Thematic analysis is the method most qualitative researchers reach for first — and the one most often done loosely. Done well, it is a rigorous, transparent way to find and report patterns of meaning across interviews, focus groups, or open text. This guide walks through it phase by phase, using Braun and Clarke's widely cited framework, and shows where AI now speeds the work without replacing your judgment.
For the wider analysis context, see our Qualitative Data Analysis guide; for fundamentals, Qualitative Research 101.
What is thematic analysis?
Thematic analysis is a method for systematically identifying, analyzing, and reporting patterns of meaning — themes — across a qualitative dataset. It is the most flexible and widely used qualitative analysis method because it is not tied to a single theoretical framework: it can be applied across disciplines and paradigms to make sense of how people experience and talk about a topic.
A theme captures something important about the data in relation to the research question and represents a patterned response or meaning across the dataset. The goal is not to summarize what was said, but to interpret what it means.
Types of thematic analysis
"Thematic analysis" is really a family of approaches. Three key distinctions shape how you do it:
- Inductive vs. deductive — themes emerge from the data (inductive) or are guided by an existing theory or codebook (deductive). Most studies blend both.
- Semantic vs. latent — analysis stays at the explicit, surface meaning (semantic) or interprets underlying assumptions and ideas (latent).
- Reflexive vs. coding-reliability — themes are developed through the researcher's active, reflexive interpretation, or via a structured codebook with multiple coders and agreement checks. Braun and Clarke's later work emphasizes the reflexive form.
Braun and Clarke's six phases
The six-phase framework is the backbone of thematic analysis. The phases are recursive, not strictly linear — you move back and forth as your understanding develops.
- Familiarization with the data — read and re-read the entire dataset, listening to recordings if available, and note initial ideas. You cannot find patterns in data you do not know well.
- Generating initial codes — work systematically through the data, attaching concise codes to features that look interesting or relevant. Code inclusively and keep some surrounding context.
- Searching for themes — sort and collate codes into candidate themes, gathering all the data relevant to each. Visual tools like maps help here.
- Reviewing themes — check candidate themes against their coded extracts (do they cohere?) and against the whole dataset (do they fit?). Split, merge, or discard as needed.
- Defining and naming themes — write a clear definition of what each theme is and is not about, and give it a concise, informative name. Each theme should have a clear focus and tell part of the overall story.
- Producing the report — weave the themes into an analytic narrative, supported by vivid, representative extracts, that answers the research question and engages with the literature.
Codes vs. themes vs. topics
These are often confused, and the confusion weakens analysis:
- A code is a short label on a specific data segment (e.g., "rationing time").
- A theme is a broader pattern, built from many codes, that makes an analytic point (e.g., "control is exercised through small daily trade-offs").
- A topic is merely a subject area (e.g., "time"). A theme says something about a topic; a topic alone does not.
Doing thematic analysis with AI assistance
AI can accelerate the most labor-intensive phases — generating initial codes across a large dataset in minutes and clustering them into candidate themes — leaving the researcher to do the interpretive work of reviewing, defining, and narrating. Used in a human-in-the-loop way, this preserves the reflexive judgment at the heart of good thematic analysis while removing the bottleneck. See GenAI-Assisted Qualitative Research for how to keep it rigorous.
Common pitfalls
- Themes that are just topics — a list of subject areas is not analysis; a theme must make a point.
- Using interview questions as themes — organizing findings by the questions asked rather than by patterns of meaning.
- Thin evidence — themes unsupported by enough data, or supported only by cherry-picked quotes.
- Overlapping themes — themes that are not distinct, signaling the reviewing phase was rushed.
- Skipping familiarization — coding before you truly know the data.
Ensuring rigor
Thematic analysis is judged on trustworthiness, not statistics. Strengthen it with an audit trail of coding decisions, clear theme definitions, representative (not cherry-picked) extracts, reflexivity about your own influence, and — for coding-reliability variants — inter-coder agreement checks.
A worked example
- Familiarize — read all 18 interviews on remote-work wellbeing twice, noting impressions.
- Code — generate codes like "always-on guilt," "blurred boundaries," and "rituals that signal stop."
- Search — cluster codes into candidate themes such as "boundaries must be actively built."
- Review — test themes against extracts and the full dataset; merge two overlapping themes.
- Define — name and define each theme precisely.
- Report — write the narrative with representative quotes and ties to the literature.
Key takeaways
- Thematic analysis systematically finds and reports patterns of meaning (themes) across qualitative data.
- It varies along inductive/deductive, semantic/latent, and reflexive/coding-reliability dimensions.
- Braun and Clarke's six recursive phases run from familiarization to writing up.
- A theme makes an analytic point; it is more than a code and more than a topic.
- AI can accelerate coding and theme-searching, but interpretation stays human.
Frequently asked questions
What is thematic analysis?
It is a flexible qualitative method for systematically identifying, analyzing, and reporting patterns of meaning (themes) across a dataset such as interviews or open-ended responses, in order to interpret what the data means.
What are Braun and Clarke's six phases of thematic analysis?
Familiarization with the data; generating initial codes; searching for themes; reviewing themes; defining and naming themes; and producing the report. The phases are recursive rather than strictly linear.
What is the difference between a code and a theme?
A code is a short label on a specific segment of data; a theme is a broader pattern built from many codes that makes an analytic point about the dataset in relation to the research question.
What is reflexive thematic analysis?
It is Braun and Clarke's emphasized approach in which themes are actively developed through the researcher's interpretation and reflexivity, rather than via a fixed codebook and inter-coder agreement.
Is thematic analysis inductive or deductive?
It can be either or both. Inductive analysis lets themes emerge from the data; deductive analysis is guided by existing theory or a codebook. Many studies combine the two.
Can AI do thematic analysis?
AI can generate initial codes and propose candidate themes across a dataset quickly, but reviewing, defining, and interpreting themes should remain with the researcher in a human-in-the-loop workflow.