Research Skills Explained: How Good Questions Become Reliable Answers

Research is often mistaken for searching.

A student is given a topic, opens a search engine, clicks several links, copies useful information into notes and begins writing. The process may produce a paper with citations, but citations alone do not make it good research.

Serious research begins earlier.

It begins when someone recognises that they do not yet know enough to answer a question responsibly.

The researcher then has to decide what is actually being asked, what kind of evidence could answer it, where that evidence is likely to exist, how reliable each source is, how competing findings relate to one another and how certain the final conclusion deserves to be.

That makes research a process of disciplined uncertainty.

The Association of College and Research Libraries captures this idea in its influential information-literacy framework through the principle “Research as Inquiry.” The Framework describes research as iterative: questions generate answers, answers generate new questions, and information discovered during the investigation may expose weaknesses in the original question itself.

Research is therefore not simply knowing where to look.

It is knowing how to move from a question to a conclusion without allowing the conclusion to become stronger than the evidence supporting it.

A Topic Is Not Yet a Research Question

“Artificial intelligence” is a topic.

“Soil pollution” is a topic.

“Remote work” is a topic.

Each is far too broad to tell a researcher what evidence to collect.

A research question creates boundaries. Instead of asking broadly about artificial intelligence, a researcher might ask: How has generative AI changed entry-level writing tasks in Indian marketing agencies since 2023?

That question immediately makes several decisions visible. It identifies the technology, the type of work, the occupational context, the geography and the period being studied.

Compare that with the phrase “AI and jobs.” Almost anything could qualify as relevant, which means the researcher has no defensible principle for deciding what should be included or excluded.

Good research questions do not merely sound academic. They control the investigation.

A Good Research Question Defines What the Answer Must Explain

Suppose someone asks, “Is social media bad for teenagers?”

The question appears simple, but several important terms remain undefined. What counts as social media? What does “bad” mean? Mental health, sleep, academic achievement, self-esteem or physical activity? Which teenagers? How much use? Active communication or passive scrolling? Over what period?

A more useful question might be: Among adolescents aged 13–17, what relationship has research found between passive social-media use and depressive symptoms?

The narrower question does not solve the research problem automatically. It makes the problem answerable.

This is one of the central disciplines of research: converting a broad area of interest into a question specific enough that evidence can actually support or challenge an answer.

Research Questions Often Improve During Research

The first question does not have to be perfect.

In fact, a question that survives the entire research process unchanged may sometimes indicate that the researcher did not allow the evidence to challenge their assumptions.

Suppose someone begins with: “Why does remote work reduce productivity?”

That wording already assumes that remote work reduces productivity.

After examining research, the investigator may discover mixed outcomes. Productivity appears to differ according to occupation, management practices, employee autonomy, task type, collaboration requirements and how productivity itself is measured.

The better question may therefore become: “Under what conditions is remote or hybrid work associated with higher or lower productivity?”

The research has not failed because the original question changed.

The process has worked.

Different Questions Require Different Evidence

A common research mistake is collecting whatever information is easiest to find rather than asking what evidence the question actually requires.

If the question concerns current law, a blog post discussing the law is not equivalent to the statute, regulation or judgment itself.

If the question concerns whether a medical intervention is effective, testimonials are not equivalent to controlled clinical evidence.

If the question concerns what happened during a historical event, archival documents, contemporary records and material evidence may carry special importance.

If the question concerns current consumer preferences, decades-old academic literature may be less relevant than a well-designed recent survey or sales dataset.

Research skill therefore begins with a match between question type and evidence type.

A descriptive question asks what exists or what is happening. A comparative question asks how two or more cases differ. A causal question asks whether one factor contributes to another outcome. An evaluative question asks how well a policy, programme or intervention performs against defined criteria.

Those questions cannot always be answered with the same methodology.

Description Is Not the Same as Causation

Suppose a survey finds that students who sleep less tend to report lower examination scores.

That may establish an association.

It does not by itself establish that reduced sleep caused the lower scores.

Students under greater stress might both sleep less and study differently. Illness could affect sleep and academic performance. Social circumstances might influence both variables.

A descriptive or correlational dataset can be extremely useful without being capable of proving every causal claim a researcher wants to make.

