If I Were Writing My Thesis in Kenya Today, Here's How I'd Use AI

Short answer: I would use AI as a tutor, a search assistant and a tough first reader, never as the author. At every stage, the ideas, the data and the final words would stay mine, and I would keep a simple log of how I used AI. Before any of that, I would read my university's rules and ask my supervisor what they allow.

I'm Rodney Maina, founder of FuKazee. I have trained more than 3,000 people since 2022, including Masters and PhD students in the ScholarShift Accelerator. This is the plan I would follow if the thesis were mine, stage by stage, from choosing a topic to standing in front of the panel.

It is written for Masters and PhD students at Kenyan universities who already use ChatGPT or Gemini a little and want to use AI properly without risking their degree.

Step zero: read the rules before you open any AI tool

As of November 2026, Kenya has no single national rule on AI in postgraduate work. In August 2026, the Commission for University Education (CUE) held a three-day national dialogue with representatives from about 54 universities to build a framework for AI in higher education, according to a report by Africa International University. Until that framework exists, your own university's rules apply.

Here is what I would do in the first week:

1. Find your university's AI guidance and plagiarism policy. Strathmore University's student guidelines, for example, say not to use generative AI where it is prohibited, to acknowledge it where it is allowed, and to save your prompts and the AI's answers in case a lecturer asks to see them.
2. Know your similarity limit. Kenyatta University's library says its revised anti-plagiarism policy allows an overall similarity index of no more than 16%, and no more than 1% from any single source. Check the figure for your own university.
3. Ask your supervisor one direct question: "Which of these uses of AI are you comfortable with in my thesis?" Bring the table below and agree it in writing, even by email.
4. Start an AI log. A simple document with the date, the tool, what you asked and what you used. It takes two minutes a day and protects you later.

The plan, stage by stage

1. Topic: use AI to argue with you

I would use AI to stress-test my idea, not to pick it. I teach the four Cs before anything else: Clarity, Context, Constraints and Customisation. Here is a prompt built on them:

I am a Masters student at a Kenyan university in [field]. My draft research question is: [paste]. Context: I have [number] months, a budget of about [amount], and access to [population or data]. Act as a sceptical supervisor. Tell me whether the question is too broad, what data I would need, and three ways it could fail. Do not suggest references.

That last line matters. You will find sources yourself in the next stage.

2. Literature: no source enters the thesis unless you opened it

AI reads fast, but it is bad at finding real sources. In a 2023 study in Scientific Reports, Walters and Wilder found that 55% of citations produced by GPT-3.5, and 18% by GPT-4, were fabricated. Tools have improved since, but the rule stays: if you have not opened the paper, it does not go in.

I would use search tools that return real papers, then a reference manager such as Zotero, then AI to help me compare papers I have already read. Our guides on AI tools for literature review and citations without fake sources go deeper.

3. Method: a sparring partner, not a decision-maker

AI can explain the difference between a case study and a survey design in plain words, or check that each objective has a matching question and method. It can draft a first interview guide for you to rewrite. It should not decide your design, and it must never see identifiable participant data. Kenya's Data Protection Act, 2019 covers personal data, and your ethics approval will set its own conditions.

4. Analysis: you run it, you check it, you interpret it

For numbers, I would ask AI to explain a test, then run it myself in Excel, SPSS, R or Python and check the output by hand on a small sample. For interviews, AI can suggest possible codes on anonymised transcripts, but I would read every transcript myself and keep the final coding mine. See AI for data analysis in postgraduate research and qualitative data analysis with AI.

5. Writing: feedback on my draft, never a draft for me

I would write every chapter myself, then ask AI where my argument is unclear, where a paragraph repeats itself, or what a supervisor's comment means. Strathmore's guidelines warn that even editing with AI can raise detection flags on tools like Turnitin, so keep your drafts and your log.

Journals set a useful standard. As of November 2026, Elsevier says AI tools should not be listed or cited as authors. It asks authors to declare the AI tools they used to prepare a manuscript, including any tool that made substantive changes to sentence structure. Basic grammar, spelling and punctuation checks need no declaration under that policy. Your university may set different rules. For the bigger picture, read how to write a Master's thesis and using AI in postgraduate research without crossing the integrity line.

