Why English? | Reporting GUIDE-AI Use Across Health Guidelines Through Tasks, Evidence, Human Review, Bias, Recommendations and Updates

ORIENTATION

What this guide owns

Current fact check

The official EQUATOR record describes AI use across the complete guideline lifecycle and a development process involving literature review, item drafting, comparison, multidisciplinary stakeholders, iterative voting and publication. It does not establish that any particular model is accurate, private or suitable for recommendation decisions.

Canonical boundary

This article owns transparent reporting of AI tasks across health-guideline work under GUIDE-AI. RIGHT owns reporting of health-practice guidelines generally, AGREE-related tools address appraisal, review-specific AI guidance owns systematic-review stages, and local guideline organisations retain responsibility for evidence and recommendations.

Worked example

Every chapter returns to a fictional national guideline team using AI to help update recommendations on adolescent sleep while human methodologists and a multidisciplinary panel retain approval authority. It is fictional, low-stakes and educational; it does not report a real school, clinical decision or guaranteed outcome.

Warm up with eduKate’s guides to explaining scientific evidence and asking better questions in class.

CHAPTER 1 OF 16

1. Name the Guideline Stage and AI Task

Back to contents

Why this move matters

State whether AI supports planning, searching, screening, extraction, synthesis, certainty, wording, implementation, evaluation or appraisal.

Place AI inside the guideline lifecycle

State whether AI supports planning, searching, screening, extraction, synthesis, certainty, wording, implementation, evaluation or appraisal. Name the stage, task, evidence input, output, model, human role and decision consequence.

Audit one AI-assisted recommendation

Worked example: AI proposes search-term expansions but does not approve eligible evidence. Trace question framing, search, screening, extraction, synthesis, certainty, wording, review and update.

Keep assistance separate from authority

AI may suggest, sort, draft or compare. State who verifies evidence, judges certainty, weighs values and approves recommendations.

Respect the development boundary

A task label can hide several models and decisions. GUIDE-AI is under development. Its official project record describes scope and methods, not a final checklist or endorsement of any tool.

Student, parent and career transfer

Make one evidence card for “Name the Guideline Stage and AI 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 medicine, nursing, allied health, public health, guideline methods, evidence synthesis, data science, regulation, patient advocacy, policy and implementation science without promising a grade, admission, funding decision or job.

CHAPTER 2 OF 16

2. Define the Guideline Question and Intended Users

Back to contents

Why this move matters

Report population, intervention or issue, comparisons, outcomes, setting, users and decisions before describing AI.

Place AI inside the guideline lifecycle

Report population, intervention or issue, comparisons, outcomes, setting, users and decisions before describing AI. Name the stage, task, evidence input, output, model, human role and decision consequence.

Audit one AI-assisted recommendation

Worked example: The update asks which school-based approaches improve adolescent sleep and for whom. Trace question framing, search, screening, extraction, synthesis, certainty, wording, review and update.

Keep assistance separate from authority

AI may suggest, sort, draft or compare. State who verifies evidence, judges certainty, weighs values and approves recommendations.

Respect the development boundary

A well-framed question does not guarantee complete evidence. GUIDE-AI is under development. Its official project record describes scope and methods, not a final checklist or endorsement of any tool.

Student, parent and career transfer

Make one evidence card for “Define the Guideline Question and Intended Users”: 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 medicine, nursing, allied health, public health, guideline methods, evidence synthesis, data science, regulation, patient advocacy, policy and implementation science without promising a grade, admission, funding decision or job.

CHAPTER 3 OF 16

3. Describe the Evidence Inputs

Back to contents

Why this move matters

Name databases, documents, dates, formats, inclusion boundaries, retrieval process and whether the model had access to full text.

Place AI inside the guideline lifecycle

Name databases, documents, dates, formats, inclusion boundaries, retrieval process and whether the model had access to full text. Name the stage, task, evidence input, output, model, human role and decision consequence.

Audit one AI-assisted recommendation

Worked example: The model receives only verified included studies and structured extraction fields. Trace question framing, search, screening, extraction, synthesis, certainty, wording, review and update.

Keep assistance separate from authority

AI may suggest, sort, draft or compare. State who verifies evidence, judges certainty, weighs values and approves recommendations.

Respect the development boundary

A curated input can still contain biased or outdated evidence. GUIDE-AI is under development. Its official project record describes scope and methods, not a final checklist or endorsement of any tool.

