Start with your course rules and what you need to learn
Before opening an assistant, check the instructions for both your course and the specific assignment. Rules can differ between courses and between assignments in the same course: one activity may allow AI for brainstorming but restrict it for writing the final answer. If the instructions do not distinguish between these uses, ask your instructor before you begin. Clarifying the scope in advance can prevent an allowed use during one stage from causing a problem at another.
Next, define what you want to achieve without delegating the learning that the assignment is meant to assess. You might want another explanation of a concept, practice questions, or help identifying which parts of an argument need review. If the assignment assesses your ability to analyze a source, summarize it yourself, or write an essay, asking an assistant to produce the whole submission could replace the very work being evaluated. The tool does not decide what is allowed: the applicable assignment instructions and your institution’s academic rules do.
If you are unsure whether a use is permitted, describe what you have in mind: “I’d like to ask for an explanation of a concept to help me study, but I’ll write the answer myself”; or “I’d like help checking the clarity of a draft I wrote.” Also ask whether you must disclose your use, what details to include, and whether you need to keep prompts or conversations. The University of Arizona, for example, says it does not apply one AI rule to every course and that instructors determine when and how students may use AI. That institutional example does not establish the policy at Hillsborough College or any other institution: check the rule for your own course.
Bring a learning goal to the conversation, not just a request to produce something. “Do my assignment” does not set out what you can verify or what you need to do yourself. By contrast, “Explain the difference between correlation and causation, then ask me three questions to check my understanding” keeps the focus on learning. Still, check that the activity allows this kind of support and evaluate the explanation before using it as a study resource.
A three-step check before you start
Use this sequence before entering text, asking for an explanation, or connecting an account.
- 01Read the course and assignment instructions. Identify permitted uses, restricted uses, and any disclosure requirements.
- 02Set a learning goal and separate the work you will do yourself from the support you want to request.
- 03If anything remains unclear, ask your instructor before using the tool and keep the answer with your course instructions.
Distinguish the assistant, the model, search, and verification
An AI assistant is the interface or service you interact with. It may provide access to one or more language models, as well as features such as web search, document uploads, or connections to other applications. These components are not interchangeable. One platform may let you compare models, while another may organize responses around materials you provide. The University of Arizona describes an institutional environment where users can interact with and compare multiple models; that description applies to its service, not to every account or tool.
A generated response does not become a source just because it sounds confident. Some systems can search web pages or academic content, summarize documents, analyze data, generate images, or carry out tasks with connected applications. A search feature does not confirm the accuracy of every summary, and citations in a response do not prove that the system interpreted them correctly. Open the original document and check whether it supports the exact claim you want to use.
Do not confuse a model’s response with a literature search, either. A search engine helps locate materials; an assistant may suggest keywords, explain a passage, or summarize text. For academic work, identify the primary or secondary source that supports each claim and cite it according to your course requirements. The Modern Language Association recommends consulting directly the secondary sources a tool mentions and citing those sources instead of attributing the fact solely to the assistant’s response.
Comparing two responses may reveal disagreements and help you frame questions, but agreement between assistants is not independent verification. Two systems may repeat the same error, rely on similar materials, or fail to show where a claim came from. Treat comparison as a signal to investigate, not as final proof. When a response could affect an important interpretation, return to the course text, the original source, or a subject expert.
Choose a tool for the task, not by ranking
There is no universally best tool for studying. The choice depends on what your course permits, which features are actually available in your account, the privacy terms, and what you can verify. The University of Arizona’s tools guide lists Gemini, Gemini Notebook, Copilot, Claude, ChatGPT, and Perplexity in different contexts, but a directory is not a comparative test or a guarantee of availability, price, or protection for every user. Features and models can change.
You can consider ChatGPT, Microsoft Copilot, Gemini, Claude, and Perplexity as names to investigate, not as a ranked list of recommendations. The University of Arizona’s documentation describes Perplexity as a service that lets users choose among models, rather than as a foundation model itself. The university also distinguishes its institutional environment from general-purpose tools. Check the service and your institution to see which features are active for your account before building a workflow around them.
Meta AI may also come up in discussions of assistants, but the institutional information cited here does not establish its current features, terms, or availability. If you are considering it, check current official documentation and your institution’s rules before using it. Do not infer data protection, free access, academic suitability, or source-citation capabilities from a product name.
