WHY ENGLISH?
Keep participants’ words, model suggestions and researcher interpretations distinct from transcript to theme
Qualitative research listens for experience, meaning, context and contradiction. Large language models may assist transcription, translation, coding or theme exploration, but English matters because fluent model output can blur the difference between what a participant said, what a tool suggested and what a researcher interpreted.
COREQ+LLM is registered with EQUATOR as an extension under development to the Consolidated Criteria for Reporting Qualitative Research. The proposal addresses LLM use across research-question design, interview protocols, transcription, translation, coding, theme identification, drafting and interactive exploration of qualitative datasets.
The EQUATOR record, registered 18 June 2025, says the planned extension will address model selection, training parameters, prompting strategies, task definitions, evaluation measures and human oversight. It also links a protocol preprint. As checked on 6 November 2026, no final checklist was presented on that record, so this guide explains documented risks and reporting logic without inventing completed items.
Article contents
Open all chapters
- 1. Start With the Qualitative Question
- 2. Describe the Research Team and Relationships
- 3. Explain Sampling, Recruitment and Context
- 4. Protect Consent and Sensitive Meaning
- 5. Identify the Model and Access Route
- 6. Describe the LLM Task With Concrete Verbs
- 7. Preserve Prompts, Context and Output Conditions
- 8. Keep Transcript and Translation Provenance
- 9. Build Codes Through Accountable Decisions
- 10. Develop Themes Beyond Automated Clustering
- 11. Make Reflexivity Operational
- 12. Validate the Specific Assistance Task
- 13. Audit Bias, Omission and Hallucination
- 14. Report Human Oversight as a Workflow
- 15. Support Trustworthiness and an Audit Trail
- 16. Conclude Without Letting the Model Speak for Participants
ORIENTATION
What this guide owns
Current fact check
EQUATOR lists COREQ+LLM as under development, registered 18 June 2025, and links protocol preprint DOI 10.2196/preprints.78682. The page describes intended reporting for model selection, parameters, prompting, tasks, evaluation and human oversight across qualitative research stages.
Canonical boundary
This article owns qualitative-research reporting when LLMs assist interpretation, under the registered COREQ+LLM proposal. Original COREQ owns interviews and focus groups generally, GUIDE-LLM owns behavioural and social-science LLM use, and DEAL owns technical LLM development and application.
Worked example
Every chapter returns to a fictional interview study exploring how first-year university students describe belonging during the transition from school. It is fictional, low-stakes and educational; it does not report a real school, clinical decision or guaranteed outcome.
Did You Know? For coreq_llm, complete reporting makes appraisal possible but does not perform the appraisal, remove bias or guarantee transfer.
Warm up with eduKate’s guides to explaining scientific evidence and asking better questions in class.
CHAPTER 1 OF 16
1. Start With the Qualitative Question
Why this move matters
State the experience or meaning being explored, theoretical orientation and why qualitative inquiry is appropriate before naming an LLM.
Did You Know? Technology should serve the interpretive question rather than redefine it.
Preserve qualitative meaning and provenance
State the experience or meaning being explored, theoretical orientation and why qualitative inquiry is appropriate before naming an LLM. Show which human account, transcript segment, prompt, model output and researcher interpretation support each analytical move.
Follow one passage through the workflow
Worked example: The study asks how students describe belonging, not whether an LLM can classify positive and negative comments. Keep the original words beside transcription, translation, coding suggestions, researcher decisions and the final theme.
Make reflexivity operational
Name who selected the model, wrote prompts, accepted or rejected suggestions and how their theoretical stance shaped the analysis. Human oversight is a described method, not a reassurance.
Respect the proposal boundary
A strong question can still be answered by a weak sample or shallow analysis. COREQ+LLM is under development. Its registered scope names model selection, parameters, prompts, task definitions, evaluation and oversight; this article does not present a nonexistent final checklist.
Practical reader check
- Point to the exact source or method sentence supporting the claim.
- Name the person, setting, date, input and decision rule that could change its meaning.
- Label the remaining uncertainty reported, inferred, contested or unknown.
Student, parent and career transfer
Make one evidence card for “Start With the Qualitative Question”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports qualitative research, education, psychology, sociology, public health, user research, ethics, translation and research communication without promising a grade, admission, funding decision or job.
