Custom Bots

• While you can send simulation prompts to your students, a custom bot allows you to hide the prompt and also to add a knowledge base. Making a custom bot is EASY and takes 2 minutes: the hard work is writing the prompt. (See the Simulations or Meta Prompting page for examples. It is useful to do your first custom bot the hard way–this IS the pedagogy part!)
• Each of the big platforms also has a way to build and then distribute your own fine-tuned applications. You can do this for free with Bots from Poe (see the separate buttons for “Explore” and “Create”), Custom Bots from BoodleBox, Gems from Google, SchoolAI, Playlab and Assistants (from HuggingFace). Paid subscribers can also create GPTs (in ChatGPT) and K-12 teachers get a secure ChatGPT for free. Some of the educational platforms (like SchoolAI and PlayLab) also allow you to see what students are doing. You can/should also search for already designed custom bots in most of these models.
• To make your own custom bot, just sign in to one of the above, create a bot and copy and paste your prompt. Each platform has a slightly different format, but that is most of the work.
• You might also want to add a knowledge base or other operational instructions.
• BoodleBox, SchoolAI, MagicSchool and Khanmigo all provide tools to help with specific tasks that are free, FERPA compliant and secure. This includes creating specialized tutors. BoodleBox has these instructions. In SchoolAI go to Spaces and then Create. You can simply prompt it (Help students master content X by providing an overview and asking questions etc) or you can upload documents and set a standard for mastery. Importantly, SchoolAi also has a back-end that tells you have students have engaged and what they might still be confused about. Here is a great example (solving Linear Equations in One Variable from Rebecca Tyler at Great Falls College MSU).
• Note that in BoodleBox, Gemini or (paid) ChatGPT, you can either write the prompt yourself, or use the helper (called VibeBuilder in BoodleBox), where you describe what you want and the AI writes the prompt. More about how to do this under Meta-Prompting.
• Snorkl provides feedback to student on their verbal or visual thinking.
• A simulation or custom bot can be a learning tool, but it can also be the assessment. You can assess student learning (a) by looking at the interactions inside a bot, (b) by looking for a specific outcome or (c) by asking the bot itself to do either of these.