6. Defence: rehearse with AI, answer alone

This is where AI is most useful and least risky. Paste your abstract and ask for the ten hardest questions a panel could ask, then answer them aloud without notes. Ask it to find the weakest link between your findings and your conclusions. Our guides on final thesis defence tips and how to prepare for defence cover the rest.

The honest limits

• AI detectors are unreliable, especially for non-native English writers. In a 2023 Stanford study of seven detectors, they classified 61.22% of TOEFL essays written by non-native English students as AI-generated. Your best protection is your drafts, your log and your ability to explain your work.
• AI does not know your field's latest Kenyan data. It may sound confident and be out of date.
• Nobody becomes an expert in a few days. A good workflow takes weeks of practice on your own thesis.
• Your supervisor's view wins. If they say no to a use, respect it.

Why I believe the ScholarShift Accelerator is the best place to learn this

We believe the ScholarShift Accelerator is the best choice for a Kenyan postgraduate who wants to use AI properly. That is our own view. Here is the evidence from our live site:

• Integrity comes first. Module 1 is AI Ethics in Academia: how to cite AI, disclose it and stay inside your university's rules. Then come prompting, literature reviews, Zotero, academic writing and defence, and building your own research assistant.
• You work on your own thesis. It is a four-week programme, live on Google Meet, Monday to Thursday, 8:00 PM to 9:30 PM, with every session recorded and available for one year, plus a weekend session in person or online.
• Researchers from many universities. Our cohorts have drawn researchers from the University of Nairobi, JKUAT, Kenyatta University, Daystar, KeMU, Kwame Nkrumah University and the University of Glasgow.
• I teach it. You get direct access to me throughout the four weeks. My background is on my trainer page.
• Clear price. As of November 2026, the fee is KSh 20,000, or a KSh 10,000 deposit and the balance midway. The next cohort is planned for late January or February 2027; ask us on WhatsApp for the confirmed date.
• Learners rate us. Our rating is 4.9 from 38 Google reviews, as shown on our reviews page on 14 November 2026.

When FuKazee may not be the right fit

• You want someone to write or "edit heavily" your thesis. We do not do that.
• Your university or supervisor has banned AI for your work.
• You need a statistician to run your analysis.
• You already have a working AI research workflow and only need one tool explained.

Questions people ask

Is it allowed to use AI for a thesis in Kenya?

It depends on your university and your supervisor. As of November 2026 there is no single national rule, and CUE is still building a framework. Read your university's policy and get your supervisor's agreement in writing.

Do I have to declare that I used AI?

Where your university allows AI, it usually expects you to acknowledge it. Strathmore's guidelines, for example, ask students to acknowledge AI use where permitted. When in doubt, declare it and keep your log.

Can AI write my literature review?

It should not. It can help you find real papers, summarise ones you have opened and map themes. The synthesis and the argument must be yours, and every citation must be checked.

Will Turnitin flag my work if I used AI?

It might, even for light editing. Detectors make mistakes, especially with non-native English writers. Keep your drafts, prompts and log so you can show how the work was made.

Can I paste my interview transcripts into ChatGPT?

Not if they contain names or anything that identifies a participant. Anonymise first, check your ethics approval, and use the most private settings your tool offers.

Which AI tools should I start with?

Start with the free version of one general assistant, a reference manager such as Zotero, and one paper-search tool. Learn them well before adding more.

Sources

All sources last checked 14 November 2026.

• Africa International University, national dialogue on an AI framework for university education (20 August 2026)
• Strathmore University Center for Teaching Excellence, student guidelines for the use of generative AI tools
• Kenyatta University Library, review of the anti-plagiarism policy
• Walters, W. H. and Wilder, E. I. (2023), "Fabrication and errors in the bibliographic citations generated by ChatGPT", Scientific Reports 13, 14045
• Elsevier, generative AI policies for journals
• Stanford HAI, "AI detectors biased against non-native English writers" (May 2023)
• Data Protection Act, No. 24 of 2019 (ODPC copy)
• FuKazee facts: ScholarShift Accelerator, About, Rodney Maina, Reviews

Where FuKazee fits in

The ScholarShift Accelerator teaches Masters and PhD students to use AI at every stage of the thesis, with integrity first and on their own topic. If you want to join the next cohort, read how to join the next ScholarShift Accelerator cohort or see AI for thesis writing in Kenya.

Tell us your university, your stage and where you are stuck. Email lead@fukazee.com