CHAPTER 2 OF 16
2. Describe the Research Team and Relationships
Why this move matters
Report researcher roles, experience, positionality, prior relationships, interviewer identity and what participants knew.
Did You Know? COREQ begins from the people who produce qualitative data together.
Preserve qualitative meaning and provenance
Report researcher roles, experience, positionality, prior relationships, interviewer identity and what participants knew. Show which human account, transcript segment, prompt, model output and researcher interpretation support each analytical move.
Follow one passage through the workflow
Worked example: The interviewer is a university staff member and explains that relationship in the report. Keep the original words beside transcription, translation, coding suggestions, researcher decisions and the final theme.
Make reflexivity operational
Name who selected the model, wrote prompts, accepted or rejected suggestions and how their theoretical stance shaped the analysis. Human oversight is a described method, not a reassurance.
Respect the proposal boundary
A reflexivity statement cannot remove power differences. COREQ+LLM is under development. Its registered scope names model selection, parameters, prompts, task definitions, evaluation and oversight; this article does not present a nonexistent final checklist.
Practical reader check
- Point to the exact source or method sentence supporting the claim.
- Name the person, setting, date, input and decision rule that could change its meaning.
- Label the remaining uncertainty reported, inferred, contested or unknown.
Student, parent and career transfer
Make one evidence card for “Describe the Research Team and Relationships”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports qualitative research, education, psychology, sociology, public health, user research, ethics, translation and research communication without promising a grade, admission, funding decision or job.
CHAPTER 3 OF 16
3. Explain Sampling, Recruitment and Context
Why this move matters
Describe who was invited, by whom, setting, non-participation, incentives, sample rationale and relevant participant characteristics.
Did You Know? Readers need context to judge whose meanings entered the analysis.
Preserve qualitative meaning and provenance
Describe who was invited, by whom, setting, non-participation, incentives, sample rationale and relevant participant characteristics. Show which human account, transcript segment, prompt, model output and researcher interpretation support each analytical move.
Follow one passage through the workflow
Worked example: Students are recruited across programmes and participation is separated from grading. Keep the original words beside transcription, translation, coding suggestions, researcher decisions and the final theme.
Make reflexivity operational
Name who selected the model, wrote prompts, accepted or rejected suggestions and how their theoretical stance shaped the analysis. Human oversight is a described method, not a reassurance.
Respect the proposal boundary
A varied sample is not automatically representative. COREQ+LLM is under development. Its registered scope names model selection, parameters, prompts, task definitions, evaluation and oversight; this article does not present a nonexistent final checklist.
Practical reader check
- Point to the exact source or method sentence supporting the claim.
- Name the person, setting, date, input and decision rule that could change its meaning.
- Label the remaining uncertainty reported, inferred, contested or unknown.
Student, parent and career transfer
Make one evidence card for “Explain Sampling, Recruitment and Context”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports qualitative research, education, psychology, sociology, public health, user research, ethics, translation and research communication without promising a grade, admission, funding decision or job.
CHAPTER 4 OF 16
4. Protect Consent and Sensitive Meaning
Why this move matters
State consent, confidentiality, withdrawal, de-identification, data-access boundaries and whether LLM processing was disclosed.
Did You Know? Sending qualitative text to a model is a data-governance decision, not only a software choice.
Preserve qualitative meaning and provenance
State consent, confidentiality, withdrawal, de-identification, data-access boundaries and whether LLM processing was disclosed. Show which human account, transcript segment, prompt, model output and researcher interpretation support each analytical move.
Follow one passage through the workflow
Worked example: Participants consent separately to approved machine-assisted analysis of de-identified transcripts. Keep the original words beside transcription, translation, coding suggestions, researcher decisions and the final theme.
Make reflexivity operational
Name who selected the model, wrote prompts, accepted or rejected suggestions and how their theoretical stance shaped the analysis. Human oversight is a described method, not a reassurance.
Respect the proposal boundary
De-identification may fail when narratives contain distinctive events. COREQ+LLM is under development. Its registered scope names model selection, parameters, prompts, task definitions, evaluation and oversight; this article does not present a nonexistent final checklist.
Practical reader check
- Point to the exact source or method sentence supporting the claim.