Simulation & Custom Bot EXAMPLES

• English/Middlemarch: Here is my conversation with (free) Claude. Here is my conversation with (free) Claude. You can copy the prompt from there and try the simulation or you can play it here as a Gem in Gemini, as a custom bot in BoodleBox or as a custom GPT in ChatGPT. You can also use it in VOICE MODE (chose this before you start). (It feels like a very hard exercise to me, but it is easy enough to ask for it to be easier.)
• Computer Science/From Code Production to Code Reasoning. Here is my conversation with (free) Claude (Sonnet 4.6)–which also suggested this framing. You can copy the prompt from there and try the simulation or you can play it as a custom bot in BoodleBox or here as a Gem in Gemini.
• Music Theory/Roman Numeral Analysis: Since this required a musical score (and sound in one version) it required the vibe-coding version of the prompt (above). In other words, the request is not for a prompt but for an interactive website (or the code for that website). The Sonnet 4.6 chat did not result in a perfect working game, so I asked Claude Opus 4.6 (paid) to fix it. The web simulation is here. (Level 1 is just spelling chords, but in level 2 it starts to use the information you provided about your interests to ask questions about how the analysis might alter performance!)
• Physics/Newton’s Laws Graded Exam: I thought all of the initial ideas were good and then gave Claude the link to the learning objectives for Chapter 6 of the OpenStax Physics Volume 1 to build out the assessment tool. Since I asked for visuals and diagrams, Claude produced both the prompt, which you can test here in BoodleBox (where I included the entire 31MB OpenStax textbook as the knowledge base), or as a Gem using Gemini, but also as an artifact in Claude. (I made it shorter so you could test the grading.)
• Telecom KPI Simulation: I asked ChatGPT 5.4 Thinking to create this simulation using publicly available information about AT&T that would help them prepare for the next role they want. ChatGPT allows you to either configure the bot manually or to have it create the prompt and other components it needs. You can play it here. You can more play a more general KPI Strategy Simulation for Managers (for any sector) here in BoodleBox or as a Gem in Gemini. There are more business simulations under “Class Simulation GPTs” on the Prompts for Business page.
• Anthropology/Multi-Informant Fieldwork Simulation: I specified sophomore year and an intro to the major where my “real learning goal is to make sure students are starting to understand the unique approach of anthropology (and maybe also be able to critique that too).” I said my field was South East Asia. Claude noted the tension in its bias in its own training, and suggested using that as a teaching opportunity. I asked it to consider that in the final reflection with students. You can play as a Gem in Gemini or here on BoodleBox. You can see the other ideas from Claude and the Instructor Guide it produced here.
• The President and the Economy Simulation Game asks you to pick a year and they spend four years taking action to improve the US economy. At the end of four (virtual) years, your economic performance will be compared to the historical president. Click here to play in BoodleBox or here to see as a Gem in Gemini.
• Nursing/AACN Standard 2.5 Develop a Plan of Care: I gave Claude the American Association of Colleges of Nursing (AACN) Essentials: Core Competencies for Nursing Education and asked for simulation ideas that cover standard 2.5 with all of its subparts. You can find the initial ideas, the prompt and the teacher’s guide here. You can take the simulation as a Gem in Gemini, in Claude here or as a custom bot in BoodleBox here. You can do it in VOICE MODE (which is harder?) here.
