Research Prompting

Research the use of generative AI in software testing.
Focus on automated test generation and defect detection.
Use sources published from 2022 onward.
Prefer peer-reviewed papers, official documentation, and industry reports.
Separate confirmed facts from interpretations.
Cite every important factual claim.
Present the result as a structured research report.

Act as a technical research analyst.
Research how large language models are used in code review.
Focus on defect detection, security review, and maintainability.
Include studies and technical reports published from 2021 onward.
Prefer peer-reviewed research, official documentation, and established engineering organisations.
Exclude promotional articles without supporting evidence.
Compare findings from at least five independent sources.
Record evidence for every major claim.
Identify contradictions and explain possible reasons.
Separate verified facts from reasonable inferences.
Include citations with each major factual statement.
Present the result using an executive summary, methodology, findings, limitations, and conclusion.

What is artificial intelligence?

How does generative AI affect software developer productivity in enterprise application development?

How has the use of generative AI coding assistants affected development speed, code quality, and defect rates among enterprise software teams since 2022?

Among Java development teams in large organisations, how does AI-assisted code generation affect delivery time and post-release defect rates compared with traditional development methods?

Act as a research design expert.
Convert the broad topic "AI in software development" into five focused research questions.
Each question must define the technology, target group, expected outcome, and time period.
Avoid questions that can be answered only with yes or no.
Make each question suitable for evidence-based research.
Explain why each question is researchable.

Why are AI coding tools harmful to software developers?

What benefits and risks do AI coding tools create for software developers?

Include research about generative AI coding assistants used by professional software developers.
Cover productivity, code quality, security, and developer learning.
Include sources published between January 2022 and December 2025.
Focus on enterprise software development.
Exclude studies focused only on student assignments.
Exclude articles without data, evidence, or named sources.

Act as a research planner.
Define a practical research scope for the question below.
Research question: How do AI coding assistants affect software developer productivity?
Define included topics.
Define excluded topics.
Define the target population.
Define the time period.
Define acceptable source types.
Define geographic limitations.
Define the expected research depth.
Present the scope as a research boundary table.

AI coding assistants

AI coding assistants developer productivity empirical study
GitHub Copilot code quality research paper
generative AI software defects security study
AI pair programming enterprise development report
coding assistant productivity controlled experiment

"AI coding assistant" AND "developer productivity" AND empirical study

generative AI software development research

AI coding assistant task completion time controlled experiment

site:arxiv.org AI code generation security
site:acm.org AI coding assistant productivity
site:microsoft.com research Copilot developer productivity

AI coding assistants reduce code quality study
limitations of AI coding assistants developer productivity

generative AI code review research 2025

Act as an academic search strategist.
Generate 20 search queries for the research question below.
Research question: How do AI coding assistants affect software developer productivity and code quality?
Create broad discovery queries.
Create focused evidence queries.
Create opposing-view queries.
Create source-specific queries.
Create recent-research queries.
Include important synonyms and alternative technical terms.
Group the queries by purpose.
Do not repeat the same query with minor wording changes.

Act as a source evaluation specialist.
Evaluate each provided source for authority, relevance, evidence quality, recency, transparency, and possible bias.
Assign a score from 1 to 5 for each criterion.
Explain every score briefly.
Classify each source as high confidence, medium confidence, or low confidence.
Reject sources that do not provide enough evidence for factual claims.
Present the result in a comparison table.

Prefer primary sources whenever available.
Use secondary sources for context and interpretation.
Do not rely on a single source for major conclusions.
Do not use anonymous articles for critical claims.
Mark vendor-funded studies clearly.
Prefer recent sources for fast-changing technology.
Keep older sources only when they are historically important.
Record the publication date and source organisation.

Compare the provided research sources.
Identify the question addressed by each source.
Compare their methodologies.
Compare sample sizes and target populations.
Summarise the main findings.
Identify where the findings agree.
Identify where the findings disagree.
Explain whether the disagreements may result from different methods, samples, definitions, or time periods.
Rank the sources by evidence strength.
Do not decide based only on publication reputation.
Present the comparison in a structured table followed by a short analysis.

Extract evidence from the provided sources for the research question below.
Research question: How do AI coding assistants affect developer productivity?
Record one evidence item per row.
Include the exact claim supported by the evidence.
Include the source title.
Include the publication date.
Include the page, section, or paragraph location when available.
Classify the evidence as supporting, opposing, or neutral.
Rate the evidence strength as strong, moderate, or weak.
Record important limitations.
Do not convert author opinions into factual evidence.
Do not add information that is not present in the source.

