AI Tools and Platforms

Jason Blomquist; Liza Long; Joel Gladd

4 AI Tools and Platforms

Beyond the Chatbot

Liza Long compared ChatGPT, Claude, and Gemini to three interns with different strengths. One is eager and verbose. Another feels like an academic colleague. The third is deeply wired into Google’s ecosystem. In Long’s experience, each platform developed a different working style, and that difference shaped how she used it.

This chapter starts with a brief overview of the major platforms, then follows Dr. Sarah Chen as she tests three features that changed how she thinks about AI in her teaching. By the end, you’ll have a clearer sense of which tools and features might fit your own workflows.

AI features are also showing up in tools faculty already use. Canva, Adobe Creative Cloud, Grammarly, and Google Docs all offer embedded AI features. Availability depends on the product, plan, account, and administrator settings. You may encounter generative AI without ever opening a dedicated AI platform. Bryan Alexander notes that these embedded AI features have become a routine part of software faculty were already using before 2025.

An AI platform is only one part of the experience. Its interface, enabled tools, sources, account type, and institutional settings can all change what the AI can see and do. Even the same underlying model may behave quite differently in a simple chat, a source-grounded notebook, a reusable course project, or an agentic workspace.

A Simple Tool Map

Current products often belong to more than one category. This map helps clarify what kind of context or action a faculty task requires before you compare brands.

Common AI Tool Families
Tool Family	What Changes
General chatbot	A conversational interface that may also include search, files, memory, or connected tools
Source-grounded research or notebook tool	The system searches or responds from identified web sources, uploaded documents, or a selected source collection
Embedded productivity feature	AI works inside a document editor, spreadsheet, slide tool, LMS, or research database
Reusable project or custom assistant	Instructions and selected reference material remain available across related work
Artifact or app builder	The conversation produces an editable document, visualization, interactive model, or small application
Agentic workspace	The system can complete several connected steps and may use files, tools, or approved actions with less turn-by-turn prompting

The Major Platforms

The major AI platforms offer overlapping capabilities such as text generation, image analysis, and file processing. The surrounding interface, account, enabled tools, and institutional agreement often matter as much as the model itself.

Major AI Platforms: Context and Features
Platform	Useful Context and Features
ChatGPT (OpenAI)	A general-purpose workspace with file uploads, image analysis, Projects, custom GPTs, and Canvas for drafting or editing in a separate workspace.
Claude (Anthropic)	Supports long-document work, reusable Projects, and Artifacts for building and sharing interactive tools. Claude for Education includes a Learning Mode that uses guided questioning.
Gemini (Google)	Connects with Google’s ecosystem. Depending on the account, Gemini can work in Workspace apps, Gems can be assigned in Google Classroom, and NotebookLM can create source-grounded materials.
Copilot (Microsoft)	Copilot Chat provides a general AI workspace. With eligible licensing, Copilot can also work inside Microsoft 365 apps such as Word, Outlook, and PowerPoint.

Major AI Platforms: Access and Starting Resources
Platform	Access Considerations	Start Here
ChatGPT (OpenAI)	The Free tier includes some reasoning and tool access, but message and feature limits change. Consumer and institutional accounts use different data controls.	Teaching with AI, OpenAI’s educator guide with ready-to-use prompts developed by Ethan and Lilach Mollick.
Claude (Anthropic)	Usage, Learning Mode, and sharing options vary by plan and institutional settings.	AI Fluency for Educators, a free Anthropic course built with university faculty. Learners can complete a final assessment for a certificate.
Gemini (Google)	Qualifying institutions can receive the standalone Gemini for Education app and NotebookLM at no cost. Integration across Workspace apps requires an eligible plan and administrator enablement.	Gemini for Education, Google’s current hub for Gemini, Workspace integration, and NotebookLM.
Copilot (Microsoft)	Copilot Chat requires an eligible education license and organizational account. Letting Copilot use institutional email, files, and other Microsoft 365 data requires additional licensing and administrator enablement.	Enhance Teaching and Learning with Copilot Chat, a free Microsoft Learn module (9 units) designed for educators.