Good research respects the limits of the method.

The conclusion must match what the evidence was capable of establishing.

Research Begins With Orientation

When entering an unfamiliar field, the first task is often not to find the final answer.

It is to learn how the field talks.

Someone interested in “why employees stop caring about work” may discover that researchers use terms such as employee disengagement, organisational commitment, burnout, job demands-resources, job satisfaction and psychological withdrawal.

Those concepts open better search pathways.

The first few useful papers may therefore matter less for their conclusions than for the vocabulary they provide.

This is why research searching is iterative. Each useful source teaches the researcher something about how to search for the next one.

Search Is a Strategy, Not a Single Query

Beginners often imagine that expert researchers simply know the perfect words to type into a search box.

In practice, strong searching involves repeated adjustment.

A researcher may begin with ordinary-language terms, discover specialist terminology, search again using that vocabulary, identify an influential paper, examine its references, find later studies citing it and move into a specialist database when a general search engine no longer provides sufficient precision.

ACRL describes this process as “Searching as Strategic Exploration.” Its Framework emphasises that searching is nonlinear and iterative, and that effective researchers adjust their vocabulary, search tools and strategies as their understanding develops.

The search is therefore part of the reasoning process.

It is not merely a gateway that occurs before the “real” research begins.

Search Terms Affect What You Can Find

Imagine investigating whether flexible working arrangements affect employee retention.

Searching only “work from home employee retention” may produce useful material, but it may miss research using terms such as telework, remote work, hybrid work, flexible work arrangements, turnover intention, employee attrition or organisational retention.

Researchers therefore need to think conceptually.

What are the central ideas in the question? What synonyms might different disciplines use? What older terminology might appear in historical studies? What spelling differences exist across countries?

Search language becomes more important as the research problem becomes more specialised.

Search Engines and Academic Databases Do Different Jobs

Google can be extraordinarily useful for research.

So can Google Scholar, PubMed, JSTOR, Scopus, Web of Science, government databases, legal databases, statistical portals, archives and specialist library catalogues.

None is universally best.

A general search engine may be ideal for locating a government report or official organisation quickly. A biomedical database may provide much better control when searching clinical literature. An archive may contain primary records that never appear prominently in ordinary web search.

The researcher's task is to choose a search environment appropriate to the question.

That is another reason “I searched Google and found nothing” is often a weak research conclusion.

Sometimes the problem is not absence of information.

It is the choice of search system.

Search Results Are Not the Evidence Base

The first page of a search engine is a ranked selection.

It is not a systematic literature review.

Search systems rank material according to their own retrieval and relevance mechanisms. Academic databases also differ in their coverage.

A source appearing first does not automatically make it the strongest source.

Likewise, a paper that is difficult to find is not necessarily less important.

Researchers therefore need to move from search results to source evaluation.

Finding information and deciding whether it deserves evidential weight are separate skills.

Primary, Secondary and Tertiary Sources Serve Different Purposes

The familiar distinction between primary and secondary sources can be useful, provided it is not turned into a simplistic hierarchy.

A primary source provides direct evidence relevant to the research question. Depending on the field, this could include an original scientific study, court judgment, statute, survey dataset, archival letter, financial filing, interview or government statistical release.

A secondary source analyses or interprets other evidence. Examples can include scholarly reviews, specialist journalism, textbooks or historical analysis.

A tertiary source synthesises or organises information at a broader level, such as an encyclopedia or reference guide.

Primary does not automatically mean better.

An individual clinical trial may be primary evidence, while a rigorous systematic review of dozens of trials may provide a stronger basis for understanding the total evidence.

The correct question is not “Is this primary?”

It is “What role should this source play in answering my specific question?”

Authority Is Contextual

Prestige alone cannot determine source quality.

A world-famous physicist is not automatically an authority on constitutional law.

A government agency may be the authoritative source for an official unemployment statistic because it produced the dataset. That does not make every political interpretation of the statistic equally authoritative.

A company is the best source for the official price and specifications of its own product. It is not automatically the best independent source for deciding whether the product is superior to competitors.

ACRL describes this principle as “Authority Is Constructed and Contextual.” Authority depends partly on the information need and the expertise relevant to the claim being evaluated.