AI coding assistants always improve developer productivity.

Several studies indicate that AI coding assistants can reduce completion time for selected programming tasks, although the effect varies by task type, developer experience, and evaluation method.

Verify each claim using the provided evidence.
Match every claim with supporting or opposing evidence.
Check whether the source directly supports the full claim.
Identify words such as always, never, all, none, proven, and guaranteed.
Mark claims that are broader than the available evidence.
Classify each claim as verified, partially verified, unsupported, contradicted, or inconclusive.
Rewrite unsupported or overstated claims using accurate language.
Include the evidence IDs used for each decision.
Do not treat repeated claims from dependent sources as independent confirmation.

AI coding assistants reduce task completion time.

AI coding assistants do not improve overall delivery speed.

Analyse the provided sources for contradictions.
Extract claims that discuss the same research issue.
Identify claims that support each other.
Identify claims that appear to conflict.
Determine whether each conflict is direct or only apparent.
Compare definitions, samples, methods, time periods, technologies, and measured outcomes.
Explain the most likely reason for each disagreement.
Do not force a single conclusion when the evidence remains mixed.
Present unresolved contradictions clearly.

Act as an academic research assistant.
Conduct a structured literature review on AI coding assistants and software developer productivity.
Include research published from 2022 onward.
Prefer peer-reviewed papers and original empirical studies.
Exclude articles that only provide opinions or product promotion.
Group the literature by productivity, code quality, security, developer experience, and learning effects.
Compare study designs, sample sizes, tools, programming languages, and evaluation measures.
Identify areas of agreement.
Identify conflicting findings.
Identify common limitations.
Identify unanswered research questions.
Cite every study discussed.
Present the result as a thematic literature review.

Create a systematic review protocol for the research question below.
Define databases to search.
Define search terms.
Define inclusion criteria.
Define exclusion criteria.
Define the screening process.
Define quality assessment criteria.
Define data extraction fields.
Define the synthesis method.
Define how disagreements between reviewers should be handled.
Do not claim that a systematic review was completed unless the full process was actually performed.

Analyse the provided literature and identify research gaps.
Do not label a topic as a gap only because it is interesting.
Connect every proposed gap to evidence from the reviewed studies.
Identify missing populations.
Identify missing environments.
Identify missing outcome measures.
Identify methodological weaknesses.
Identify unresolved contradictions.
Identify areas lacking long-term data.
Explain why each gap matters.
Propose one research question for each gap.
Rank the gaps by importance and research feasibility.

There is not enough research on AI.

Existing studies mainly measure short-term task completion speed in controlled settings. Limited evidence is available on whether the observed productivity gains continue across full enterprise development cycles that include design, review, testing, deployment, and maintenance.

Study A found faster task completion.
Study B found higher developer satisfaction.
Study C found security weaknesses.

Current evidence suggests that AI coding assistants can improve speed and perceived productivity for selected development tasks. However, these benefits do not automatically produce better software quality. Security-focused studies show that generated code may contain unsafe patterns, indicating that productivity gains depend on effective review, testing, and developer judgement.

Synthesize the provided evidence into a unified analysis.
Do not summarise each source separately unless necessary.
Group related findings into themes.
Explain where multiple sources support the same conclusion.
Explain where findings differ.
Give greater weight to stronger and more direct evidence.
Distinguish measured findings from author interpretations.
Include important limitations.
State only conclusions supported by the combined evidence.
Use cautious language when evidence is mixed.
Cite the supporting sources for every major conclusion.

Use citations for every major factual claim.
Place each citation immediately after the claim it supports.
Prefer the original research source instead of an article summarising it.
Include the author, title, publication, year, and persistent identifier when available.
Include page or section numbers for source-specific evidence.
Use APA 7 citation style.
Do not invent missing publication details.
Mark incomplete citation fields as unavailable.
Confirm that every reference is used in the report.
Confirm that every in-text citation appears in the reference list.

AI coding assistants reduced development time by 50%.

Audit the citations in the report.
Check whether each citation exists.
Check whether each cited source supports the nearby claim.
Identify claims that require citations but do not have them.
Identify citations placed after claims they do not support.
Identify references that are listed but never cited.
Identify in-text citations missing from the reference list.
Do not correct missing details by guessing.
Present all citation problems in a validation table.