Evaluate each option with the same task and the same authorized material. Record the account type, the features available, any relevant limits, and the date you checked. A feature shown in an institutional account may not be available in a personal account, or vice versa. A test describes that specific setup; it does not show that the tool will respond the same way in other accounts, languages, or at other times.
A comparison matrix that does not assume features
Copy the matrix and fill it in with conditions you can check in your account. Make clear when a detail has not been confirmed.
| Criterion | What to record | Question to check |
|---|---|---|
| Permission | Course rule for this specific assignment | Does it allow the specific use I plan to make? |
| Account | Institutional or personal account and associated institution | Am I using the intended, authorized version? |
| Data | Material entered or connected and its sensitivity | Am I permitted to share it with this service? |
| Feature | Capability observed in my session | Is this feature available now in this account? |
| Limits | Restrictions shown by the service or set by the course | Are there usage, file, or access limits that affect the task? |
| Cost | Cost checked for my account, if any | Is this condition still current and does it apply to my plan? |
| Response | The specific claim I plan to use | Which original source supports it? |
| Decision | Use, change the plan, or do not use | Can I justify the decision using the rules and checks I recorded? |
| Date | Date you checked the terms and setup | Should I check them again before another assignment? |
Protect your data and review connection permissions
Do not enter personal, confidential, or protected information unless you have explicit authorization and understand the applicable terms. This includes information that identifies other people, records, someone else’s drafts, passwords, and internal materials. If an assignment requires you to work with a text, use a public version or your own material that you are allowed to share. If you cannot reduce the risk safely, do not upload the content; ask your instructor or library about an alternative.
Distinguishing training, retention, history, memory, and sharing helps you assess privacy. Turning off a training option does not, by itself, prove that a service deletes data immediately, keeps no records for other reasons, or prevents a shared conversation from being accessed. A temporary conversation may change what appears in your history, but that alone does not establish every storage practice. Read the current terms and the settings for the account you plan to use.
Protections for an institutional account do not automatically apply to a personal account or students at other institutions. The University of Arizona describes specific protections for certain tools when someone signs in with a university account; its guide also warns that its approved tools are not suitable for every type of sensitive information. Do not assume that example applies at Hillsborough College. Check your own institution’s policy and account type before entering course material.
Take extra care with agents or features that can act on files, websites, applications, and accounts. The University of Arizona’s guide notes that some tools can perform multi-step tasks and connect to external services; it also warns that access to private files or accounts can have consequences if the system makes a mistake. For a study activity, grant only the minimum necessary permissions, avoid access to important folders or accounts, and review each action before confirming changes or submissions.
Before uploading a file or connecting an application
Assess the content and scope of permission before you start an action.
- 01Identify whether the file contains personal, confidential, protected, or other people’s information.
- 02Check that the assignment instructions and your institution’s policy allow you to use that content with the selected tool and account.
- 03Review which files, applications, or permissions the assistant will receive; reduce access and avoid authorizing unnecessary actions.
- 04If you cannot confirm authorization or data handling, do not share the file. Ask your instructor or library about an alternative.
Try a low-risk task and verify the result
A small, focused test can help you decide whether a tool suits a particular task without mistaking an impression for a broad evaluation. Choose a permitted learning goal and public text or your own authorized material. Read the material first and write a short explanation in your own words. Then ask the assistant, for example, to explain a concept in simple terms, identify the main ideas, or write review questions. The activity is for comparing kinds of support, not for letting the assistant decide what is correct.
If you compare tools, use the same text and question, and record the account, date, and relevant settings. Do not change the material, goal, and instructions all at once, or you will not know what caused a difference. This kind of test does not establish which tool is best overall; it describes only how that setup responded to that task. One configuration or response does not guarantee accuracy in later uses.
Hypothetical example: in a history class, you read a public excerpt from a document and write down three main ideas. Then you ask for an explanation of one term in the excerpt and a review question. Check the explanation against the text and course materials; check that the question tests the concept actually discussed. If the assistant adds dates, causes, or interpretations that the excerpt does not contain, do not accept them as facts: find support in an appropriate source or remove those claims.