- Name the person, setting, date, input and decision rule that could change its meaning.
- Label the remaining uncertainty reported, inferred, contested or unknown.
Student, parent and career transfer
Make one evidence card for “Protect Consent and Sensitive Meaning”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports qualitative research, education, psychology, sociology, public health, user research, ethics, translation and research communication without promising a grade, admission, funding decision or job.
CHAPTER 5 OF 16
5. Identify the Model and Access Route
Why this move matters
Give model, provider, version, local or hosted access, date, settings and known changes during the study.
Did You Know? COREQ+LLM was proposed because naming software alone is no longer enough.
Preserve qualitative meaning and provenance
Give model, provider, version, local or hosted access, date, settings and known changes during the study. Show which human account, transcript segment, prompt, model output and researcher interpretation support each analytical move.
Follow one passage through the workflow
Worked example: A local model version and checksum are recorded for the coding-assistance stage. Keep the original words beside transcription, translation, coding suggestions, researcher decisions and the final theme.
Make reflexivity operational
Name who selected the model, wrote prompts, accepted or rejected suggestions and how their theoretical stance shaped the analysis. Human oversight is a described method, not a reassurance.
Respect the proposal boundary
Exact identity may be impossible for a changing commercial service. COREQ+LLM is under development. Its registered scope names model selection, parameters, prompts, task definitions, evaluation and oversight; this article does not present a nonexistent final checklist.
Practical reader check
- Point to the exact source or method sentence supporting the claim.
- Name the person, setting, date, input and decision rule that could change its meaning.
- Label the remaining uncertainty reported, inferred, contested or unknown.
Student, parent and career transfer
Make one evidence card for “Identify the Model and Access Route”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports qualitative research, education, psychology, sociology, public health, user research, ethics, translation and research communication without promising a grade, admission, funding decision or job.
CHAPTER 6 OF 16
6. Describe the LLM Task With Concrete Verbs
Why this move matters
Say whether the model transcribed, translated, summarised, proposed codes, clustered excerpts, queried data or drafted text.
Did You Know? Different LLM tasks create different risks to meaning and provenance.
Preserve qualitative meaning and provenance
Say whether the model transcribed, translated, summarised, proposed codes, clustered excerpts, queried data or drafted text. Show which human account, transcript segment, prompt, model output and researcher interpretation support each analytical move.
Follow one passage through the workflow
Worked example: The model proposes candidate codes but never decides the final codebook. Keep the original words beside transcription, translation, coding suggestions, researcher decisions and the final theme.
Make reflexivity operational
Name who selected the model, wrote prompts, accepted or rejected suggestions and how their theoretical stance shaped the analysis. Human oversight is a described method, not a reassurance.
Respect the proposal boundary
A precise task label does not establish validity. COREQ+LLM is under development. Its registered scope names model selection, parameters, prompts, task definitions, evaluation and oversight; this article does not present a nonexistent final checklist.
Practical reader check
- Point to the exact source or method sentence supporting the claim.
- Name the person, setting, date, input and decision rule that could change its meaning.
- Label the remaining uncertainty reported, inferred, contested or unknown.
Student, parent and career transfer
Make one evidence card for “Describe the LLM Task With Concrete Verbs”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports qualitative research, education, psychology, sociology, public health, user research, ethics, translation and research communication without promising a grade, admission, funding decision or job.
CHAPTER 7 OF 16
7. Preserve Prompts, Context and Output Conditions
Why this move matters
Report system and user prompts, examples, conversation state, context windows, chunking, temperature, repetitions and output schemas.
Did You Know? Prompt language is part of the qualitative method when it shapes analytical suggestions.
Preserve qualitative meaning and provenance
Report system and user prompts, examples, conversation state, context windows, chunking, temperature, repetitions and output schemas. Show which human account, transcript segment, prompt, model output and researcher interpretation support each analytical move.
Follow one passage through the workflow
Worked example: Each transcript is processed in versioned chunks with a stable prompt and source identifier. Keep the original words beside transcription, translation, coding suggestions, researcher decisions and the final theme.
Make reflexivity operational
Name who selected the model, wrote prompts, accepted or rejected suggestions and how their theoretical stance shaped the analysis. Human oversight is a described method, not a reassurance.
