UC San Diego is more direct about AI than most schools. Its Academic Integrity Office doesn't hedge: unless your instructor has said yes, don't use generative AI on graded work. That one rule shapes everything else in this guide.
Below is what the official pages say, how to read them as a student, and which models are worth your time for the kind of work UCSD is known for.
What UC San Diego actually says
We read the university's own pages in October 2026. Three are worth bookmarking.
The Academic Integrity Office's tips for students. The top 10 tips page says not to use ChatGPT or other generative AI on assessments unless the instructor has explicitly allowed it, and then only in the way they allowed. It also warns against using chatbots as a research tool, because "they will give you false sources." On summarizing readings instead of reading them, the page's verdict is two words: "Bad idea." If you're unsure whether something counts as cheating, the office takes questions at aio@ucsd.edu.
The office's AI in education page. The generative AI page frames the goal as AI literacy, and is careful to say this "is not about promoting AI use." It asks students to judge output critically, notice when AI isn't helping them learn, and know that AI detectors may not be valid or accurate.
The TritonGPT terms. UC San Diego runs its own assistant, TritonGPT, and the terms of use open it to students as well as staff. The terms say you may not submit its output as your own work for academic credit unless your instructor authorizes it, and that if they do, you should keep your chat history in case it's requested.
What that means in practice
- Silence is not permission. If a syllabus says nothing about AI, treat graded work as off limits and ask.
- Studying is a different question from submitting. The office says using AI to study may be fine, but warns that chatbots make things up, so you can end up learning something wrong.
- Keep your chats. TritonGPT's terms ask for this, and it's good practice with any tool. A saved conversation shows how you worked.
- Don't trust a citation you haven't opened. UCSD calls this out directly. Every source goes through the library or Google Scholar before it goes in your paper.
Which models suit UCSD coursework
UC San Diego is known for biology and bioengineering, cognitive science, data science, engineering at the Jacobs School and ocean and climate science at Scripps. Different models are better at different parts of that work.
Life sciences and pre-med: explain, then quiz
Use AI to understand a mechanism, then test yourself. Claude is good at patient, step-by-step explanations of long material.
Try: "Explain how the lac operon is regulated when glucose is low and lactose is present. Then give me five short-answer questions on it, without the answers, and grade my replies one at a time."
That's study use. It doesn't touch a graded assignment, and you do the thinking.
Data science: debug what you wrote
For a data science or statistics class, paste your own code and ask why it fails. Don't ask for a finished solution unless the course allows it.
Try: "Here is my pandas code and the error it throws. Don't rewrite it. Point to the line that causes the error, explain why, and tell me what concept I should review."
ChatGPT and Claude are both strong at code. If you want to see how they differ, our ChatGPT vs Claude for students post goes through it.
Scripps and climate science: read the paper first
Gemini handles very long inputs well, which helps when a reading is a 40-page report. Read it yourself, then use AI to check your understanding.
Try: "I've read this IPCC chapter. Here's my one-paragraph summary. Tell me what I got wrong or left out, and quote the section that shows it."
For finding sources, use a search model that links to real pages, such as Perplexity, then open every link. The model is a starting point, not a citation.
Cognitive science and writing-heavy classes
Use AI as a reader, not a writer. Paste your draft and ask where the argument is weakest, where a claim needs evidence, or where a paragraph loses the thread. Then fix it yourself.
How to use AI at UCSD without breaking the rules
- Read the AI section of every syllabus in week one. If there isn't one, email the instructor before using AI on anything graded.
- Use AI for studying and feedback before production. Explanations, practice questions and critiques of your own drafts are much safer than generated text.
- Disclose when it's allowed. A short note saying which tool you used and what for takes a minute and answers the question before anyone asks it.
- Verify everything. Facts, numbers and especially sources. The Academic Integrity Office's own warning about false sources is the reason.
- Keep sensitive data out of public tools. Don't paste other students' work, health information or unpublished lab data into a consumer chatbot.
- When unsure, ask the office. aio@ucsd.edu exists so you don't have to guess.
Where LazySusan fits
TritonGPT is the first stop for campus work where your instructor allows it. LazySusan is for everything around that: studying, personal projects, side work and comparing how different models handle the same question.
It puts ChatGPT, Claude, Gemini, Grok, Perplexity, DeepSeek and image and video models in one workspace with one bill. The part students use most is Multi-Chat: send one prompt to several models and read the answers side by side. When two models agree and one doesn't, that's usually where you need to check a source yourself.
The Student plan is $99 a year with a .edu email, with 5,000 tokens a month that roll over. If you're new to prompting, Prompt School covers the basics. You can sign up here.
LazySusan isn't affiliated with or endorsed by UC San Diego. Your instructor's syllabus and the Academic Integrity Office always come first.
The short version
At UC San Diego, the default for graded work is no AI unless your instructor says otherwise. Studying with AI is a gray area the school handles with caution, not a ban. Read the syllabus, keep your chats, check every source and ask the Academic Integrity Office when you're not sure.