When checking a response, separate three questions: Does it represent the material faithfully? Does an external claim have support? Does the final work comply with the assignment rules? Open proposed references and review the relevant passage. If you cannot locate a source or it does not support the claim, do not cite it as if it did. If the assistant paraphrases a document, compare the meaning, nuance, and conditions with the original. A short answer can omit an important exception even when the rest seems correct.
Treat errors as information about your process, not as a reason to trust a different answer without checking it. You can clarify an ambiguous question, provide permitted context, or ask the assistant to distinguish what the text says from an inference. Those changes may make a response more useful, but they do not turn the tool into a deterministic checking mechanism. The person submitting the work remains responsible for the final review.
Brief test log
One row describes one session; do not turn a single observation into a universal conclusion.
| Date | Permitted task | Account and tool | Authorized material | Claim to check | Original source reviewed | Result and decision |
|---|---|---|---|---|---|---|
| Fill in when testing | Fill in the course instruction | Fill in the setup used | Fill in the authorized text | Copy the specific claim, not the whole response | Record the document and passage reviewed | Use, correct, check further, or discard |
Disclose and cite your use transparently
Follow your instructor’s instructions and the citation style required for the course. Some rules may ask for a brief disclosure; others may require a citation for generated content, a list of prompts, or an appendix. Do not present generated text that you did not write as your own, and do not attribute an external source to the assistant. Your disclosure should describe what you actually did, not a generic list of features you did not use.
The Modern Language Association’s revised guidance does not recommend treating the AI tool as an author. For a Works Cited entry, describe the generated content, name the tool, identify its model or version as specifically as possible, identify the company responsible, and give the date the content was generated. For the location, MLA prefers a stable, shareable URL when one is available; if the tool does not provide that feature, it allows the general URL for the tool. That preference does not authorize publishing a conversation containing sensitive information: protect privacy and ask your instructor what alternative documentation is acceptable. Follow your course instructions and the current version of the required citation style.
The MLA recommends directly consulting and citing secondary sources linked by a tool instead of citing only the assistant’s response as support for those sources. It also warns that tools can invent references or inaccurately summarize what a source says. Open each original, check the claim, and cite the document you actually consulted. An AI response may need acknowledgment for its contribution, but it does not replace citing the material that supports the fact.
If a shareable link would expose a conversation you do not want to make public, do not share it without authorization. Ask your instructor how to document your use in a way that protects privacy. Keep a private record with the date, the tool and model name if you can identify it, the task, relevant instructions, the parts you checked, and the original sources you cited. You do not need to keep or disclose sensitive information to demonstrate transparency.
What to do when the tool is not enough
If you cannot confirm that a use is permitted, do not have an appropriate account, or cannot check a response, stop before incorporating it into an assignment. The library, your instructor, and course materials may offer guidance or an alternative. Not having access to an assistant does not prevent you from studying: you can work with readings, notes, practice questions, and support available through your institution.
Do not treat an AI-text detector as a verdict on who wrote something. The University of Arizona’s academic integrity guide says these tools can produce false positives and false negatives and are not reliable enough to prove that a student used AI. That caution does not, by itself, determine the procedures at another institution. Check your own institution’s rules and support options if an accusation or academic-integrity concern arises.
Keep drafts, notes, outlines, and versions of your work if you want to document your process. These materials may help explain how you developed an idea, but they do not by themselves prove that every choice complied with the assignment rules. If you are concerned about a review, explain specifically what you did, what assistance you used, and which sources you consulted, then ask for guidance through your institution’s channels.
Before submitting, return to the assignment instructions and check your content, citations, and disclosure. Review the service terms if you plan to use the tool again: menus, setting names, models, and features can change. An earlier check describes only the account, date, and setup you reviewed; it does not confirm that the terms remain the same.
Final check before you submit
Assess permission, learning, and accuracy separately; do not let a fluent response replace these checks.
- 01Confirm that your specific use follows the course and assignment instructions.
- 02Check every claim you plan to keep against the original materials; correct or remove anything you cannot support.
- 03Make sure the final writing reflects your understanding and does not present someone else’s or generated content as your own without acknowledgment when required.
- 04Add the disclosure or citation your course requires, and do not share conversations or sensitive data without authorization.
- 05If a question about privacy, permission, or accuracy remains unclear, ask your instructor or library before submitting.
Keep exploring
Sources consulted
Corrections and transparency
If you spot incorrect or outdated information, send us a correction with the page and source we should review.
Submit a correction