"We used thematic analysis" is one of the least informative sentences in a methods section. It names a family of approaches that differ in philosophy, in how they treat a codebook, and in what counts as good quality. A reviewer who sees the phrase will immediately ask which one you mean, and the answer decides whether your reliability claims make sense or contradict your own method.
Three versions, not one
Virginia Braun and Victoria Clarke have argued in their 2021 paper on quality practice in thematic analysis that the label covers three broad versions. Knowing which one you are working in is the single most useful decision you can make before you start coding.
Coding reliability thematic analysis
This version sits close to classic content analysis. A coding frame is fixed early, often before the analysis proper begins; several coders apply it independently; agreement between them is measured and treated as evidence of quality. Boyatzis is the usual reference, and Guest and colleagues and Joffe are associated with this tradition too. The underlying worry is bias, and the remedy is consistency.
Codebook thematic analysis
Here a structured coding frame is also developed early, but the working assumptions are more qualitative. Framework analysis, template analysis and matrix analysis belong in this group. Teams are common, the frame can be revised as the data push back, and agreement statistics are used pragmatically rather than as the headline quality measure. It is a sensible fit for applied and team-based projects with a deadline and a defined question.
Reflexive thematic analysis
This is the approach Braun and Clarke themselves develop. There is no codebook. Coding is open and evolves; the researcher's subjectivity is treated as a resource rather than a contaminant; and themes are the result of coding, not the input to it. A theme here is a pattern of shared meaning organised around a central idea, not simply a topic that came up repeatedly. Because the method does not treat coding as measurement, coder agreement is not meaningful as a quality criterion, and the authors explicitly reject it as one.
What this means for reliability claims
The practical consequence is a simple matrix. If you work in the coding reliability tradition, a kappa between independent coders is expected, and leaving it out needs an explanation. If you work in the codebook tradition, a second coder is useful for refining definitions and for team consistency, but the number is not the point. If you work reflexively, reporting a kappa is not a sign of rigour; it signals that you have not understood the method you cite, and rigour has to be shown through transparency about your decisions, your positioning and how the analysis developed.
For the content-analytic case, our article on intercoder reliability when AI does the coding shows how to run a defensible check. It does not apply to a reflexive analysis, and that is not a gap in the method.
Typical mistake: the unacknowledged hybrid
The most common problem is a methods section that cites the reflexive approach, then describes a fixed codebook defined in advance and a reliability coefficient between two coders. Braun and Clarke are not against combining ideas, but they ask that a combination be justified and coherent. Citing the reflexive label while using procedures from the other two, without saying so, creates a procedural contradiction that is easy for a reviewer to spot.
A second variant is the topic-summary theme. In the first two versions, themes often are domains of data, such as "communication" or "workload". That is acceptable there. Under the reflexive version it is the thing to avoid, since a theme should make a claim about the data. Our walkthrough of the six phases of reflexive thematic analysis covers how to tell the difference.
A quick way to decide which one you are doing
Ask yourself four questions. Did you fix categories before reading the material, or did they emerge from your coding? Will a second person code independently, and will you report their agreement? Do you see your own position as something to account for or something to control? Is your supervisor or journal community qualitative-interpretive or closer to content analysis? If the answers lean towards fixed categories and agreement, you are in the first or second family. If they lean towards emergence and positioning, you are in the third. Mixed answers are a signal to say explicitly what you borrowed from where.
What software can and cannot do in each version
In a codebook or coding reliability design, the mechanical work is large: applying a defined frame consistently across hundreds of responses. Themera is built for that. Every code is backed by a verbatim quote, and quotes that are not exactly in the material are rejected automatically. A coverage indicator shows which responses received no code, you can edit the codebook and re-code the material, and the intercoder agreement check (blind re-coding with Cohen's kappa) is there when your tradition calls for it.
In a reflexive design, the tool should be positioned more modestly. It can give you a fast, quote-grounded first pass that helps you see the material, but the interpretive work of developing themes is the method itself and stays with you. If you use such a pass, say so, and describe how you revised it. Our guide to coding qualitative data covers the mechanics of that first stage.
A formulation for your methods section
Adapt this to what you actually did. The wording below is a template, not a citation.
We conducted a codebook thematic analysis in the sense described by Braun and Clarke (2021). A coding frame was developed from the research questions and refined after coding the first ten interviews. Themes were defined at the level of topic domains. A second researcher independently coded a subset of the material to check the clarity of the category definitions; we treat the resulting agreement as a check on the codebook rather than as a measure of objectivity. Computational support was used to apply the frame across all responses, and every assignment was verified against its verbatim extract.
For a reflexive analysis, replace the second half with a description of how your coding evolved, your positioning, and the points at which you revised your themes. When in doubt about which framing your department expects, ask your supervisor; this is a methodological, not a technical question.
In short
Name the version of thematic analysis you are using, match your quality claims to it, and say plainly where you borrowed from another. That takes one paragraph and prevents the most common objection.
See a worked example analysis with codebook, verbatim quotes and coverage, no account needed, or try it on your own data: three analyses free, no card needed.