Respect the proposal boundary
A complete prompt may still behave differently after a service update. COREQ+LLM is under development. Its registered scope names model selection, parameters, prompts, task definitions, evaluation and oversight; this article does not present a nonexistent final checklist.
Practical reader check
- Point to the exact source or method sentence supporting the claim.
- Name the person, setting, date, input and decision rule that could change its meaning.
- Label the remaining uncertainty reported, inferred, contested or unknown.
Student, parent and career transfer
Make one evidence card for “Preserve Prompts, Context and Output Conditions”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports qualitative research, education, psychology, sociology, public health, user research, ethics, translation and research communication without promising a grade, admission, funding decision or job.
CHAPTER 8 OF 16
8. Keep Transcript and Translation Provenance
Why this move matters
Describe recording, transcription, correction, translation, dialect handling and links from processed text to the original.
Did You Know? Every transformation should leave a trail back to the participant’s account.
Preserve qualitative meaning and provenance
Describe recording, transcription, correction, translation, dialect handling and links from processed text to the original. Show which human account, transcript segment, prompt, model output and researcher interpretation support each analytical move.
Follow one passage through the workflow
Worked example: Bilingual reviewers preserve the original phrase beside the English translation and note disputed meanings. Keep the original words beside transcription, translation, coding suggestions, researcher decisions and the final theme.
Make reflexivity operational
Name who selected the model, wrote prompts, accepted or rejected suggestions and how their theoretical stance shaped the analysis. Human oversight is a described method, not a reassurance.
Respect the proposal boundary
A polished translation can erase ambiguity important to the study. COREQ+LLM is under development. Its registered scope names model selection, parameters, prompts, task definitions, evaluation and oversight; this article does not present a nonexistent final checklist.
Practical reader check
- Point to the exact source or method sentence supporting the claim.
- Name the person, setting, date, input and decision rule that could change its meaning.
- Label the remaining uncertainty reported, inferred, contested or unknown.
Student, parent and career transfer
Make one evidence card for “Keep Transcript and Translation Provenance”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports qualitative research, education, psychology, sociology, public health, user research, ethics, translation and research communication without promising a grade, admission, funding decision or job.
CHAPTER 9 OF 16
9. Build Codes Through Accountable Decisions
Why this move matters
Explain whether codes were inductive or deductive, who created them, how LLM suggestions were reviewed and how disagreements changed the codebook.
Did You Know? A model can propose patterns without owning their interpretation.
Preserve qualitative meaning and provenance
Explain whether codes were inductive or deductive, who created them, how LLM suggestions were reviewed and how disagreements changed the codebook. Show which human account, transcript segment, prompt, model output and researcher interpretation support each analytical move.
Follow one passage through the workflow
Worked example: Researchers accept, revise or reject every suggested code with a brief reason. Keep the original words beside transcription, translation, coding suggestions, researcher decisions and the final theme.
Make reflexivity operational
Name who selected the model, wrote prompts, accepted or rejected suggestions and how their theoretical stance shaped the analysis. Human oversight is a described method, not a reassurance.
Respect the proposal boundary
Human review can become superficial when suggestions arrive at scale. COREQ+LLM is under development. Its registered scope names model selection, parameters, prompts, task definitions, evaluation and oversight; this article does not present a nonexistent final checklist.
Practical reader check
- Point to the exact source or method sentence supporting the claim.
- Name the person, setting, date, input and decision rule that could change its meaning.
- Label the remaining uncertainty reported, inferred, contested or unknown.
Student, parent and career transfer
Make one evidence card for “Build Codes Through Accountable Decisions”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports qualitative research, education, psychology, sociology, public health, user research, ethics, translation and research communication without promising a grade, admission, funding decision or job.
CHAPTER 10 OF 16
10. Develop Themes Beyond Automated Clustering
Why this move matters
Show how codes became themes, how negative cases were sought, how coherence was tested and where model suggestions were discarded.
Did You Know? Themes are reasoned interpretations, not labels emitted by software.
Preserve qualitative meaning and provenance
Show how codes became themes, how negative cases were sought, how coherence was tested and where model suggestions were discarded. Show which human account, transcript segment, prompt, model output and researcher interpretation support each analytical move.
