AI may change research as much as the internet did. Before online research became routine, finding sources could mean a trip to the library and a search through card catalogs. The internet made that work far more efficient. AI offers another leap, but it also raises a question for doctoral students: How do you use a powerful research tool without letting it do the thinking you came here to learn?
At City Vision, our values are Jesus, justice, and technology. We want to make cutting-edge use of tools that help us address difficult problems in our organizations and communities. That means integrating AI into the doctoral program while being clear about where assistance ends and your own intellectual work begins.
The tools will change, and screens or subscription offers may look different by the time you set them up. Use current product instructions when that happens. The principles below matter more than any particular button.
Key Takeaways
- Use Gemini and NotebookLM to support research while continuing to engage deeply with course material.
- Treat AI research reports as secondary sources, and review their claims and citations.
- Write your own assignments rather than submitting AI-created or substantially AI-rewritten prose as your work.
- Build effective prompts by defining the task, context, expertise, format, and constraints.
How AI fits into the doctoral program
Doctoral students are required to maintain a Google Gemini Pro subscription throughout the program. We integrate Gemini and NotebookLM into nearly all courses. You will also keep a publicly facing research journal on Substack, using articles and podcasts to share your work.
After you complete a course, we provide access to a NotebookLM containing that course’s content. The timing is intentional. If we handed you a ready-made way to query everything on day one, it would be too easy to bypass sustained engagement with the material. Once you have done that work, the notebook becomes a resource you can return to.
We use AI when preparing courses, too. It helps us create highly customized, integrated material for topics where an existing textbook does not fit. Consider research methods in this program: we emphasize practical application and action research while integrating a Christian worldview. Rather than asking students to piece those elements together from an assortment of books, we can develop course-specific articles that bring them into conversation. Teaching faculty also have access to course content in NotebookLM.
That is the overall approach: use technology to make research and teaching more useful, while preserving the work that actually develops a researcher.
Choose your Google account and organize your research
One of your first decisions is whether to subscribe to Gemini through a personal Google account or a work account. Neither choice is automatically right for everyone.
- A personal account may qualify for Google student offers when they are available. It can also be useful if you want to create public NotebookLM content. Check the current offer terms rather than assuming a discount is available.
- A work account may be more convenient if your employer uses Google Workspace and you need to work with documents already in that environment. Your employer may also be willing to cover the subscription.
If you pursue a student offer, use your enrollment verification from the student portal and follow the current instructions for the offer. The course discussion assignment provides the program’s setup steps.
Next, set up Google Drive for research with separate public and private folders. The public folder gives you a place for resources you may want to share alongside an article, podcast, or paper. The private folder keeps material that is not meant for public distribution separate. Deciding which is which at the beginning is much easier than untangling everything later.
Use Deep Research to find a direction, not to skip source evaluation
Google Deep Research can help you survey a topic quickly, but a broad request will not necessarily produce the academic sources you need. You have to tell it what kind of research to seek. Our course resources include suggested prompts for academic research, citation guidance, and instructions for moving Gemini Deep Research citations into Zotero.
For practice-oriented projects, I often recommend a sequence that moves from broad discovery toward a more focused set of peers:
- Scope the organizational landscape. Identify the types of organizations working on the problem and the approaches they describe publicly.
- Narrow to promising peer organizations. Look for organizations whose practices or experience may be relevant to your question.
- Investigate those peers more closely. Ask targeted questions about their published work, then examine the sources behind the AI report.
If you work with a rescue mission, the program has a directory of rescue missions that can help you build that peer group. If you work in another movement or field, create a comparable list. Larger organizations with substantial information online are generally more useful for this kind of AI-assisted search than small organizations with little published material. That does not make the smaller organizations unimportant. It means there is less available for an AI tool to investigate.
Remember what a Deep Research report is: a secondary source. It may point you toward primary sources, but the report itself is not a substitute for reading and evaluating them. If you use an AI-written report as a source, cite it directly. If you publish one, review it first for factual problems and claims inconsistent with your values. Fast research still needs human judgment.
Learn the tools by solving the how-to problems
Parts of this program are deliberately practical. You may need to set up Drive, work with NotebookLM, or figure out a Deep Research workflow. We provide starting instructions and collections of how-to resources, but doctoral-level research includes learning how to resolve a technical problem when a tool changes.
You do not need to work through every optional how-to resource if you already know how to complete the task. A specific resource assigned for a substantive part of a course is different, and you should treat it accordingly. When you are stuck on setup, ask Gemini for more detailed instructions. You can describe the technical task and the point where you are having trouble.