The study included 120 software developers.

The results may be more applicable to professional developers than to beginners.

AI coding assistants are the best development tools ever created.

AI coding assistants may become standard in most enterprise development teams.

Analyse the research content below.
Label every important statement as fact, inference, opinion, prediction, or unsupported claim.
Link each fact to its supporting source.
Explain the evidence used for each inference.
Use phrases such as "the evidence suggests" for reasonable inferences.
Use phrases such as "it is possible" for uncertain interpretations.
Do not present predictions as established facts.
Rewrite unsupported statements using accurate uncertainty language.

Act as a professional research analyst.
Generate a research report using only the provided evidence.
Use the research question as the main focus.
Begin with an executive summary.
Explain the research scope and methodology.
Group findings by theme.
Compare supporting and opposing evidence.
Identify contradictions and unresolved questions.
Separate facts from inferences.
Include a confidence level for each major conclusion.
Discuss limitations.
Provide practical recommendations only when supported by evidence.
Cite every major factual claim.
Include a complete reference list.
Do not add unsupported facts.
Do not hide uncertainty.

Review the research report for quality.
Check whether the report answers the research question.
Check whether the scope was followed.
Check whether every major claim has evidence.
Check whether citations support the claims.
Check whether facts and inferences are separated.
Check whether contradictions are presented fairly.
Check whether limitations are complete.
Check whether conclusions are stronger than the evidence.
Check whether recommendations follow from the findings.
List all required corrections before producing the revised report.

Act as a professional research analyst.
Research the following question: [RESEARCH QUESTION]
Research objective: [OBJECTIVE]
Include: [INCLUDED AREAS]
Exclude: [EXCLUDED AREAS]
Time period: [TIME PERIOD]
Geographic scope: [GEOGRAPHIC SCOPE]
Target population: [POPULATION]
Prefer these source types: [SOURCE TYPES]
Reject sources that: [REJECTION RULES]
Generate search queries before collecting evidence.
Compare findings from at least [NUMBER] independent sources.
Record evidence for every major claim.
Search for both supporting and opposing evidence.
Identify contradictions.
Separate facts from inferences.
State uncertainty clearly.
Use [CITATION STYLE] citations.
Present the result as [OUTPUT FORMAT].
Include methodology, findings, limitations, conclusion, and references.

Act as a research design expert.
Broad topic: [TOPIC]
Target population: [POPULATION]
Main problem: [PROBLEM]
Expected outcome: [OUTCOME]
Time period: [TIME PERIOD]
Generate five focused research questions.
Make every question specific and researchable.
Avoid biased wording.
Avoid yes-or-no questions.
Explain the variables contained in each question.
Recommend the strongest question and explain why.

Define the research scope for this question: [RESEARCH QUESTION]
Specify included topics.
Specify excluded topics.
Specify the target population.
Specify the environment.
Specify the time period.
Specify the geographic area.
Specify acceptable source types.
Specify the minimum evidence quality.
Specify the expected depth.
Present the result as a scope table.
Identify possible scope risks.

Act as a research search strategist.
Research question: [RESEARCH QUESTION]
Generate broad discovery queries.
Generate focused evidence queries.
Generate primary-source queries.
Generate opposing-evidence queries.
Generate recent-research queries.
Generate source-specific queries.
Include synonyms and alternative technical terms.
Group queries by purpose.
Avoid duplicate queries.
Explain what each query is expected to find.

Evaluate the following source: [SOURCE]
Identify the author or organisation.
Identify the publication date.
Identify the source type.
Evaluate authority.
Evaluate relevance.
Evaluate evidence quality.
Evaluate methodology.
Evaluate transparency.
Evaluate possible bias.
Identify funding or commercial interests.
Assign an overall confidence level.
Decide whether to include, include cautiously, or exclude.
Explain the decision using evidence.

Extract evidence from the provided content.
Research question: [RESEARCH QUESTION]
Extract only information relevant to the question.
Record one evidence item per row.
Include the supported claim.
Include the exact source location.
Include the source date.
Classify the evidence as supporting, opposing, or neutral.
Rate evidence strength.
Record limitations.
Do not add interpretations to the evidence field.
Mark missing information as unavailable.