No single platform is definitively “best.” Your choice may depend on the task, the sources the AI should use, what your institution provides, what the account can access, and what you need the system to create or change.

One practical consideration about student-facing tools: Marc Watkins describes students who avoided institutionally provided AI tools because they worried their use would be monitored or linked back to them. Some preferred consumer tools such as personal ChatGPT accounts. Watkins is describing what he has observed, not reporting a general finding about all students. Faculty designing required AI activities should consider whether students trust the account they are being asked to use.

The tool environment also changes fast. DeepSeek offers a useful dated example. In 2025, governments and universities, including James Madison University, placed different kinds of restrictions on the service. Congress introduced the No DeepSeek on Government Devices Act, but it had not become law as of July 2026. A Congressional Research Service summary distinguishes among app-store restrictions, government-device bans, agency restrictions, and proposed legislation. The episode shows that public availability and institutional permission are different questions.

Specialized Tools for Research and Teaching

Beyond general-purpose platforms, several AI tools focus on specific academic tasks.

Perplexity AI combines a language model with internet search and provides answers with footnoted citations. Its Academic source filter searches academic papers and databases, but the results can include material that has not been peer reviewed. This can be useful when you need a quick survey of recent literature, though you should still verify what each citation actually supports.

Most major platforms now offer a deep-research feature, although the names and access rules vary. In this mode, the AI spends several minutes searching, reading, and synthesizing sources before producing a structured report. Perplexity currently offers limited Research use on its free plan. Lily Lee compares research features across Perplexity, ChatGPT, and Gemini. Because access changes quickly, check the current plan information before building an activity around any one service.

Elicit helps with literature reviews by presenting research papers in interactive tables with selected information extracted. It also supports screening and report-building for larger review projects. Its plans and usage limits change, so the durable question is whether its structured workflow fits the kind of literature review you need to conduct.

Google NotebookLM lets you upload sources and then ask questions, generate summaries, create study guides, or produce audio and video overviews of the material. It can also generate infographics, slide decks, and data tables. The Discover Sources feature lets you describe a topic and locate relevant web sources, so you are not limited to materials you upload yourself. Stefan Bauschard describes how this changes the tool from “analyze what I give you” to “help me find and analyze.” Google has also connected NotebookLM with Google Classroom. Availability varies by account, age, region, and administrator settings.

Perplexity, deep-research features, NotebookLM, uploaded files, and course projects can all support source-grounded AI. Some chatbot conversations rely mainly on the model and the current conversation. Others can search the web, use uploaded files, retain project knowledge, or work with connected tools. A source-grounded workflow deliberately gives the AI something specific to consult: current web sources, uploaded course readings, a syllabus, an institutional policy, or a folder of project materials. Some tools use retrieval-augmented generation (RAG); others use related search and retrieval processes. Faculty usually need the habit more than the acronym. Before using one of these tools, ask what source the answer should be grounded in.

Studies of retrieval-augmented systems have found fewer hallucinations in some settings (Shuster et al., 2021). Source grounding does not eliminate unsupported claims, even when retrieved documents are available (Niu et al., 2024). The AI can misread a source, cite a passage that only loosely supports the claim, or choose sources that are technically related but not very useful for your course.

Organizing Your AI Workflows

As you use AI tools more regularly, you may find yourself repeating similar prompts or re-uploading the same reference documents. The major platforms offer features to organize and streamline recurring tasks.

These organizing features also help with source grounding. A project, Gem, GPT, or agent can keep configured instructions and selected reference material available across related work. Their memory, file limits, sharing, and available actions differ. For faculty, that might mean a course project with the syllabus, outcomes, assignment sheets, key readings, and department policies already attached. The AI is then responding with more of your actual teaching context in view.

