Open-Ended Response Coding with ChatGPT

Before going to ChatGPT, get your data in order.

1. Open your Qualtrics dataset or SPSS output
2. Export the open-ended column to Excel
3. Clean obvious garbage: remove blank rows, responses that are just punctuation, and responses shorter than 3 words
4. Number each row (you'll use these reference numbers in the coded output)

What you should see: A numbered list of clean verbatim responses in a spreadsheet. Troubleshooting: If you have over 500 responses, split them into batches of 75-100 rows. This prevents the AI from truncating output.

Start a fresh ChatGPT conversation. Copy your first 50-75 responses and paste them with this prompt.

1. Go to chatgpt.com and click New chat
2. Paste this prompt, replacing the brackets:

Here are open-ended responses to the survey question: "[paste your exact question text]"

[paste 50-75 verbatims here]

Please:
1. Identify 6-8 recurring themes across all responses
2. For each theme, give it a short label (2-4 words) and a 1-sentence definition
3. Note which 3 responses best illustrate each theme

Output only the theme list for now. We will code responses in the next step.

1. Review the theme list carefully. Edit theme names or definitions that don't match your data

What you should see: A numbered theme list with definitions and example quotes. Troubleshooting: If themes are too broad (e.g., "Negative Experience"), ask: "Break theme 3 into more specific subcategories."

In the same conversation (so the AI remembers your theme list), send this:

1. Type:

Now using the theme list above, code each of the following responses.
For each response, assign it to 1-2 of the themes from our list.
Output as a table: Row # | Response (first 8 words) | Theme 1 | Theme 2

Responses to code:
[paste rows 1-75 with their row numbers]

1. Wait for the full table output. It may take 30-60 seconds for large batches
2. Copy the coded table into a new Excel tab

What you should see: A table with each response coded to 1-2 theme labels from your approved list.

Continue in the same ChatGPT conversation:

1. Type: "Continue coding with the same theme list. Here are the next batch:" and paste the next 75 rows
2. Repeat until all responses are coded
3. Keep all batch outputs in the same Excel tab

Troubleshooting: If the AI starts inventing new theme names, say: "Stay with the original 8 themes we defined. Do not create new categories."

After all responses are coded:

1. In ChatGPT, type: "Based on all the coding we've done, give me a frequency table showing: Theme | Count | % of Total | Top 3 illustrative quotes"
2. Alternatively, use Excel's COUNTIF function on your coded column to generate the count yourself. This is more reliable for large datasets

What you should see: A summary table ready to paste into your report.