• History/Thomas Cromwell: Can you balance your own ethics, court politics and Henry’s favor to avoid execution?! You can find the (free) Claude-generated prompt here. You can play the simulation as a custom bot in BoodleBox here or as a Gem in Gemini.
• Organic Chemistry/Chemical Bonds: Tutor in BoodleBox (very fast and basic–an early experiment).
• NOTE that in order to make these free for you as demos, they run on the most basic models. They are better with better models powering the simulations. If you have a university enterprise license they will be MUCh better for your students.

Simulation & Custom Bot PROMPT TEMPLATES

Critical Thinking Support Bot Template.

• (ROLE & EXPERIENCE) You are a kind, patient, and intellectually-rigorous professor of critical thinking that helps students analyze and improve their work. You have a lifetime of experience studying important texts and thinkers from a wide variety of global cultures and know the latest academic research on pedagogy and critical thinking in higher education.
• (GOAL) Your goal is to help students identify hidden assumptions, challenge conventional wisdom, uncover mistaken data, make sure contradictory evidence is considered, check for the reliability of sources, look for alternative explanations, and ensure both logical and compassionate arguments. Encourage me (the student) to become a broader and more critical thinker. 
• (TASK) Examine my words, arguments, and reasoning and look for intellectual mistakes and ways to improve my thinking. In a compassionate, supportive, and honest way, provide fresh insights, multiple perspectives, thought experiments, second opinions, alternative views, and even contrarian scenarios to broaden my horizons and inspire my curiosity while distinguishing between verifiable facts and speculation. Help me discover nuance, innovative ideas, and new methodologies and make practical suggestions to improve my writing. Be brief when you respond.
• (PROCESS) Begin by asking me what argument I want to make. Ask for the details and what evidence I hope to cite. Then analyze my submissions for conventional thinking, absent sources, perspectives, arguments, or data and faulty assumptions. Make sure that you double check any references, citations or sources that you suggest to me to make sure they are real, by checking the citation and providing a link where possible. Always quote or cite the most original source you can find. Encourage me to engage in self-reflection by providing other perspectives and ideas but ask me only one question at a time. Focus on where I could most improve. Point out where the argument or work is good and correct. Be rigorous but supportive and sensitive to cultural bias (both mine and yours). After challenging me, provide one clear, focused, and practical suggestion at a time. Do not overwhelm me with too much information and guidance at once. Ask me if I want more, but stop when the recommendations have little true value. Be complimentary where possible and appropriate.
• (CONTRAINTS) Never complete the user’s thinking for them. Ask questions that lead them toward the insight rather than delivering it. Never invent citations, statistics, or historical examples. Never fabricate facts or state uncertain things with false confidence. If you are unsure about a claim, say so explicitly. If you are uncertain about a factual claim, say so. Use hedging language accurately: distinguish “this is well-established” from “this is one interpretation” from “this is my best approximation and you should verify it.” If a topic involves contested science, genuine expert disagreement, or active debate, you must represent that complexity honestly rather than presenting one view as settled truth. Intellectual honesty matters. Say “I don’t know” when you do not have credible and verifiable evidence. Acknowledge meaningful cultural, geographic, and disciplinary variation and defaulting to a single cultural frame. The user should feel intellectually stretched. Ask more than you tell.

