
- Learning analytics: aggregate and individual data surfaces patterns in student engagement and understanding
Together, these features make teacher-directed platforms particularly useful for formative and alternative assessment at this time.
Creative AI Tools
The tools in this category empower students to create multimodal products, such as videos, podcasts, slide presentations, images, and diagrams, to demonstrate their learning. Teachers can ask students to produce an explanation, a visual argument, or a narrated presentation. AI dramatically lowers the technical barrier to these forms of expression, making them accessible to more students.
NotebookLM
| NotebookLM | Creative AI Tool (Google) | |
| Overview | Google’s AI-powered research and synthesis tool that allows users to upload source documents and interact with an AI grounded in those sources. Increasingly capable of generating multimedia outputs. |
| Key Features | Source-grounded AI responses (reduce hallucination); podcast-style audio generation from documents; study guide and FAQ generation; mind map and visual summary creation; video overview generation. |
| Assessment Use | Students can upload a set of sources and ask NotebookLM to generate a podcast, video overview, or visual explanation of a concept, thereby creating a multimedia artifact that demonstrates synthesis and understanding. Teachers can assess the quality of the artifact as evidence of learning. |
| Teacher Oversight | No direct teacher oversight or transcript visibility built in. Teachers assess the final product rather than the process. |
One way to use Notebook-LM for assessment is to ask students to upload primary or secondary sources and then produce a NotebookLM-generated podcast or video that explains a concept, event, or argument. The quality of the output reflects the quality of the sources students selected, and the prompts they used can reveal meaningful understanding. Students then can analyze the results, identifying what AI did well and what important information and perspectives it misses. Notably, however, selecting lower quality texts and prompts may make it harder for students to demonstrate more sophisticated analyses.
Canva (with AI Tools)
| Canva | Creative AI Tool | |
| Overview | A widely used design platform that has integrated a suite of AI tools for image generation, photo editing, presentation creation, video production, and more. |
| Key Features | AI image generation and editing; AI-assisted slide presentation design; video creation with AI voiceover; photo background removal and editing; “Magic Write” for text generation; diagram and infographic templates. |
| Assessment Use | Students can edit images for a visual argument, build an AI-assisted presentation to explain a concept, create an infographic to synthesize research, or produce a short video demonstrating the process they went through. These artifacts serve as alternative assessments of understanding. |
| Teacher Oversight | No built-in teacher monitoring of AI interaction. Assessment focuses on the finished product. Canva for Education offers classroom management features, including assignment sharing. |
Canva’s value for alternative assessment lies in its accessibility and range. Students who might struggle with a traditional essay may demonstrate a more sophisticated understanding through a well-designed infographic or a narrated video. Photo-editing tasks, such as adjusting an image for a presentation, can also demonstrate visual literacy and design thinking alongside content knowledge.
Both NotebookLM and Canva shift assessment from the process of thinking to its product, which can have both benefits and drawbacks. For instance, these tools enable simultaneous assessment of content knowledge and creative and technical skills, and the reduced technical barrier means more students can demonstrate learning in ways that suit their strengths. However, teachers must develop criteria for assessing multimodal products, not just written responses. Furthermore, the artifact itself (the video, the podcast, the presentation) becomes both the assessment and the evidence of learning, but neither tool currently provides teachers with visibility into the AI interactions that produced the artifact.
Multimodal General AI
General-purpose AI assistants like Google Gemini are increasingly capable across modalities by generating text, producing images, creating video, and supporting interactive experiences. As these tools add educational features (like Google Gems), they occupy a middle ground between creative tools and teacher-directed platforms. They are powerful and flexible, but their oversight features are currently more limited than those of purpose-built EdTech platforms.
Google Gemini
| Google Gemini | Multimodal General AI (Google) | |
| Overview | Google’s flagship AI assistant, available across the Google ecosystem. Gemini is increasingly multimodal, capable of generating text, images, and video, and of reasoning across multiple formats. |
| Key Features | Text generation and conversation; image generation; video generation (Veo integration); code generation; document analysis; integration with Google Workspace tools. |
| Assessment Use | Gemini can quiz students on a topic, provide feedback on submitted writing or work, and help students explore ideas through conversation. Its multimodal output capabilities allow students to create a range of artifacts. |
| Limitation | At this time, Gemini does not provide teachers with access to transcripts of student conversations. Assessment of AI-assisted work must rely on final products or student self-reporting. |
| Access | Gemini Advanced is available through Google accounts with a Google One subscription. Integrated into Google Workspace for Education. |
Gemini’s growing multimodal capabilities make it a versatile tool for student creation. A student might ask Gemini to generate an image illustrating a concept, then embed it in a presentation with an explanation they have written. Or they might submit a draft essay and receive detailed feedback, an experience that mirrors personalized writing conferences. However, the absence of teacher-visible transcripts is a key limitation for use in formal assessment. Teachers can assess what students produce with Gemini’s help, but not the quality of their reasoning within the conversation itself.