Student, parent and career transfer

Make one evidence card for “Describe the Evidence Inputs”: 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 medicine, nursing, allied health, public health, guideline methods, evidence synthesis, data science, regulation, patient advocacy, policy and implementation science without promising a grade, admission, funding decision or job.

CHAPTER 4 OF 16

4. Trace Search, Screening and Extraction

Back to contents

Why this move matters

Report which steps are automated, thresholds, calibration, sampling, reviewer roles, disagreement and missed-record checks.

Place AI inside the guideline lifecycle

Report which steps are automated, thresholds, calibration, sampling, reviewer roles, disagreement and missed-record checks. Name the stage, task, evidence input, output, model, human role and decision consequence.

Audit one AI-assisted recommendation

Worked example: Humans screen all exclusions from a low-confidence AI band and audit a random sample of the rest. Trace question framing, search, screening, extraction, synthesis, certainty, wording, review and update.

Keep assistance separate from authority

AI may suggest, sort, draft or compare. State who verifies evidence, judges certainty, weighs values and approves recommendations.

Respect the development boundary

Sampling may miss rare systematic errors. GUIDE-AI is under development. Its official project record describes scope and methods, not a final checklist or endorsement of any tool.

Student, parent and career transfer

Make one evidence card for “Trace Search, Screening and Extraction”: 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 medicine, nursing, allied health, public health, guideline methods, evidence synthesis, data science, regulation, patient advocacy, policy and implementation science without promising a grade, admission, funding decision or job.

CHAPTER 5 OF 16

5. Identify Model, Platform and Version

Back to contents

Why this move matters

Name provider, model, release, access date, interface, retrieval system, updates and whether the service is reproducible.

Place AI inside the guideline lifecycle

Name provider, model, release, access date, interface, retrieval system, updates and whether the service is reproducible. Name the stage, task, evidence input, output, model, human role and decision consequence.

Audit one AI-assisted recommendation

Worked example: The team freezes the model version and records the hosted-service date for each run. Trace question framing, search, screening, extraction, synthesis, certainty, wording, review and update.

Keep assistance separate from authority

AI may suggest, sort, draft or compare. State who verifies evidence, judges certainty, weighs values and approves recommendations.

Respect the development boundary

A vendor name alone cannot preserve a changing service. GUIDE-AI is under development. Its official project record describes scope and methods, not a final checklist or endorsement of any tool.

Student, parent and career transfer

Make one evidence card for “Identify Model, Platform and Version”: 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 medicine, nursing, allied health, public health, guideline methods, evidence synthesis, data science, regulation, patient advocacy, policy and implementation science without promising a grade, admission, funding decision or job.

CHAPTER 6 OF 16

6. Preserve Prompts, Context and Configuration

Back to contents

Why this move matters

Report system instructions, templates, examples, context windows, retrieval settings, decoding, tool calls and prompt revisions.

Place AI inside the guideline lifecycle

Report system instructions, templates, examples, context windows, retrieval settings, decoding, tool calls and prompt revisions. Name the stage, task, evidence input, output, model, human role and decision consequence.

Audit one AI-assisted recommendation

Worked example: The certainty-summary template and every revision are archived with reasons. Trace question framing, search, screening, extraction, synthesis, certainty, wording, review and update.

Keep assistance separate from authority

AI may suggest, sort, draft or compare. State who verifies evidence, judges certainty, weighs values and approves recommendations.

Respect the development boundary

Published prompts may not reveal hidden platform instructions. GUIDE-AI is under development. Its official project record describes scope and methods, not a final checklist or endorsement of any tool.

Student, parent and career transfer

Make one evidence card for “Preserve Prompts, Context and Configuration”: 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 medicine, nursing, allied health, public health, guideline methods, evidence synthesis, data science, regulation, patient advocacy, policy and implementation science without promising a grade, admission, funding decision or job.

CHAPTER 7 OF 16

7. Protect Confidential and Licensed Material

Back to contents

Why this move matters

Explain lawful access, personal data, unpublished stakeholder input, retention, training use, security and deletion.

Place AI inside the guideline lifecycle

Explain lawful access, personal data, unpublished stakeholder input, retention, training use, security and deletion. Name the stage, task, evidence input, output, model, human role and decision consequence.