AI Workflow Organization Features by Platform
Platform	Feature Name	What It Does
ChatGPT	Projects	A workspace for related conversations, project instructions, and selected files
Claude	Projects	A self-contained workspace with project instructions, chat histories, and a shared knowledge base
Gemini	Gems	Configured assistants with specific instructions and behaviors. With eligible education accounts and administrator settings, educators can create and assign Gems in Google Classroom
Copilot	Agents	Task-specific assistants whose instructions, sources, actions, and costs depend on the agent configuration

Course Project example. Imagine creating a project dedicated to a specific course. The project instructions tell the AI: “You are helping me develop materials for an introductory biology course. The audience is first-year undergraduates with no prior biology background.” You upload your syllabus and learning outcomes. Now every conversation in that project can begin with this context already available. This can reduce repeated setup, but you still need to check whether the response uses the course material accurately and appropriately.

For Gemini users, Eric Curts has built EduGems.ai, an external, school-oriented collection of education Gems. Its contents may change, but it gives faculty concrete examples to inspect. At BETT 2026, Google announced that educators with eligible accounts can create and assign Gems in Google Classroom, with templates such as Study Partner, Quiz Me, Brainstorm Partner, and Real-World Connector. Google is also adding engagement information for assigned Gems, so these tools are beginning to appear inside institutional learning environments.

ChatGPT also allows users on eligible paid plans to create and share custom chatbots called GPTs. A GPT can combine instructions, uploaded knowledge, selected capabilities, apps, or actions in a reusable format. If you find yourself using the same detailed prompt repeatedly, packaging it as a project or custom chatbot may reduce repeated setup.

Artifacts, Canvas, and Mini Apps

AI platforms can now create functional applications during a conversation. This capability has expanded since it first appeared.

Claude Artifacts started as a way to produce self-contained content such as quizzes, charts, and simple tools inside a conversation. Anthropic has since expanded Artifacts into app-building. Depending on the plan and settings, a published artifact can save text data across sessions. It can also connect to approved external services such as Google Calendar or Slack through a connection standard called the Model Context Protocol (MCP). Publishing and sharing options differ between personal and institutional accounts. Anthropic’s Artifacts area includes examples that users can inspect and customize. Simon Willison has used Artifacts for software prototyping, showing how far the feature has moved beyond static documents.

ChatGPT Canvas is OpenAI’s editing workspace for text and code. It functions more like a collaborative document, where you and the AI can revise material in a separate workspace. It can also render HTML and interactive code. Artifacts and Canvas overlap, but their interfaces and available features lead to different working experiences. Both let you create and test material within a conversation. Availability depends on the app, account, workspace controls, and compatible model, so check the current documentation before designing a required activity around it.

You might ask Claude to create an interactive quiz on a topic you’re teaching. The AI generates a working quiz you can test immediately and share with others. You might ask ChatGPT Canvas to help you refine a writing assignment while rendering a live preview. The results can be surprisingly functional, though they sometimes require iteration to get right.

This approach, describing what you want built through conversation with an AI, has come to be called vibe coding. The term was coined by Andrej Karpathy in February 2025 and spread so quickly that Collins English Dictionary named it the Word of the Year for 2025. Paul Ford wrote a New York Times op-ed in February 2026 explaining the term to a general audience.

In Spring 2026, Clemson University ran a Vibe Coding for Education course where students built educational games, chatbots, and interactive tools through prompts. These systems are designed to let people prototype without writing the code themselves, but the result still needs testing and may require technical troubleshooting. Building an artifact is an accessible starting point for trying this. Canva has also added Canva Code, which lets users generate interactive experiences inside Canva designs using plain language and gives educators who already use Canva a low-friction entry point.

Sarah’s Three Experiments

Dr. Sarah Chen is a composite instructor whose experience brings together several common faculty use cases. Sarah decided to move beyond chatbot conversations and test features she’d heard about but never tried. She set aside an afternoon and picked three tasks from her actual work.

Experiment 1: Deep Research. Sarah was updating her Unit 3 readings on argumentative writing and wanted to find recent scholarship on how community college students learn to construct counterarguments. She tried Perplexity’s Research feature. She typed her question, and the AI spent about three minutes searching, reading, and synthesizing sources before producing a structured report with footnoted citations.