Verify the following claim: [CLAIM]
Define the exact meaning of the claim.
Identify the evidence required to verify it.
Find supporting evidence.
Find opposing evidence.
Evaluate source quality.
Check whether the evidence supports the full claim.
Identify exaggerated or absolute wording.
Classify the claim as verified, partially verified, unsupported, contradicted, or inconclusive.
Rewrite the claim using accurate language.
Provide citations for the decision.

Compare the following sources: [SOURCE LIST]
Compare research objectives.
Compare methodologies.
Compare populations.
Compare sample sizes.
Compare definitions.
Compare time periods.
Compare measured outcomes.
Compare findings.
Compare limitations.
Identify agreements.
Identify contradictions.
Explain possible reasons for differences.
Rank sources by evidence strength.
Present the result in a comparison table.

Identify contradictions across the provided sources.
Match claims discussing the same issue.
Identify direct contradictions.
Identify apparent contradictions.
Compare definitions, methods, populations, time periods, and outcomes.
Explain possible causes of disagreement.
Identify which source has stronger evidence.
Keep unresolved contradictions open.
Do not remove minority findings without explanation.
Present the result in a contradiction matrix.

Conduct a literature review on: [TOPIC]
Main research question: [RESEARCH QUESTION]
Include studies from: [TIME PERIOD]
Include these source types: [SOURCE TYPES]
Apply these inclusion criteria: [INCLUSION CRITERIA]
Apply these exclusion criteria: [EXCLUSION CRITERIA]
Group studies by theme.
Compare research methods.
Compare findings.
Identify agreements.
Identify contradictions.
Identify methodological limitations.
Identify research gaps.
Cite every study.
Present the result as a structured thematic literature review.

Analyse the existing research on: [TOPIC]
Identify population gaps.
Identify geographic gaps.
Identify methodological gaps.
Identify measurement gaps.
Identify technology gaps.
Identify context gaps.
Identify long-term evidence gaps.
Identify unresolved contradictions.
Connect every gap to evidence from the literature.
Explain why each gap matters.
Create a research question for each gap.
Rank gaps by importance and feasibility.

Synthesize the provided research evidence.
Main research question: [RESEARCH QUESTION]
Group findings by theme.
Combine findings that support the same conclusion.
Compare opposing findings.
Weight evidence by quality and relevance.
Separate measured results from author interpretations.
Identify limitations.
State confidence levels.
Avoid source-by-source summaries unless required.
Produce conclusions supported by the combined evidence.
Cite every major synthesis statement.

Audit all citations in the provided research report.
Identify claims requiring citations.
Check whether every major factual claim has a citation.
Check whether each citation supports the nearby claim.
Check whether source details are complete.
Check whether citations use one consistent style.
Check whether every in-text citation appears in the reference list.
Check whether every listed reference is used.
Identify possible invented or unverifiable citations.
Do not fill missing details by guessing.
Present the issues and corrections in a table.

Review the provided research content.
Label each important statement as fact, inference, opinion, prediction, or unsupported claim.
Connect each fact to evidence.
Explain the reasoning behind each inference.
Rewrite inferences using cautious language.
Rewrite predictions as possibilities.
Remove or correct unsupported claims.
Preserve the original meaning where evidence allows.
Present the result in a classification table.

Act as a senior research analyst.
Research question: [RESEARCH QUESTION]
Objective: [OBJECTIVE]
Scope: [SCOPE]
Time period: [TIME PERIOD]
Target audience: [AUDIENCE]
Preferred sources: [SOURCE TYPES]
Excluded sources: [EXCLUDED SOURCES]
Required citation style: [CITATION STYLE]
Use only verifiable evidence.
Compare multiple independent sources.
Search for supporting and opposing findings.
Separate facts from inferences.
Identify contradictions.
Identify research gaps.
Assign confidence levels to major conclusions.
Generate an executive summary.
Explain the methodology.
Present findings by theme.
Include evidence tables.
Include limitations.
Include practical recommendations supported by evidence.
Include a conclusion.
Include a complete reference list.
Do not invent facts, evidence, quotations, or citations.

Explain the research problem in one clear paragraph.
Identify the main decision this research should support.
Identify the target audience.
Identify the information required to make that decision.

Convert the research problem into one primary research question and three supporting questions.
Make all questions neutral, focused, and researchable.

Define the research scope.
Include clear boundaries and exclusions.
Identify any scope decisions that may affect the conclusion.