Audit one AI-assisted recommendation

Worked example: Unpublished conflict declarations are excluded from a public model and processed in an approved environment. Trace question framing, search, screening, extraction, synthesis, certainty, wording, review and update.

Keep assistance separate from authority

AI may suggest, sort, draft or compare. State who verifies evidence, judges certainty, weighs values and approves recommendations.

Respect the development boundary

A provider assurance does not replace governance review. GUIDE-AI is under development. Its official project record describes scope and methods, not a final checklist or endorsement of any tool.

Student, parent and career transfer

Make one evidence card for “Protect Confidential and Licensed Material”: 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 medicine, nursing, allied health, public health, guideline methods, evidence synthesis, data science, regulation, patient advocacy, policy and implementation science without promising a grade, admission, funding decision or job.

CHAPTER 8 OF 16

8. Evaluate Error, Bias and Reproducibility

Back to contents

Why this move matters

Define reference standards, metrics, subgroups, hallucinations, omissions, citation accuracy, run variation and external checks.

Place AI inside the guideline lifecycle

Define reference standards, metrics, subgroups, hallucinations, omissions, citation accuracy, run variation and external checks. Name the stage, task, evidence input, output, model, human role and decision consequence.

Audit one AI-assisted recommendation

Worked example: AI evidence summaries are compared with methodologist summaries and errors are classified by consequence. Trace question framing, search, screening, extraction, synthesis, certainty, wording, review and update.

Keep assistance separate from authority

AI may suggest, sort, draft or compare. State who verifies evidence, judges certainty, weighs values and approves recommendations.

Respect the development boundary

Agreement with one reviewer is not ground truth. GUIDE-AI is under development. Its official project record describes scope and methods, not a final checklist or endorsement of any tool.

Student, parent and career transfer

Make one evidence card for “Evaluate Error, Bias and Reproducibility”: 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 medicine, nursing, allied health, public health, guideline methods, evidence synthesis, data science, regulation, patient advocacy, policy and implementation science without promising a grade, admission, funding decision or job.

CHAPTER 9 OF 16

9. Define Human Roles and Decision Rights

Back to contents

Why this move matters

Name who prompts, checks, adjudicates, can override and approves outputs; report expertise, training, workload and access to sources.

Place AI inside the guideline lifecycle

Name who prompts, checks, adjudicates, can override and approves outputs; report expertise, training, workload and access to sources. Name the stage, task, evidence input, output, model, human role and decision consequence.

Audit one AI-assisted recommendation

Worked example: A methodologist verifies citations, while the panel alone judges recommendation direction and strength. Trace question framing, search, screening, extraction, synthesis, certainty, wording, review and update.

Keep assistance separate from authority

AI may suggest, sort, draft or compare. State who verifies evidence, judges certainty, weighs values and approves recommendations.

Respect the development boundary

The phrase human in the loop does not prove meaningful oversight. GUIDE-AI is under development. Its official project record describes scope and methods, not a final checklist or endorsement of any tool.

Student, parent and career transfer

Make one evidence card for “Define Human Roles and Decision Rights”: 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 medicine, nursing, allied health, public health, guideline methods, evidence synthesis, data science, regulation, patient advocacy, policy and implementation science without promising a grade, admission, funding decision or job.

CHAPTER 10 OF 16

10. Report Disagreement and Overrides

Back to contents

Why this move matters

Preserve when AI and humans differ, why an output is accepted or rejected and whether disagreement patterns change the process.

Place AI inside the guideline lifecycle

Preserve when AI and humans differ, why an output is accepted or rejected and whether disagreement patterns change the process. Name the stage, task, evidence input, output, model, human role and decision consequence.

Audit one AI-assisted recommendation

Worked example: The team logs a rejected AI summary that omitted harms and updates the prompt and audit plan. Trace question framing, search, screening, extraction, synthesis, certainty, wording, review and update.

Keep assistance separate from authority

AI may suggest, sort, draft or compare. State who verifies evidence, judges certainty, weighs values and approves recommendations.

Respect the development boundary

Overrides may reflect human error or deference, not correctness. GUIDE-AI is under development. Its official project record describes scope and methods, not a final checklist or endorsement of any tool.

Student, parent and career transfer

Make one evidence card for “Report Disagreement and Overrides”: 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 medicine, nursing, allied health, public health, guideline methods, evidence synthesis, data science, regulation, patient advocacy, policy and implementation science without promising a grade, admission, funding decision or job.