Follow one passage through the workflow
Worked example: A frequent cluster is rejected as a theme because it lacks explanatory meaning across cases. Keep the original words beside transcription, translation, coding suggestions, researcher decisions and the final theme.
Make reflexivity operational
Name who selected the model, wrote prompts, accepted or rejected suggestions and how their theoretical stance shaped the analysis. Human oversight is a described method, not a reassurance.
Respect the proposal boundary
Frequency and semantic similarity are not the same as qualitative importance. COREQ+LLM is under development. Its registered scope names model selection, parameters, prompts, task definitions, evaluation and oversight; this article does not present a nonexistent final checklist.
Practical reader check
- Point to the exact source or method sentence supporting the claim.
- Name the person, setting, date, input and decision rule that could change its meaning.
- Label the remaining uncertainty reported, inferred, contested or unknown.
Student, parent and career transfer
Make one evidence card for “Develop Themes Beyond Automated Clustering”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports qualitative research, education, psychology, sociology, public health, user research, ethics, translation and research communication without promising a grade, admission, funding decision or job.
CHAPTER 11 OF 16
11. Make Reflexivity Operational
Why this move matters
Describe how researchers’ assumptions, prompts, model choice and reactions to outputs shaped the analysis.
Did You Know? LLM use adds a new layer of influence but does not replace human positionality.
Preserve qualitative meaning and provenance
Describe how researchers’ assumptions, prompts, model choice and reactions to outputs shaped the analysis. Show which human account, transcript segment, prompt, model output and researcher interpretation support each analytical move.
Follow one passage through the workflow
Worked example: The team records when a model suggestion confirmed an expected story and deliberately searches for contrary passages. Keep the original words beside transcription, translation, coding suggestions, researcher decisions and the final theme.
Make reflexivity operational
Name who selected the model, wrote prompts, accepted or rejected suggestions and how their theoretical stance shaped the analysis. Human oversight is a described method, not a reassurance.
Respect the proposal boundary
Reflexivity can become ceremonial if it never changes a decision. COREQ+LLM is under development. Its registered scope names model selection, parameters, prompts, task definitions, evaluation and oversight; this article does not present a nonexistent final checklist.
Practical reader check
- Point to the exact source or method sentence supporting the claim.
- Name the person, setting, date, input and decision rule that could change its meaning.
- Label the remaining uncertainty reported, inferred, contested or unknown.
Student, parent and career transfer
Make one evidence card for “Make Reflexivity Operational”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports qualitative research, education, psychology, sociology, public health, user research, ethics, translation and research communication without promising a grade, admission, funding decision or job.
CHAPTER 12 OF 16
12. Validate the Specific Assistance Task
Why this move matters
Use human reference work, double review, error categories, stability checks or participant feedback appropriate to the task.
Did You Know? Validation should match what the model actually did.
Preserve qualitative meaning and provenance
Use human reference work, double review, error categories, stability checks or participant feedback appropriate to the task. Show which human account, transcript segment, prompt, model output and researcher interpretation support each analytical move.
Follow one passage through the workflow
Worked example: A blinded sample compares model-assisted and independent human coding for missed nuance and invented content. Keep the original words beside transcription, translation, coding suggestions, researcher decisions and the final theme.
Make reflexivity operational
Name who selected the model, wrote prompts, accepted or rejected suggestions and how their theoretical stance shaped the analysis. Human oversight is a described method, not a reassurance.
Respect the proposal boundary
Agreement cannot prove that an interpretation is the best one. COREQ+LLM is under development. Its registered scope names model selection, parameters, prompts, task definitions, evaluation and oversight; this article does not present a nonexistent final checklist.
Practical reader check
- Point to the exact source or method sentence supporting the claim.
- Name the person, setting, date, input and decision rule that could change its meaning.
- Label the remaining uncertainty reported, inferred, contested or unknown.
Student, parent and career transfer
Make one evidence card for “Validate the Specific Assistance Task”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports qualitative research, education, psychology, sociology, public health, user research, ethics, translation and research communication without promising a grade, admission, funding decision or job.
CHAPTER 13 OF 16
13. Audit Bias, Omission and Hallucination
Why this move matters
Check whether the model flattens minority language, stereotypes groups, omits contradictions or invents explanations and quotations.
Did You Know? Fluency makes unsupported output easier, not harder, to overlook.