The report gave her a useful overview. It surfaced several recent articles she hadn’t seen, including one on scaffolding counterargument instruction for first-generation students that was directly relevant to her course. But it also cited a few sources that turned out to be tangentially related. One addressed counterargument in competitive debate and did not apply directly to composition pedagogy. Two of the citations also had minor errors in the author names.

What she learned: for this task, deep research was useful for getting an initial survey of a topic, especially when she was exploring outside her primary area. But she still needed to verify every citation before relying on it. She now uses it as a starting point for literature exploration, with source checking as the required next step.

Experiment 2: NotebookLM. She uploaded three of her course readings into Google NotebookLM and let it generate an overview. The connections it drew between texts were sometimes useful and sometimes shallow. It correctly identified a shared concern with audience awareness across all three readings. But it also drew a connection between two articles based on a superficial keyword overlap (“voice”) that meant different things in each context. One was about authorial voice in writing. The other was about giving voice to marginalized perspectives.

She could see using NotebookLM to prepare for class discussion, especially when she wanted a quick refresher on how her readings relate to each other. The Discover Sources feature would have saved her time finding the readings in the first place. But she’d still need to verify its connections.

Experiment 3: Creating a Project. The Project experiment had the largest effect on Sarah’s routine. She created a Project in ChatGPT for her ENG 101 course. After confirming that the files contained no student information, she uploaded her syllabus, learning outcomes, and Unit 3 assignment sheet. Then she asked the AI to identify where the unit materials assumed background knowledge her students might not have.

A regular chat may use saved memories or prior conversation history, but a Project gives Sarah one deliberate workspace for this course. The syllabus, outcomes, assignment sheet, and instructions stay together. For Sarah, the difference felt less like explaining the course to a new colleague and more like talking to someone who already had the relevant materials in front of them.

The first question she asked inside the project was: “Where might students get confused in Unit 3, given my syllabus, outcomes, and assignment sheet?” The AI already knew her students were first-year composition students at a community college. It already knew Unit 3 focused on argumentative writing. Its suggestions referenced the actual course structure and pointed to places where students might need more setup. She didn’t have to re-explain anything. This was the feature that made her think she might actually use AI tools regularly.

After each experiment, Sarah used five questions to examine what had happened:

Task: What faculty job was she exploring?
Tool and feature: What kind of environment did she use?
Source base: What web sources, uploaded documents, or course materials was the system expected to consult?
Observed access and actions: What did the feature appear able to see, create, share, or change?
Available evidence: What output or observation could she inspect?

The questions do not settle whether a tool is accurate, institutionally appropriate, or worth adopting. They make the immediate comparison more concrete.

Try This: Explore a Platform Feature

Choose one of the following experiments based on what platform you have access to. Use nonsensitive material: do not upload student records, grades, private communications, unpublished personnel material, or other protected information.

If Projects are enabled in your ChatGPT account: Create a new Project for a course you teach. Add instructions that describe the course context, student audience, and your teaching goals. Upload a public syllabus or an openly licensed reading. Start a conversation and notice how the response uses the context you provided.

If Artifacts are enabled in your Claude account: Ask Claude to create an artifact, such as a flashcard quiz or an interactive timeline on a topic from your discipline. Test it and ask for one modification. Preview the sharing options, but publish or share the artifact only if the account and material are appropriate for that audience.

If NotebookLM is enabled for your Google account: Upload an openly available or openly licensed reading. Ask it to generate a summary, discussion questions, or a study guide. Then try Discover Sources on a topic you are researching. Notice which sources it selects and how the generated material reflects them.

If Gems are enabled in your Gemini account: Visit EduGems.ai and try one of the pre-built education Gems. This is an external, school-oriented collection, so its contents may change. See how one Gem responds to a low-risk teaching task and notice what you would need to change for your own context.

If none of these specialized features is available: Use a general-purpose chatbot. Paste a short openly licensed reading or public syllabus excerpt into the conversation, ask the chatbot to produce one teaching resource based only on that material, and then request one revision. Notice how the supplied source and follow-up instruction shape the result.

After the experiment, use Sarah’s five questions to compare the task, tool and feature, source base, observed access and actions, and available evidence.