CHAPTER 11 OF 16

11. Keep Certainty Judgements Explainable

Back to contents

Why this move matters

State whether AI extracts inputs, drafts rationales or assigns certainty; show how risk of bias, inconsistency, indirectness, imprecision and publication bias are handled.

Place AI inside the guideline lifecycle

State whether AI extracts inputs, drafts rationales or assigns certainty; show how risk of bias, inconsistency, indirectness, imprecision and publication bias are handled. Name the stage, task, evidence input, output, model, human role and decision consequence.

Audit one AI-assisted recommendation

Worked example: AI drafts a table, but trained reviewers make and justify each certainty judgement. Trace question framing, search, screening, extraction, synthesis, certainty, wording, review and update.

Keep assistance separate from authority

AI may suggest, sort, draft or compare. State who verifies evidence, judges certainty, weighs values and approves recommendations.

Respect the development boundary

A fluent rationale can still misuse a framework. GUIDE-AI is under development. Its official project record describes scope and methods, not a final checklist or endorsement of any tool.

Student, parent and career transfer

Make one evidence card for “Keep Certainty Judgements Explainable”: 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 medicine, nursing, allied health, public health, guideline methods, evidence synthesis, data science, regulation, patient advocacy, policy and implementation science without promising a grade, admission, funding decision or job.

CHAPTER 12 OF 16

12. Keep Values, Equity and Feasibility Visible

Back to contents

Why this move matters

Report how patient values, acceptability, resources, equity, feasibility and local context enter deliberation and whether AI is used.

Place AI inside the guideline lifecycle

Report how patient values, acceptability, resources, equity, feasibility and local context enter deliberation and whether AI is used. Name the stage, task, evidence input, output, model, human role and decision consequence.

Audit one AI-assisted recommendation

Worked example: Youth representatives review AI-generated themes against original consultation comments. Trace question framing, search, screening, extraction, synthesis, certainty, wording, review and update.

Keep assistance separate from authority

AI may suggest, sort, draft or compare. State who verifies evidence, judges certainty, weighs values and approves recommendations.

Respect the development boundary

Summarisation can flatten minority views. GUIDE-AI is under development. Its official project record describes scope and methods, not a final checklist or endorsement of any tool.

Student, parent and career transfer

Make one evidence card for “Keep Values, Equity and Feasibility Visible”: 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 medicine, nursing, allied health, public health, guideline methods, evidence synthesis, data science, regulation, patient advocacy, policy and implementation science without promising a grade, admission, funding decision or job.

CHAPTER 13 OF 16

13. Draft Recommendations Without Delegating Authority

Back to contents

Why this move matters

Describe AI involvement in wording, alternatives, conditions, strength and rationale; identify final panel approval and dissent.

Place AI inside the guideline lifecycle

Describe AI involvement in wording, alternatives, conditions, strength and rationale; identify final panel approval and dissent. Name the stage, task, evidence input, output, model, human role and decision consequence.

Audit one AI-assisted recommendation

Worked example: AI offers plain-language versions, while the panel checks that conditions and certainty remain accurate. Trace question framing, search, screening, extraction, synthesis, certainty, wording, review and update.

Keep assistance separate from authority

AI may suggest, sort, draft or compare. State who verifies evidence, judges certainty, weighs values and approves recommendations.

Respect the development boundary

Clear wording can overstate a weak recommendation. GUIDE-AI is under development. Its official project record describes scope and methods, not a final checklist or endorsement of any tool.

Student, parent and career transfer

Make one evidence card for “Draft Recommendations Without Delegating Authority”: 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 medicine, nursing, allied health, public health, guideline methods, evidence synthesis, data science, regulation, patient advocacy, policy and implementation science without promising a grade, admission, funding decision or job.

CHAPTER 14 OF 16

14. Report Conflicts, Vendors and Independence

Back to contents

Why this move matters

Disclose funding, tool providers, commercial relationships, donated services, model access and influence over design or publication.

Place AI inside the guideline lifecycle

Disclose funding, tool providers, commercial relationships, donated services, model access and influence over design or publication. Name the stage, task, evidence input, output, model, human role and decision consequence.

Audit one AI-assisted recommendation

Worked example: A vendor supplies credits but cannot access deliberations or veto results. Trace question framing, search, screening, extraction, synthesis, certainty, wording, review and update.