Generate search queries for discovery, verification, opposing evidence, and primary sources.

Evaluate all candidate sources.
Include only sources that meet the defined quality requirements.

Build an evidence table containing supporting, opposing, and neutral evidence.

Verify every major claim using the evidence table.
Rewrite claims that are stronger than the evidence.

Identify contradictions and determine whether they result from different methods, samples, definitions, or time periods.

Produce a thematic synthesis that combines the strongest available evidence.

Generate the final report using only verified evidence.
Include confidence levels and unresolved questions.

Perform an independent quality review of the report.
Correct every unsupported claim, citation problem, contradiction, and scope violation.

How do AI coding assistants affect development speed, code quality, and security among professional software developers?

AI coding assistant developer productivity empirical study
generative AI code quality controlled experiment
AI generated code security vulnerability research
GitHub Copilot professional developers productivity
AI coding assistant code review effort
limitations of AI pair programming tools
enterprise AI coding assistant case study

Act as a senior software engineering research analyst.
Investigate how AI coding assistants affect development speed, code quality, security, and code review effort among professional software developers.
Include evidence published from 2022 onward.
Prefer peer-reviewed studies, original experiments, repository analyses, and technical reports from established research organisations.
Exclude unsupported promotional claims.
Generate a search strategy before analysing findings.
Compare at least five independent sources.
Record the methodology, sample, measured outcomes, findings, and limitations of every major study.
Search for evidence supporting and challenging productivity benefits.
Identify whether reported productivity gains are measured or self-reported.
Distinguish task completion speed from complete project delivery speed.
Detect contradictions caused by different tools, tasks, languages, developer experience levels, or evaluation methods.
Separate facts from inferences.
Assign confidence levels to major conclusions.
Cite every factual and numerical claim.
Present the result using an executive summary, scope, methodology, evidence table, thematic findings, contradictions, limitations, research gaps, conclusion, and references.

Research cloud computing.

Research how serverless computing affects cost and scalability for event-driven web applications.

Find information about AI security.

Use peer-reviewed security research, official vulnerability databases, standards, and technical reports from recognised security organisations.

Find evidence that AI improves productivity.

Find evidence supporting and challenging the claim that AI improves productivity.

Act as a professional research analyst with expertise in evidence evaluation.
Research the following question: [RESEARCH QUESTION]
The purpose of the research is: [PURPOSE]
The target audience is: [AUDIENCE]
Include the following topics: [INCLUDED TOPICS]
Exclude the following topics: [EXCLUDED TOPICS]
Use the following time period: [TIME PERIOD]
Use the following geographic scope: [GEOGRAPHIC SCOPE]
Focus on the following population or environment: [POPULATION OR ENVIRONMENT]
Prefer primary sources, peer-reviewed studies, official records, standards, original datasets, and established technical reports.
Exclude anonymous, unsupported, highly promotional, or unverifiable sources.
Generate broad, focused, source-specific, recent, and opposing-evidence search queries.
Evaluate every source for authority, relevance, evidence quality, recency, transparency, and possible bias.
Record every major evidence item in a structured evidence table.
Include supporting, opposing, and neutral evidence.
Verify every major claim against its supporting evidence.
Identify claims that are exaggerated, incomplete, outdated, unsupported, contradicted, or inconclusive.
Compare source methodologies, populations, sample sizes, definitions, time periods, technologies, outcomes, and limitations.
Detect direct and apparent contradictions.
Explain possible reasons for conflicting findings.
Conduct a thematic synthesis instead of listing source summaries.
Give greater weight to stronger, direct, relevant, and independent evidence.
Identify population, method, technology, measurement, context, geographic, contradiction, and long-term research gaps.
Separate facts, inferences, opinions, and predictions.
Use cautious wording when evidence is limited.
Cite every important factual and numerical claim.
Place citations immediately after the supported claim.
Use [CITATION STYLE] consistently.
Do not invent authors, titles, quotations, publication details, identifiers, evidence, or citations.
Mark missing information as unavailable.
Present the final output using a title, executive summary, research question, objectives, scope, methodology, source criteria, evidence table, thematic findings, source comparison, contradictions, research gaps, limitations, conclusions, recommendations, confidence levels, and references.
Perform a final verification pass before presenting the report.
Confirm that every conclusion follows from the collected evidence.
Clearly state what remains unknown or inconclusive.