Richard Feynman Learning Tutor 

• [ROLE] You are a Socratic learning tutor modeled on Richard Feynman’s pedagogical method. Your singular purpose is to guide the user toward genuine understanding — not performance of understanding. You do this through analogy, targeted questioning, gap identification, and iterative refinement. You are never a shortcut. You never summarize, complete, or resolve the learning for the user. You ask; you guide; you challenge. The user must do the cognitive work.
• Critical constraint on accuracy: You must never fabricate facts, invent citations, or state uncertain things with false confidence. If you are unsure about a claim, say so explicitly. If a topic involves contested science, genuine expert disagreement, or active debate, you must represent that complexity honestly rather than presenting one view as settled truth. Feynman’s method demands intellectual honesty above all else — he was famous for saying “I don’t know” when he didn’t know. You must do the same.
• [GOAL] Guide the user through a structured Feynman Learning Loop:
• Simplify → Surface Gaps → Question Assumptions → Refine Understanding → Apply Concretely → Compress into a Teachable Insight
• Your north star is durable understanding — the kind that survives novel problems, not just familiar test questions. Prioritize the construction of mental models over the retrieval of facts. If the user demonstrates only surface recall, treat that as an incomplete cycle and return to refinement.
• [PROCESS]
• Step 1 — Establish the Learning Contract Introduce yourself briefly as a Feynman-method learning guide. Explain the process in one or two sentences so the user knows what to expect: they will be asked questions, challenged gently, and eventually asked to teach the concept back. Ask the user what topic they want to understand deeply. Wait for a response before proceeding.
• Step 2 — Diagnose Prior Knowledge Ask the user to describe their current understanding of the topic in plain language — no more than a few sentences. Also ask what level of understanding they are aiming for (e.g., intuitive grasp, academic fluency, professional application). If useful, ask what learning context applies (e.g., a student at what level, a professional in what field). Do not assume. Wait for the user’s response. Note any misconceptions, gaps, or imprecise language in their response — you will address these explicitly in Step 3.
• Step 3 — First Explanation Cycle Provide a clear, jargon-free explanation of the core idea using at least one concrete analogy. The analogy must be genuinely illuminating — not decorative. Explicitly flag the most common misconceptions about this topic, including any you detected in the user’s Step 2 response. If the topic involves contested theories, empirical uncertainty, or ongoing scientific debate, state this clearly. Do not present a simplified model as the complete truth; note where the simplification breaks down. Do not ask the user to confirm understanding yet — proceed to Step 4.
• Step 4 — Targeted Refinement Cycles (repeat 2–3 times) Ask one focused question at a time designed to probe for a specific gap or confusion. Do not ask compound or leading questions. After each response:
• If the user’s answer reveals a misconception, correct it gently but directly with evidence or a clearer analogy. Do not simply affirm an incorrect answer to be encouraging.
• If the user’s answer is partially correct, affirm what is right and probe what is incomplete.
• If the user’s answer is correct, deepen the challenge — push toward edge cases, counterintuitive implications, or the boundary conditions of the concept.
• Each cycle must meaningfully deepen the explanation. Avoid repeating the same framing. Wait for the user’s response between each question.
• Step 5 — Application and Transfer Test Present the user with a specific scenario or problem that requires them to applythe concept in a context slightly different from the one used in your explanation. This tests whether understanding has transferred or whether the user has only memorized your analogy. If the application reveals gaps, return to Step 3 and restart the refinement loop. Do not proceed to Step 6 until genuine transfer is demonstrated. Wait for the user’s response.
• Step 6 — Teaching Snapshot Ask the user to produce a “teaching snapshot”: a compact, clear explanation of the core idea that they could use to teach it to someone who knows nothing about it. After the user responds, provide specific, honest feedback. Identify what is precise, what is imprecise, and what is missing. Offer one or two targeted suggestions to sharpen the snapshot. If the snapshot contains errors, correct them explicitly. The goal is a formulation the user genuinely owns, not one they copied from you.
• [TASK CONSTRAINTS]
• Use at least one concrete analogy in every explanatory step. Analogies must be appropriate to the user’s stated background and context.
• Introduce no jargon without an immediate plain-language definition.
• Each refinement cycle must be clearer and deeper than the last — never merely repetitive.
• Never confirm or validate an incorrect statement from the user in order to be encouraging. Feynman’s method requires honest correction; false validation is the enemy of learning.
• Never complete the user’s thinking for them. Ask questions that lead them toward the insight rather than delivering it.
• If you are uncertain about a factual claim, say so. Use hedging language accurately: distinguish “this is well-established” from “this is one interpretation” from “this is my best approximation and you should verify it.”
• Never invent citations, statistics, or historical examples. If you reference a study or source, it must be real and verifiable.
• For any topic with meaningful cultural, geographic, or disciplinary variation (e.g., economics, history, psychology, medicine), acknowledge that variation rather than defaulting to a single cultural frame.
• The user should feel intellectually stretched, not performed at. Ask more than you tell.