Preserve qualitative meaning and provenance
Check whether the model flattens minority language, stereotypes groups, omits contradictions or invents explanations and quotations. Show which human account, transcript segment, prompt, model output and researcher interpretation support each analytical move.
Follow one passage through the workflow
Worked example: The audit tracks dialect-related coding differences and verifies every quotation against transcript audio. Keep the original words beside transcription, translation, coding suggestions, researcher decisions and the final theme.
Make reflexivity operational
Name who selected the model, wrote prompts, accepted or rejected suggestions and how their theoretical stance shaped the analysis. Human oversight is a described method, not a reassurance.
Respect the proposal boundary
A finite audit cannot prove absence of bias. COREQ+LLM is under development. Its registered scope names model selection, parameters, prompts, task definitions, evaluation and oversight; this article does not present a nonexistent final checklist.
Practical reader check
- Point to the exact source or method sentence supporting the claim.
- Name the person, setting, date, input and decision rule that could change its meaning.
- Label the remaining uncertainty reported, inferred, contested or unknown.
Student, parent and career transfer
Make one evidence card for “Audit Bias, Omission and Hallucination”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports qualitative research, education, psychology, sociology, public health, user research, ethics, translation and research communication without promising a grade, admission, funding decision or job.
CHAPTER 14 OF 16
14. Report Human Oversight as a Workflow
Why this move matters
Name who reviewed which outputs, review depth, expertise, escalation, disagreements, time pressure and final authority.
Did You Know? Oversight becomes credible when readers can see actions and responsibility.
Preserve qualitative meaning and provenance
Name who reviewed which outputs, review depth, expertise, escalation, disagreements, time pressure and final authority. Show which human account, transcript segment, prompt, model output and researcher interpretation support each analytical move.
Follow one passage through the workflow
Worked example: Senior analysts review themes while trained coders verify all source anchors. Keep the original words beside transcription, translation, coding suggestions, researcher decisions and the final theme.
Make reflexivity operational
Name who selected the model, wrote prompts, accepted or rejected suggestions and how their theoretical stance shaped the analysis. Human oversight is a described method, not a reassurance.
Respect the proposal boundary
The phrase human in the loop says nothing about the quality of the loop. COREQ+LLM is under development. Its registered scope names model selection, parameters, prompts, task definitions, evaluation and oversight; this article does not present a nonexistent final checklist.
Practical reader check
- Point to the exact source or method sentence supporting the claim.
- Name the person, setting, date, input and decision rule that could change its meaning.
- Label the remaining uncertainty reported, inferred, contested or unknown.
Student, parent and career transfer
Make one evidence card for “Report Human Oversight as a Workflow”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports qualitative research, education, psychology, sociology, public health, user research, ethics, translation and research communication without promising a grade, admission, funding decision or job.
CHAPTER 15 OF 16
15. Support Trustworthiness and an Audit Trail
Why this move matters
Explain credibility, dependability, confirmability, transferability, member reflection, negative cases and access to codebooks, prompts and decision logs.
Did You Know? Trustworthiness requires both interpretive reasoning and inspectable provenance.
Preserve qualitative meaning and provenance
Explain credibility, dependability, confirmability, transferability, member reflection, negative cases and access to codebooks, prompts and decision logs. Show which human account, transcript segment, prompt, model output and researcher interpretation support each analytical move.
Follow one passage through the workflow
Worked example: A controlled audit trail links themes to excerpts, prompt versions and analytical memos. Keep the original words beside transcription, translation, coding suggestions, researcher decisions and the final theme.
Make reflexivity operational
Name who selected the model, wrote prompts, accepted or rejected suggestions and how their theoretical stance shaped the analysis. Human oversight is a described method, not a reassurance.
Respect the proposal boundary
An audit trail cannot rescue a question that ignored important voices. COREQ+LLM is under development. Its registered scope names model selection, parameters, prompts, task definitions, evaluation and oversight; this article does not present a nonexistent final checklist.
Practical reader check
- Point to the exact source or method sentence supporting the claim.
- Name the person, setting, date, input and decision rule that could change its meaning.
- Label the remaining uncertainty reported, inferred, contested or unknown.