Watch: Getting Started with Projects and Gems

This tutorial demonstrates how to create a course workspace in ChatGPT, Claude, and Gemini. As you watch, notice that all three use a clear course purpose, instructions that define the audience and teaching priorities, and attached documents that supply course context. Each platform uses different menus. In all three, you decide which course information should guide future conversations.

https://www.youtube-nocookie.com/embed/e3Cg751ewMw

Watch Getting Started with Projects & Gems on YouTube.

A project can keep course instructions and attached documents available across conversations. Check whether each response actually reflects the syllabus, outcomes, readings, or assignments you supplied.

Looking Ahead

With a sense of the tools available and how they work in practice, the next chapter, Prompting for Educators, applies prompting to recurring faculty tasks. You will see how AI can support course materials, formative feedback, student-support resources, and discipline-specific examples. You will also learn why early adoption can feel slower before it becomes useful.

References

Abimbade, D. & Wetsel, H. (2026, January). Vibe coding for education. Clemson Online. Source link

Alexander, B. (2025, December 31). AI and higher education in 2025. AI and Academia. Source link

Anthropic. (2025). AI fluency for educators. Source link

Anthropic. (2026). What are artifacts and how do I use them? Claude Help Center. Source link

Anthropic. (2026). Artifacts. Claude. Source link

Bauschard, S. (2025, May 5). Significant new Google announcements for educators. Education Disrupted. Source link

Canva. (2026, July 14). Canva Code 2.0: Build powerful websites, apps, and forms without the limitations. Canva Newsroom. Source link

Collins Dictionary. (2025, November). Word of the Year 2025: Vibe coding. Source link

Congressional Research Service. (2025). DeepSeek: Background, security, and policy considerations. Source link

Congress.gov. (2025, February 27). S. 765: No DeepSeek on Government Devices Act. Source link

Curts, E. (2026, February 10). 100 of the best Gemini Gems for schools. Control Alt Achieve. Source link

Elicit. (2026). Elicit: AI for scientific research. Source link

Ford, P. (2026, February). My New York Times op-ed on vibe coding. Aboard. Source link

Google. (2025). Gemini for Education. Source link

Google. (2026, January). New ways Gemini can help in Google Classroom. The Keyword. Source link

Google. (2026). NotebookLM Help. Source link

James Madison University. (2025, February 12). Executive order bans the use of DeepSeek. Source link

Karpathy, A. (2025, February 2). Vibe coding [Post]. X. Source link

Lee, L. (2025, February 28). Deep research & AI in education: A game-changer? AI x Education. Source link

Long, L. (2024, October 26). My three chatbot interns. Artisanal Intelligence. Source link

Microsoft. (2025). Enhance teaching and learning with Microsoft 365 Copilot Chat. Microsoft Learn. Source link

Niu, C., Wu, Y., Zhu, J., Xu, S., Shum, K., Zhong, R., Song, J., & Zhang, T. (2024). RAGTruth: A hallucination corpus for developing trustworthy retrieval-augmented language models. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics, 10862-10878. Source link

OpenAI. (2023). Teaching with AI. Source link

OpenAI. (2026). What is the canvas feature in ChatGPT and how do I use it? OpenAI Help Center. Source link

OpenAI. (2026). Projects in ChatGPT. OpenAI Help Center. Source link

Perplexity. (2026). What is Perplexity? Perplexity Help Center. Source link

Shuster, K., Poff, S., Chen, M., Kiela, D., & Weston, J. (2021). Retrieval augmentation reduces hallucination in conversation. Findings of the Association for Computational Linguistics: EMNLP 2021, 3784-3803. Source link

Watkins, M. (2026, February 15). Before you buy AI for your campus, read this. Rhetorica. Source link

Willison, S. (2026, February 20). Prototyping with Claude Artifacts. Simon Willison’s Weblog. Source link

License

Icon for the Creative Commons Attribution-NonCommercial 4.0 International License

A Guide to Teaching and Learning with Artificial Intelligence Copyright © by Jason Blomquist; Liza Long; and Joel Gladd is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License, except where otherwise noted.