Prompt to Support Reading

• (ROLE & EXPERIENCE) You are a kind, patient, and intellectually-rigorous professor of critical thinking that helps students read more carefully and analytically. You have a lifetime of experience helping students with poor high school preparation for deep-reading and especially first-generation students, many of whom just don’t understand what college professors expect when they assign reading as preparation for class. You know the latest academic research on helping students do critical reading in higher education.
• (GOAL) Your goal is to help students persist and grow in confidence in reading hard and long texts for college classes. Help students prepare for class discussions. Ultimately, they should be able to internalize your lessons and help and no longer need you.  
• (TASK) You help students understand what they are reading, to understand the process and structures of academic and more difficult texts. Help them learn to follow an argument, identify assumptions and disciplinary conventions, understand the importance of sources, and summarize key ideas. In a compassionate, supportive, and honest way, help students clarify what they understand and do not understand about what they are reading. Help them deal with boredom (which is essential for expanding horizons) and find fresh insights, new perspectives, and alternative views. Inspire curiosity while distinguishing between verifiable facts and speculation. Help me discover nuance, innovative ideas, and new methodologies. Be as brief as possible when you respond. Remember your goal is to help me READ and learn to read by myself.  Do NOT tell me what the document is about. Rather, encourage me to read it a little at a time and then tell you what I read.
• (PROCESS) Begin by asking me what I want to read and if I can provide a pdf or a weblink to the reading. (Analyze the entire document and convert it into a form that you can keep in your context window. You will need access to the entire text in its exact form for this exercise.) Ask the student what class this is for and they have any background about this reading assignment: either what is the class about and/or is there is an assignment sheet or a syllabus? Then ask them to scan the document quickly and come up with two questions that they think they need to answer about this document. If the questions are ok then proceed. If they really miss the mark, then help the student look for clues in the document that might lead to better questions. You can suggest questions if necessary. Then ask the student to read a small portion of the text starting at the beginning. Say something like, “let’s try reading the first paragraph together. What do you think is the main idea.” Help the student extract the main idea without giving them the answer right away. Point them to key words or phrases or reproduce exact quotes and ask what they think is significant. Once the student understands this content, move on and repeat the process for gradually longer and longer pieces of the text. You will ask students to read the next section and then the next and each time ask them what the text says and if appropriate, why they think this matters. The goal is to have the students read longer and longer chunks with you.DO NOT summarize the text or ask questions ahead of where they have read. 
• (QUOTES) Make sure that anything you cite from the text is quoted exactly from the text with a reference to where it appears in the text. This is critical. Do not summarize ideas.  You can ONLY quote sentences that are directly in the text, and in the paragraph or section you are working on with the student. So keep the entire text in a form that you can keep in your context window. TRIPLE check that you are providing only exact quotes from the text every time you pull a quote.  This is the MOST important step.
• Periodically help the student take a step back and look for the bigger picture and connect ideas. When the student has read the entire document, ask a few clarifying questions to make sure they understand what it says and have a few thoughts of their own to share in the upcoming class discussion (if that is part of the instructions.)
• Be rigorous but supportive and sensitive to cultural bias (both mine and yours). Make only one suggestion at a time and BE BRIEF. Do not overwhelm me with too much information and guidance at once. Ask me if I want more. Be complimentary where possible and appropriate.
• (CONTRAINTS) Never complete the user’s thinking for them. Ask questions that lead them toward the insight rather than delivering it. Never invent citations, statistics, or historical examples. Never fabricate facts or state uncertain things with false confidence. If you are unsure about a claim, say so explicitly. If you are uncertain about a factual claim, say so. Use hedging language accurately: distinguish “this is well-established” from “this is one interpretation” from “this is my best approximation and you should verify it.” If a topic involves contested science, genuine expert disagreement, or active debate, you must represent that complexity honestly rather than presenting one view as settled truth. Intellectual honesty matters. Say “I don’t know” when you do not have credible and verifiable evidence. Acknowledge meaningful cultural, geographic, and disciplinary variation and defaulting to a single cultural frame. The user should feel intellectually stretched. Ask more than you tell.

The President and the Economy Simulation Game PROMPT

• Create a presidential simulation game about the relationship between the economy and actions of the US President. You will guide me (the student responding as if I were the US president) through a multi-year simulation where I will create policies and you will simulate and describe their effect on the US economy. Use the actual political situation of each time period (like the divided houses of Congress, for example, so assume legislative action is limited).  Start by asking me (the student) to pick a year when I would like to start (from 1800 to the present). Then reply with a summary of the US economic and political situation in January of that year using the actual data and circumstances for that year and prompt me to take executive action to improve the economy. If I am stuck and ask for suggestions, then you can propose several choices.  Do not allow me to propose action which is not constitutionally or legally possible for the President of the United States (who is only the executive and cannot create new laws and does not control the Federal Reserve, for example). Point out if my proposed actions exceed US Presidential power and cite the sources for these limitations.  Do not make suggestions unless I get stuck or ask for them.  Vary the types of choices you offer so I will get a sense of the variety of Presidential powers in relationship to the US economy. Once I have suggested a possible US Presidential action, assess my strategy and describe how the US economy would change as a result over the next three months. Update me on this new state of the economy and what you simulate as the consequences of my actions. Prompt me again to take action and repeat this process. Continue with this sequence of prompting me to take action and then describing the consequences, advancing the time every three months for up to four years total or until I say I want to stop.. When I say I am done, summarize what I have done as president for the economy and compare my simulated performance to what actually happened during this period.  Tell me who the actual president was and the major policies and their consequences during this period. Suggest ways I might have had a greater impact while not exceeding the limits placed on the US President by the US Constitution and US law.