There is an important boundary here. Asking AI how to configure a tool helps you complete the setup. Asking it to complete an intellectual assignment for you removes the work the assignment was designed to teach. Use the same research independence for both situations, but do not confuse them.
Think of AI as an electric bike, not a replacement for training
I like to compare the doctoral journey to preparing to bike up Mount Kilimanjaro. Your doctoral project is the climb. Each course helps build the research and thinking fitness you will need when the climb gets difficult.
AI is like an electric bike. It can help you move faster, and there are good reasons to use that assistance when you are tackling major challenges in society or in your organization. But if the electric bike carries you through every training session, you will not build the fitness required for the final climb.
The same is true of research and writing. AI makes a steep task feel easier. That is valuable when it helps you locate material, explore connections, or solve a technical obstacle. It becomes a problem when it takes over the work through which you develop your own analysis. I do not want students to reach the doctoral project without the intellectual fitness to complete it.
Keep a boundary between AI research and your writing
In technology, an air gap separates systems so that information cannot pass directly between them. As a working principle, I want something close to an air gap between an AI research report and the prose you submit as your own. Read the research, assess its sources, develop your argument, and write it yourself. The act of writing is part of how you learn to think.
You can write your own article and cite an AI research report where appropriate. You can also publish an AI-generated report as such, after reviewing and correcting it. What you should not do is take AI-generated prose, or prose substantially rewritten by AI, and present it as your unaided assignment or project.
That includes asking a tool to “clean up” a paper if the result is that it performs much of your writing. City Vision encourages appropriate AI use precisely because technology is one of our core values. We are also strict about misuse. Do not submit text created or manipulated by AI without directly citing it. The goal is not to make you afraid of the tools. It is to keep authorship clear and the educational purpose intact.
Write prompts like a researcher, not search queries
As AI becomes a regular research assistant, the quality of your questions matters enormously. A short search-style question may give you a quick answer. A useful research prompt requires you to define the project.
A practical structure is persona, task, context, format, and constraints:
- Persona: What domain of expertise should the response draw on?
- Task: What research or analysis do you want completed?
- Context: What problem, organization, audience, or history makes the task specific?
- Format: What form should the result take, such as a report?
- Constraints: What source types, citation style, or boundaries should shape the work?
For example, a prompt about gospel rescue missions can specify an audience of executive directors, ask for a consulting report, and require the research to consider both theology and sociology. Depending on the purpose, you might ask for APA-style citations or limit the search to peer-reviewed sources. Those details change what the assistant has a chance to produce.
Study the prompts attached to AI-generated articles in your courses. They are not merely instructions hidden behind a finished result. They model how to frame a research problem. Notice what terminology they use, how they identify relevant sources, and how they connect the requested output to a particular audience.
Connect knowledge domains to get richer research
Strong prompts depend on more than knowing the right words. They depend on understanding how ideas relate across fields. As you move through your courses, you will gain terminology that lets you ask more precise questions. You will also learn when a concept in one discipline helps illuminate a concept in another.
One example is the relationship between theological vision and the sociological idea of reframing. In work on gospel rescue missions, I use both sets of concepts to investigate how a movement might describe its mission in response to changing circumstances. The prompt asks about diagnostic, prognostic, and motivational framing while also asking what a faithful Christian theological vision would mean for the movement.
That is not a simple request for a history of rescue missions. It is a research assignment with a framework. It asks the AI to examine examples from a substantial list of organizations and develop five possible reframing positions. Those possibilities must aim to help the movement achieve critical mass while remaining faithful to its historical values. Related concepts, such as ministry expression and the sociological idea of repertoires, add another way to connect vision with practice.
The AI can work through a large body of publicly available material quickly. Your contribution is knowing which question is worth asking, which domains need to be connected, which organizations belong in the comparison, and whether the resulting analysis holds up. Treat the tool as a research assistant available whenever you need it, not as the researcher in charge.
Be the engineer who knows where to use the tool
There is an old story about an engineer called in to fix a large machine. The repair takes only moments, but the bill reflects the expertise required to know exactly where to act. The point applies to AI prompts: entering instructions is easy; understanding the problem well enough to give the right instructions is the skilled part.
That is also how I use AI when developing course materials. There may be more useful lectures I would like to create than I have time to record. I can articulate the purpose of a proposed lecture, supply the conceptual connections and other hooks it needs, and use a carefully framed prompt to generate supplemental material. I still examine the result, correct problems, and add diagrams where they make the ideas clearer.
The aim is not to press an easy button until research disappears. It is to become wise enough to know when AI can extend your work, when it needs close scrutiny, and when you need to put it aside and do the writing yourself. Set up the tools, learn to direct them well, and keep building the intellectual fitness your doctoral project will demand.