AI Twin for Professors: What Your Twin Should Answer, What Stays Off Limits, and How to Set One Up

An AI twin for a professor is a first-person AI version of you that students can question at any hour, answering with your course facts and your way of explaining. A generic chatbot answers as itself and guesses at your policies, while your twin answers as you. Faculty also search for this as an AI teaching assistant of yourself, a custom AI for a course, AI office hours, or an AI clone of a professor. A campus course tool covers one class, while your twin speaks for you across your intro course, your seminar and your lab.

Timeless AI™ Twin answers students in the first person, in your own words, on three doors: your website (one script tag), an API, and AI assistants over MCP (the Model Context Protocol), which lets other AIs ask your twin questions and read who you are. A badge names your twin, and when a student asks directly, your twin says in one line that the student is talking to your twin. The Timeless AI™ Twin guide walks through each door, and our page on professional AI makes the wider case for experts.

Georgia Tech, Harvard, the University of Michigan, Morehouse College and the California State University system all put AI assistants in front of students, and the best measured results came from assistants held to material their instructors controlled.

For a single professor, Georgia Tech's result matters most: restricted to validated course material, Jill Watson answered correctly 78.7% of the time where OpenAI's general Assistant managed 30.7%. That gap is the case for a twin built from your own answers rather than a chatbot guessing at your syllabus. Harvard's CS50 adds the policy lesson, banning ChatGPT while allowing its own duck, with the stated goal of a 1:1 tutor for students 24 hours a day (CS50, 2025).

Your twin should handle questions with one right answer you have already given, and stay away from anything that judges a student or touches a student's record.

Sort each question by whether every student would get the same answer from you. If so, write that answer into your twin once and stop retyping it. If the answer depends on one student's grades, circumstances or wellbeing, your twin hands the question to you or the right campus office.

Yes. FERPA does not ban AI tools; it limits how schools disclose education records, and a twin built from your syllabus and explanations needs none of them. FERPA (20 U.S.C. § 1232g, with regulations at 34 CFR Part 99) applies to schools that receive funds under any program administered by the US Department of Education (34 CFR 99.1). At college the rights belong to the student, passing from parents when a student turns 18 or enters a postsecondary institution at any age (US Department of Education, Student Privacy Policy Office).

For outside tools, the clause that matters is the "school official" exception. A school may treat a contractor as a school official only if the contractor performs a service the school would otherwise use employees for, is under the school's direct control with respect to the use and maintenance of education records, and follows the redisclosure rules (34 CFR 99.31). That is a contract your institution signs, not a decision a professor makes alone, so keep records out of your twin entirely:

• Never add grades, rosters, student ID numbers or accommodation letters to your twin's memories.

• Tell students to keep grades, ID numbers and health details out of the chat, because twin conversations are stored.

• Keep feedback on individual work in your learning management system (LMS), such as Canvas or Blackboard.

• Ask your registrar or IT office before you wire any system that holds student records into your twin's API.

Your twin needs your course facts, your explanations and your boundaries, written the way you talk in office hours. Start with the questions you answer most often, because students ask those first.

• Logistics: dates, exam format, submissions, and late and regrade policies.

• A reading map: which chapter, paper or slide deck covers each week.

• Core explanations: your go-to analogy for each hard concept, and the misconception you correct every term.

• Boundaries: what your twin will not do, and where those questions go.

• Your booking link: the office-hours page you already share, such as Calendly or Microsoft Bookings.

• Who you are: your research, your lab, and whether you take undergraduate researchers.

Example (hypothetical): a statistics professor adds a memory that explains p-values with the coin-flip setup from lecture. A student who asks at midnight why a p-value of 0.04 is not a 96% chance the effect is real gets that coin-flip framing, not a textbook definition. Write one memory like that for each concept students trip over.

When a student speaks to your twin on your website, your twin can reply in your own cloned voice, while typed questions get text. Persona Video turns one 30-second recording into video messages of you, in your face and your own voice, for a week-one welcome, and your first Persona Video is free. Creating your voice is free too.

Your twin can run the part of office hours that is really a queue of repeat questions, and send everyone who needs you to your booking link. Harvard's CS50 staff put the case plainly in their course notes: with the duck, students who hit a wall do not have to wait for support staff to be available (CS50, 2023). Your twin also answers at 2 a.m. before a deadline, so your own hours go to the students who need a person.

AI assistants connected to your twin's MCP door can ask your twin questions and read who you are, as our AI twin MCP server guide explains. The same twin answers prospective graduate students about your lab and clients if you consult, which our guides to licensing your expertise as an AI and monetizing your expertise with AI cover.