Keep assistance separate from authority

AI may suggest, sort, draft or compare. State who verifies evidence, judges certainty, weighs values and approves recommendations.

Respect the development boundary

Disclosure alone does not remove influence. GUIDE-AI is under development. Its official project record describes scope and methods, not a final checklist or endorsement of any tool.

Student, parent and career transfer

Make one evidence card for “Report Conflicts, Vendors and Independence”: 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 medicine, nursing, allied health, public health, guideline methods, evidence synthesis, data science, regulation, patient advocacy, policy and implementation science without promising a grade, admission, funding decision or job.

CHAPTER 15 OF 16

15. Plan Implementation, Monitoring and Updating

Back to contents

Why this move matters

Explain dissemination, decision support, local adaptation, performance monitoring, incident review, model change and evidence surveillance.

Place AI inside the guideline lifecycle

Explain dissemination, decision support, local adaptation, performance monitoring, incident review, model change and evidence surveillance. Name the stage, task, evidence input, output, model, human role and decision consequence.

Audit one AI-assisted recommendation

Worked example: The team monitors citation errors and pauses AI support after a model update until revalidation. Trace question framing, search, screening, extraction, synthesis, certainty, wording, review and update.

Keep assistance separate from authority

AI may suggest, sort, draft or compare. State who verifies evidence, judges certainty, weighs values and approves recommendations.

Respect the development boundary

Monitoring may detect harm only after exposure. GUIDE-AI is under development. Its official project record describes scope and methods, not a final checklist or endorsement of any tool.

Student, parent and career transfer

Make one evidence card for “Plan Implementation, Monitoring and Updating”: 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 medicine, nursing, allied health, public health, guideline methods, evidence synthesis, data science, regulation, patient advocacy, policy and implementation science without promising a grade, admission, funding decision or job.

CHAPTER 16 OF 16

16. Conclude With a Bounded Guideline Claim

Back to contents

Why this move matters

Tie any benefit to task, evidence, model, evaluation, oversight, setting and date; do not promise better recommendations without comparative evidence.

Place AI inside the guideline lifecycle

Tie any benefit to task, evidence, model, evaluation, oversight, setting and date; do not promise better recommendations without comparative evidence. Name the stage, task, evidence input, output, model, human role and decision consequence.

Audit one AI-assisted recommendation

Worked example: The evidence supports supervised drafting assistance in this update, not autonomous guideline production. Trace question framing, search, screening, extraction, synthesis, certainty, wording, review and update.

Keep assistance separate from authority

AI may suggest, sort, draft or compare. State who verifies evidence, judges certainty, weighs values and approves recommendations.

Respect the development boundary

Transparent reporting cannot certify clinical correctness. GUIDE-AI is under development. Its official project record describes scope and methods, not a final checklist or endorsement of any tool.

Student, parent and career transfer

Make one evidence card for “Conclude With a Bounded Guideline Claim”: 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 medicine, nursing, allied health, public health, guideline methods, evidence synthesis, data science, regulation, patient advocacy, policy and implementation science without promising a grade, admission, funding decision or job.

PRACTICE LAB

Build a GUIDE-AI decision trail

Student route

1. Choose one original report inside this guide’s scope.
2. Copy its objective and underline actor, action, context, evidence and intended decision.
3. Draw the specific reasoning chain used in this article.
4. Mark each step reported, inferred, contested or unknown.
5. Rewrite the conclusion so it preserves the narrowest important limitation.

Parent or teacher route

Ask the learner to explain how English helps us show exactly where ai enters a health guideline, what it changes, who checks it and which decisions remain human 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 GUIDE-AI 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.

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 RIGHT health-practice guideline guide for a related but different reader job. Return here when the question is specifically how to show exactly where ai enters a health guideline, what it changes, who checks it and which decisions remain human. 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 GUIDE-AI?

A project developing reporting guidance for AI use across the health-guideline enterprise.

Is it final?

No. EQUATOR lists the project as under development.

Can AI make guideline recommendations?

A study may investigate assistance, but authority, verification and approval must be explicit.

Why report prompts and versions?

They can materially change outputs and reproducibility.

Does human review remove bias?

No. Human roles, expertise, workload, evidence and disagreement still matter.

What belongs in an update plan?

Evidence surveillance, model changes, revalidation, incidents, responsibilities and dates.