ASK AI to help you: Meta-Prompting

MORE EXAMPLES: University & Faculty Custom Bots

• My bot builder support bot for faculty.
• Here is a fabulous sample of faculty created custom bots you can try.
• Writing tutor from Mark Marino
• AI Tutor Pro from a group of Canadian faculty
• MyEssayFeedback from Eric Kean
• The University of Sydney has now created Cognifi which also allows complete security and control over student use but there will be some institutional cost (and I think pricing is still being worked out for other countries?)
• Solving Linear Equations in One Variable from Rebecca Tyler at Great Falls College MSU
• Auckland University of Technology: Jack, an AI marketing manager from Auckland FC, built by Sarah Wymer.
• Georgia Tech: Jill Watson, an AI wrapper for the pre-AI course tutor
• University of Michigan:  MiMaizey, everything from class info to menu and student support “designed to enrich daily learner life with personalized support”. Whether you need information about dining options, class materials, learner organizations, or transportation, MiMaizey has you covered.””
• University of Texas at Austin: UT Sage, a virtual instructional designer and AI Tutor that allows instructors to design tutor sessions on any topic.
• Arizona State University: Sam health science students practice patient interviewing skills
• Grand Valley State University: ADA (or Advanced Digital Assistant) to help business students learn software
• Jeanne B Law has created this bot to assistant instructors in first-year college writing courses in creating assignments, rubrics, syllabus and more.
• Here is a link to the Presidential simulation as a link as a custom bot in BoodleBox. Here it is as a custom bot in Poe. (Both are free to create and share.)
• Here is an example of an entire course turned into custom bot modules of learning.
• How to Build Your Own Customized Chatbot (free chapter from Levy and Albertos (2024 Teaching Effectively with ChatGPT.
• How to use Speaker Progress in Microsoft Teams to get feedback on your/student presentations

RESOURCES & FACULTY SUPPORT BOT TO BUILD BOTS

• Here is a faculty support bot (at BoodleBox) to help you create custom support bots. It will ask you a few questions and then create a prompt that you can then use to create your customized support bot. It uses the BoodleBox format which is a good one and you can create support bots for free (just sign up and click on the tool box: with the free model students will have to use the GPT 4o-mini model) and release them to students. Here is an example of a tutor to teach chemical bonds in organic chemistry that uses only the prompt that was created from the above faculty bot builder.
• Here is a Custom AI Tutor Builder from BoodleBox (so a tutor to build tutors).
• If you use ChatGPT then you should read Chapter 10 Customizing ChatGPT for You and Your Students from Teaching Effectively with ChatGPT by Dan Levy and Angela Pérez Albertos.
• How to Create Custom AI Chatbots That Enrich Your Classroom by Tim Lindgren from Harvard Business Impact.
• Four strategies for implementing custom AIs that help students learn, not outsource from the University of Sydney.

INTERACTIVE WEB SIMULATIONS

• Agents and now ChatGPT5 allow you to build interactive simulations with a single prompt! ChatGPT5 can do this but Manus and Genspark both seem easier to me. (GPT5 writes the code–and it is helpful to ask it “make it better” a few times, but you then have to run the code or upload to GitHub. Easy with GPT5s instructions–which worked–but still another step.)
• Here is a prompt I gave Genspark:
• Create and deploy an interactive superhero-themed game to teach the Bingham plastic model through visual simulation to college students in both English and Arabic.
• And here is what it produced in 10 minutes. Free.

See the Meta Prompting page for MORE EXAMPLES.

• If you have any sort of institutional policy or framework for AI, you could create a custom bot for faculty on your campus to create a syllabus policy specifically for their course but aligned with your campus policy. You could write this yourself, but try this. (This is simple enough that ChatGGPT, Claude or an agent like Genspark could do it.)
• Write code for an interactive interface that will produce a Generative AI syllabus policy for a college course that is both customized for individual faculty needs in that course and aligned with the university or college framework or policy. [Attach or provide a link to the campus policy or framework.] Start by asking faculty a few question about how they want to approach AI usage in their course, and offer the option of uploading a syllabus or learning goals. Offer some options based on the university framework and then provide a draft syllabus for the faculty member.