Build your Persona free, add your course in your own words, switch on Timeless Pro, and paste one script tag on your course site. The first three steps cost nothing, so you can talk with your Persona and check your answers before you pay.

1. Build your Persona free. No card: 25 memories, up to 5 Trusted Contacts and Executor Lock™ setup, and your free build never expires.

2. Add your course as memories. Use the checklist above, in the words you use with students.

3. Create your voice. Record your own voice so your twin can answer spoken questions in your cloned voice; creating your voice is free.

4. Switch on Timeless Pro. $59.99 a month or $599 a year for everything in Eternal plus Timeless AI™ Twin (see pricing).

5. Add your booking link. Use the office-hours page you already share.

6. Test as a student. Ask the ten questions you get most, and add memories wherever an answer is thin or wrong.

7. Open the website door. Paste one script tag on your course site or faculty page, link that page from your LMS, and add the syllabus statement below.

8. Set your long-term choices. Name your executor and Trusted Contacts, and choose which twin doors keep answering after a verified death; each door starts switched off.

Step 8 runs on Executor Lock™, the patent-pending governance layer from Idy®, which has 78 patents pending. After a verified death, your Persona is locked as a perfect snapshot that nobody can edit, retrain or use to impersonate you, and your twin passes to your executor.

Timeless AI™ Twin comes with Timeless Pro at $59.99 a month or $599 a year, after a free build you can test first. Whether that pays off depends on how much of your week goes to repeat questions.

Example, with assumed numbers: you teach two sections with 150 students in total and get 30 logistics emails a week at three minutes each. That is 90 minutes a week, 22.5 hours across a 15-week semester, and 45 hours across two semesters. On the annual plan, $599 divided by 45 hours is about $13.31 per hour of your time returned. If your twin absorbs only half of those emails, the cost is about $26.62 per hour, so weigh that against what an hour of your research or teaching time is worth. Ask your chair whether professional development or course funds can cover Timeless Pro, and bring this math with your own numbers.

Use your campus tool if your school licenses one and course facts are all you need. Pick Timeless AI™ Twin when you want students to get your own answers, in your voice and with your booking link, across every course you teach.

Our verdict: build your own only with a research team behind you, and keep custom GPTs for private experiments, since students have to go into ChatGPT to use one. For first-person answers on your own site, Timeless AI™ Twin is the stronger pick, and Timeless Pro costs less than Delphi's Builder plan. Compare the do-it-yourself route in how to make a custom GPT of yourself and our Delphi alternative page.

Tell students what your twin is, what your twin can and cannot do, that conversations are stored, and that the syllabus wins any conflict. Copy this wording, then edit the name and links.

Sample syllabus statement

AI twin of your professor. This course has an AI twin of me, Professor [Your Name], built from my own course materials and the way I explain things. You can ask my twin about deadlines, policies, readings and core concepts at any hour, by typing or by speaking on the course website, and my twin can send you to my office-hours booking page. My twin does not grade, cannot see your grades or records, and cannot grant extensions or exceptions, so email me about anything that affects your grade. Do not type grades, student ID numbers, health information or other private details into the chat, because conversations are stored. If an answer from my twin conflicts with this syllabus or with what I said in class, the syllabus and I take priority. My twin is labeled as an AI twin, and if you ask, my twin will tell you so.

Put the statement next to your AI-use policy, mention your twin in the first lecture, and keep the wording in line with your institution's generative AI guidance. On your LMS home page, two lines are enough: what students can ask your twin, and that your twin does not grade or see their records.

• Georgia Tech via Newswise: Georgia Tech's Jill Watson Outperforms ChatGPT in Real Classrooms

• CS50 at Harvard: Artificial Intelligence lecture notes, CS50x 2025

• CS50 at Harvard: Artificial Intelligence lecture notes, Fall 2023

• Harvard Gazette: Professor tailored AI tutor to physics course, engagement doubled

• BestColleges: University of Michigan to Introduce 2 New AI Assistants This Fall

• Morehouse College: AI Teaching Assistants

• CSUN Newsroom: California State University to Become First AI-Powered University System in US with OpenAI

• Pew Research Center: About a quarter of US teens have used ChatGPT for schoolwork

• US Department of Education, Student Privacy Policy Office: What is FERPA?

• Cornell Law School LII: 34 CFR 99.1, to which institutions the FERPA rules apply

• Cornell Law School LII: 34 CFR 99.31, disclosure without prior consent

• Delphi: Pricing