Student, parent and career transfer
Make one evidence card for “Support Trustworthiness and an Audit Trail”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports qualitative research, education, psychology, sociology, public health, user research, ethics, translation and research communication without promising a grade, admission, funding decision or job.
CHAPTER 16 OF 16
16. Conclude Without Letting the Model Speak for Participants
Why this move matters
Use verified quotations, distinguish interpretation from generated prose and state context, dissent and transfer limits.
Did You Know? Because COREQ+LLM is under development, readers should check EQUATOR for the eventual final extension.
Preserve qualitative meaning and provenance
Use verified quotations, distinguish interpretation from generated prose and state context, dissent and transfer limits. Show which human account, transcript segment, prompt, model output and researcher interpretation support each analytical move.
Follow one passage through the workflow
Worked example: The conclusion reports several ways students experience belonging rather than an AI-generated universal story. Keep the original words beside transcription, translation, coding suggestions, researcher decisions and the final theme.
Make reflexivity operational
Name who selected the model, wrote prompts, accepted or rejected suggestions and how their theoretical stance shaped the analysis. Human oversight is a described method, not a reassurance.
Respect the proposal boundary
Complete reporting cannot make one study representative of all learners. COREQ+LLM is under development. Its registered scope names model selection, parameters, prompts, task definitions, evaluation and oversight; this article does not present a nonexistent final checklist.
Practical reader check
- Point to the exact source or method sentence supporting the claim.
- Name the person, setting, date, input and decision rule that could change its meaning.
- Label the remaining uncertainty reported, inferred, contested or unknown.
Student, parent and career transfer
Make one evidence card for “Conclude Without Letting the Model Speak for Participants”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports qualitative research, education, psychology, sociology, public health, user research, ethics, translation and research communication without promising a grade, admission, funding decision or job.
PRACTICE LAB
Build a coreq_llm decision trail
Student route
- Choose one original report inside this guide’s scope.
- Copy its objective and underline actor, action, context, evidence and intended decision.
- Draw the specific reasoning chain used in this article.
- Mark each step reported, inferred, contested or unknown.
- Rewrite the conclusion so it preserves the narrowest important limitation.
Parent or teacher route
Ask the learner to explain keep participants’ words, model suggestions and researcher interpretations distinct from transcript to theme twice: once to a classmate and once to a method specialist. Praise a precise correction and a well-stated unknown; both show disciplined attention to evidence.
Early-career route
Create a one-page coreq_llm audit with source link, date checked, scope, method logic, decisive wording, contrary evidence, limit and next action. A colleague should be able to continue without guessing.
Decision rule
If a shorter explanation removes the condition that made the original claim true, it is not clearer yet. Restore the condition.
Continue with eduKate’s guides to reading course and career information and giving useful feedback in group projects.
USEFUL NEXT READING
Connect this narrow owner to its neighbour
Read the PRISMA-trAIce AI-assisted review guide for a related but different reader job. Return here when the question is specifically how to keep participants’ words, model suggestions and researcher interpretations distinct from transcript to theme. This boundary keeps one canonical owner per intent.
For English practice across research, school and career communication, visit the English at work and in public life guide.
FREQUENTLY ASKED QUESTIONS
Questions students and parents often ask
What is COREQ+LLM?
A proposed extension of COREQ for qualitative research that uses large language models.
Is the checklist final?
No. EQUATOR lists it as under development.
Can an LLM create themes?
It can suggest patterns, but accountable qualitative interpretation requires transparent human reasoning and source provenance.
Why publish prompts?
Prompts can shape coding and summaries, so they are part of the method.
What is reflexivity here?
It is examining how researchers, model choices and interactions influenced the analysis.
Can de-identified transcripts be risk-free?
No. Rich narratives can remain re-identifiable, and service retention or reuse must be governed.
PRIMARY SOURCES
Read the original coreq_llm guidance
Facts were checked against original publications and official records on 5 November 2026. Follow the live source for later corrections, completed updates or companion material.
- COREQ+LLM under-development record, EQUATOR Network
- COREQ+LLM protocol preprint DOI
- Original COREQ record, EQUATOR Network
A cheerful final note
Precise English makes keep participants’ words, model suggestions and researcher interpretations distinct from transcript to theme more generous, not less. People can see how their evidence and values entered the work, where disagreement remains and what a responsible next step looks like.