Google Gems
| Google Gems | Configurable AI Agents within Gemini | |
| Overview | Gems are customizable AI personas within Google Gemini. Teachers or institutions can create Gems with specific roles (a research assistant, a writing coach, a practice quiz partner) and share them with students. |
| Key Features | Custom AI persona creation with defined roles and instructions; ability to set a specific knowledge focus; shareable with student groups; supports Socratic questioning, practice environments, and research assistance. |
| Assessment Use | A teacher can create a Gem configured to quiz students on a specific topic, ask follow-up questions, or provide scaffolded feedback on submitted work. Students interact with a purpose-built AI experience within the broader Gemini environment. |
| Limitation | Like Gemini generally, Gems do not currently provide teachers with access to conversation transcripts. The teacher sees neither what the students asked nor how the Gem responded. |
| Comparison Note | Gems offer similar configurability to teacher-directed EdTech platforms (Magic School AI, School AI, Flint), but currently lack the teacher-visibility features that make those platforms most useful for assessment. |
Google Gems represent an important development: a mainstream AI platform that adds teacher-configurable experiences. However, Gems’ usefulness for assessment depends on whether or not future versions add teacher visibility into student interactions.
Implications for Assessment Design
Each category of tool implies a different approach to assessment design. Understanding these distinctions helps educators choose the right tool for the right assessment purpose.
Process-Visible Assessment
When using teacher-directed platforms (Magic School AI, School AI, Flint AI), the conversation itself is an assessment artifact. Teachers can evaluate:
- The depth and relevance of questions a student asks the AI
- How a student responds when the AI pushes back or asks a clarifying question
- The progression of a student’s thinking across a multi-turn conversation
- Whether a student recognizes when the AI has made an error
This form of assessment is particularly valuable because it surfaces metacognitive processes that traditional written products often obscure.
Product-Based Assessment
When using creative tools (NotebookLM, Canva) or general AI (Gemini), assessment focuses on what the student creates. Rubrics for product-based assessment should address:
- Accuracy and depth of content knowledge demonstrated
- Quality of synthesis across multiple sources or ideas
- Clarity and effectiveness of communication in the chosen medium
- Evidence of original thinking beyond AI-generated content
- Appropriate attribution of AI-assisted elements
Product-based AI assessment works best when teachers pair it with reflections or oral explanations, giving students the opportunity to articulate what they made and why, which provides additional assessment evidence and can help reveal when students have not thought through or critically analyzed their use of AI.
Feedback and Practice Environments
Both teacher-directed platforms and Google Gems can serve as low-stakes practice and feedback environments that provide students with iterative responses to their work without the pressure of a formal grade. These uses include:
- Submitting a draft for AI feedback before a teacher conference
- Practicing for a presentation by interacting with a Gem configured as an audience member
- Working through a problem set with an AI tutor that provides hints rather than answers
In these contexts, the AI functions less as an assessment instrument and more as a rehearsal space that frees teacher time for higher-order feedback and conferring.
Looking Ahead
The landscape of AI tools in education is evolving rapidly. Several trends are worth watching:
- Growing teacher oversight of general AI tools. Platforms like Google Gemini and its Gems feature may add transcript visibility and classroom management features as they develop education-specific versions. This would significantly expand their usefulness for assessment.
- Richer multimodal assessment. As students gain facility with tools like NotebookLM and Canva, educators will have the opportunity to design assessments that require video explanations, well-designed presentations, or AI-assisted research products, pushing assessment beyond text-based demonstrations of knowledge.
- The conversation as curriculum. In teacher-directed platforms, the transcript of a student’s conversation with AI is a new kind of learning record that documents not just what a student knows, but how they think. Assessment frameworks will need to evolve to make productive use of this data.
- Questions of attribution and academic integrity. As AI becomes embedded in the creation process, assessment design must grapple with what it means to demonstrate learning when AI has contributed to the product. Clear expectations, reflection requirements, and process-visible assessment approaches all play a role.
Next week: “Micro-Innovations” in Assessment for Specific Subjects, Levels, and Contexts: AI, New Technologies and the Future of Assessment (Part 4)
International Education News 