Qualitative coding software that keeps the quote next to the code

Qualitative coding software should make one thing easy: going from a theme back to the exact words a participant said. The Qualitative analysis workbook in AI for Research is built around that. You upload interview or focus group transcripts, select a passage and code it against your codebook. Every coded excerpt keeps the exact quote and where it came from, negative cases are flagged instead of dropped, and each theme lists the excerpts that support or contradict it. It is for master's and PhD students doing thematic analysis, academics running interview studies, and researchers in health, education, business and policy.

What it does

On top of those, the workbook now has:

• Transcript upload. Add de-identified transcripts as sources for the workbook.
• Select text to code. Highlight a passage in a transcript and attach one or more codes; the exact quote and its locator are saved with the excerpt.
• AI code suggestions. Ask for suggested codes on a transcript. Each suggestion quotes the text it is based on, so you can check it against the passage before you accept it.
• Code frequency by source. See how often each code appears in each transcript, which shows when a code rests on one talkative participant.

When the analysis is ready, the workbook's output is a Draft findings report: the themes with their central meaning and the original excerpts as evidence, ready to edit in the editor. From there you can send the write-up to the Reviewer and check it against COREQ 2007 or SRQR 2014 with the reporting guideline checker.

Why negative cases have their own flag

In thematic analysis, a theme is stronger when you can show where it does not hold. Excerpts that contradict a theme are easy to lose: they get left uncoded, or coded and then forgotten when the themes are written up. The workbook makes the evidence direction a required field on every coded excerpt, and each theme has a place for contradictory excerpt IDs next to the supporting ones. When you write the findings, the exceptions are already listed, and a reader can see that you looked for them. Examiners and reviewers often ask how negative cases were handled; this gives you the record to answer.

How it works, step by step

1. Open Workbooks and start a Qualitative analysis workbook. (You can reach every workspace from the tools page.)
2. Upload your transcripts. De-identify them first and use speaker labels such as P01 and P02.
3. Build the codebook. For each code write a definition and, ideally, when to include and exclude it. A deductive codebook can be written before you read; an inductive one grows as you code.
4. Read a transcript, select a passage and code it. Set the evidence direction: supports, negative case or mixed, and add your interpretation.
5. Optionally press for AI code suggestions (3 credits). Review each suggested code against its quoted passage and accept, edit or reject it.
6. Check code frequency by source to see how codes spread across participants.
7. Develop themes: give each a central meaning, link its codes, and list the excerpts that support and contradict it. Move it from candidate to reviewed or rejected as you test it.
8. Write memos as you go, then create the Draft findings report and open it in the editor.

A worked example

Example (invented for illustration): a PhD student in education has twelve de-identified interviews with early-career teachers about leaving the profession.

1. She starts a Qualitative analysis workbook and uploads the twelve transcripts. Cost: 0 credits.
2. She writes a starting codebook from her literature review: workload, autonomy, mentoring, pay. She codes three transcripts by hand and adds an inductive code, "being watched", with include and exclude rules.
3. She asks for AI code suggestions on the remaining nine transcripts, three runs at 3 credits each. She accepts about half the suggestions, edits some and rejects the rest where the quoted passage does not fit the code's definition.
4. Code frequency by source shows "being watched" comes almost entirely from two participants. She records that in a memo and keeps the theme as a candidate.
5. A theme on mentoring has two negative-case excerpts from teachers who stayed despite poor mentoring. She keeps them linked as contradictory evidence and reports them.
6. She creates the Draft findings report, edits it in the editor and checks the write-up against COREQ in the Reviewer: 5 credits for a chapter under 10,000 words.

Total: 14 credits, inside the Free plan's 45 a month.

What it does not do

• It does not do the analysis for you. AI suggests codes; the researcher decides, and themes are yours.
• It does not transcribe audio. Upload text transcripts; see the post on interview transcription for options.
• It does not de-identify transcripts. Remove identifying details before upload, as your ethics approval requires.
• It does not calculate inter-coder agreement. If you double-code, compare codes outside the workbook; the post on Cohen's kappa explains how.
• It is not a full replacement for NVivo or other dedicated qualitative data analysis software. It is a focused workbook for coding, themes and a findings draft in the same place as your literature and writing.

Who uses it

• Master's and PhD students doing thematic analysis for a dissertation: see AI for master's students and AI for PhD students.
• Academics running interview and focus group studies: see AI for academics and social sciences research.
• Health researchers coding patient or staff interviews: see AI for medicine.
• User researchers and evaluators in organizations: see AI for industry.

Learn more

• How to code qualitative data
• Inductive vs deductive coding
• AI for qualitative research
• Interview transcription for researchers

Last updated 2026-11-13