AI in Africa
How can you use ChatGPT to analyse survey results for a…
GOSPELTRADER Research Desk · 6 November 2026 · 5 min read
Learn how to clean survey data, calculate percentages and interpret results with ChatGPT for a Nigerian research project, while protecting respondents.
You can use ChatGPT to help clean anonymised survey data, produce summary tables and explain patterns in your Nigerian research project. Check every calculation against your spreadsheet or statistical software, and follow your department’s rules on AI use before including the results in your work.
1. Prepare your survey spreadsheet and remove personal details
Export your responses from Google Forms or your survey platform to a spreadsheet. Keep an untouched original, then create a separate copy for cleaning and analysis.
Use one row per respondent and one column per question. Before sharing anything with ChatGPT, remove names, phone numbers, email addresses, matriculation numbers and other identifying details. Replacing names with codes is not enough if other details can still identify someone.
Check your consent statement and your institution’s data-handling requirements. If you are not allowed to share respondent-level data with an external AI service, use approved tools locally and share only permitted, non-identifying summaries.
- • Use clear column names, such as age_group, study_level and internet_access.
- • Standardise equivalent responses, such as Yes, YES and yes, without changing their meaning.
- • Distinguish unanswered questions from responses such as No or Not applicable.
- • Review apparent duplicates before removing them; similar answers do not necessarily mean duplicate respondents.
2. Give ChatGPT your research question and coding guide
Explain what your study investigates, how you selected respondents and what each column means. Without this context, ChatGPT may recommend unsuitable comparisons or misinterpret numerical codes.
If file uploads and data analysis are available in your account, upload the cleaned, permitted dataset. Otherwise, paste a small anonymised summary table and ask for guidance or spreadsheet formulas rather than expecting reliable calculations from a large pasted dataset.
Try this prompt: “You are helping me analyse a student survey about internet access at a Nigerian university. Each row represents one respondent. My research question is: [insert question]. Here is my coding guide: [insert guide]. First identify unclear codes, missing values and possible data-quality problems. Propose an analysis plan; do not invent responses or change the dataset.”
- • Explain numerical codes, such as 1 = Yes and 2 = No.
- • Identify questions where respondents could select more than one answer.
- • State the labels and direction of any rating scale.
- • Provide your actual sampling method, rather than asking ChatGPT to infer it.
3. Start with counts, percentages and clear tables
Begin with descriptive results: how many respondents selected each answer, what percentage they represent and how many left the question unanswered. Ask ChatGPT to state the denominator used for every percentage.
For example, if 60 out of 80 people who answered a question selected Yes, the percentage is 75%. If your total sample was 100, you must explain whether you are reporting a percentage of valid answers or of the full sample.
For rating-scale questions, report the response distribution first. If your department expects a mean score or a combined scale score, confirm the coding and scoring method before calculating it.
- • Prompt: “Create a frequency table for each question. Show counts, percentages, valid responses and missing responses. Explain how you calculated each percentage.”
- • For multiple-response questions, state whether percentages use respondents or total selections as the denominator.
- • Check that single-response category counts add up to the number of valid responses.
- • Do not treat missing responses as zero unless your documented method specifically requires it.
4. Verify statistical tests and open-ended themes
Use ChatGPT to discuss which statistical test might fit your research question, not to choose whichever test produces a significant result. The appropriate method depends on your variables, sample design and assumptions. Confirm it with your supervisor or a statistics adviser.
Check numerical outputs in Excel, SPSS, R or another suitable tool. Ask for the formulas or code used, and verify that any filtering or exclusions match your analysis plan.
For open-ended answers, ChatGPT can suggest themes and a coding framework. Review the original responses yourself, allow a response to receive more than one theme where appropriate, and never present an AI-generated sentence as a respondent’s quotation.
- • Prompt: “Suggest an appropriate analysis for [research question]. Explain the variable types, assumptions and limitations. Do not calculate a p-value without the necessary data.”
- • Prompt: “Suggest themes for these anonymised responses. Define each theme, flag ambiguous answers and do not invent quotations.”
- • Keep a record of cleaning decisions, coding rules and verified calculations so you can explain your method.
5. Write your findings without overstating what they show
Give ChatGPT your verified tables and ask for a plain-English draft of the findings. Require it to use only the supplied results, retain the correct figures and separate observations from possible explanations.
Try this prompt: “Draft a results paragraph using only this verified table. State the sample size and missing responses where relevant. Do not invent citations, claim causation or generalise beyond the sampled population. Flag anything the table cannot establish.”
A survey from one university does not automatically represent all Nigerian students or students across Africa. Likewise, an association between internet access and academic performance does not by itself prove that one caused the other.
Check every sentence against your tables, revise the wording yourself and disclose AI assistance if your institution requires it. You remain responsible for the analysis and the submitted project.
Frequently asked questions
Can you analyse survey results with the free version of ChatGPT? You can ask for analysis plans, spreadsheet formulas and explanations of summary tables. File uploads and data-analysis tools may be available with usage limits, so check your account’s current features. If those tools are unavailable, calculate results in your spreadsheet and ask ChatGPT to help interpret a non-identifying summary.
Can ChatGPT replace SPSS for your final-year project? ChatGPT can assist with analysis, but it should not replace a verifiable statistical workflow. If your department requires SPSS, use it and retain your output files. Whatever tool you use, you should be able to reproduce the calculations and explain the method.
Can you ask ChatGPT to fill in missing questionnaire responses? Do not ask it to invent answers and present them as collected data. Record missing responses and follow a documented method for handling them. If statistical imputation is appropriate, agree the approach with your supervisor, preserve the original data and report the method transparently.