That principle helps researchers move beyond the crude rule that prestigious sources should always be trusted and obscure sources should always be rejected.

Source Evaluation Should Begin Before Writing

A weak workflow collects everything first and evaluates credibility later.

By then the draft may already depend on poor evidence.

A stronger workflow evaluates sources while they are being collected.

Who created the information? What method produced it? What population or period does it describe? Is it current enough for the question? What limitations did the authors identify? Does the source have a commercial or political interest? Is the claim supported directly by evidence, or is one source simply repeating another?

These questions determine how much weight the source deserves.

They should not be postponed until the bibliography is formatted.

Relevance and Reliability Are Different

A source can be highly reliable but irrelevant.

A prestigious study of American university students may be excellent research but still provide weak evidence for a question specifically about rural secondary-school students in India.

Likewise, a source can be directly relevant but methodologically weak.

Researchers need both dimensions.

Does this source address my question?

And:

How much confidence should I place in what it says?

The best-known source is not necessarily the most useful source.

Freshness Depends on the Question

Newer is not always better.

For a question about today's inflation rate, a report from ten years ago is outdated.

For a question about the original formulation of a philosophical argument, a centuries-old primary text may be indispensable.

A 1970s study can remain historically important even if modern research has revised its conclusions.

Researchers therefore should not apply a universal “use sources from the last five years” rule.

Freshness must be interpreted according to the subject.

Current data need current sources.

Historical questions need historically appropriate evidence.

Foundational theories may require older material.

Research Requires a Trail

One of the easiest ways to weaken research is to lose track of where information came from.

A researcher reads ten pages, copies a statistic into notes and assumes they will remember the source later.

Weeks later, the number remains but its provenance has disappeared.

A good research trail records enough information that the evidence can be reconstructed.

For each useful source, preserve the citation, link or identifier, the claim it supports, relevant page numbers or sections, important limitations and notes about how it relates to other evidence.

This may feel slower initially.

It saves enormous time when writing and fact-checking.

Notes Should Separate the Source From the Researcher

Poor note-taking creates two risks simultaneously: accidental plagiarism and confused reasoning.

Suppose a researcher writes a sentence in their notes without marking whether it is a quotation, paraphrase or personal interpretation.

Months later, the distinction may be impossible to reconstruct.

Good notes make that boundary visible.

The researcher should be able to tell whether a statement is the source's exact wording, a paraphrase of the source, their own inference or a question that still needs investigation.

This discipline matters because research writing constantly moves between what the evidence says and what the researcher concludes from it.

Those are not the same thing.

Citation Is Part of the Evidence Chain

Citations are sometimes treated as formatting imposed by universities.

Their deeper purpose is verification.

A citation allows another person to inspect the source and ask whether it really supports the statement for which it was used.

That creates accountability.

A good citation therefore does more than prove that the writer read something.

It connects a claim to the evidence behind it.

A paper full of citations can still be poorly researched if the sources do not actually support the claims attached to them.

Citation quantity is not evidence quality.

Ten Sources Are Not Automatically Ten Independent Pieces of Evidence

This distinction becomes especially important in journalism, policy research and online research.

Suppose ten news articles report the same statistic.

At first glance, that looks like ten sources confirming the fact.

But if all ten stories cite one government press release, there is really one underlying source being repeated ten times.

Research requires source lineage.

Where did the claim originate? Are the sources independently verifying it, or are they copying one another?

Repeated publication can create an illusion of corroboration.

Independence matters.

Citation Chaining Can Reveal the Structure of a Field

Once a researcher finds a particularly useful paper, the reference list becomes another search tool.

Looking backward through its citations can reveal foundational studies and earlier debates.

Looking forward to research that later cited the paper can show how the field developed, whether the result was replicated and whether later evidence challenged it.

This technique is often called citation chaining.

It is valuable because good research rarely exists as isolated papers.

Scholarship is a conversation in which studies respond to one another.

Understanding that conversation helps a researcher see which findings became influential and which were later revised.

Synthesis Is More Than Summarising Sources

A literature review that says, “Study A found this. Study B found that. Study C found something else,” has collected information.

It has not necessarily synthesised it.

Synthesis asks how those findings relate.

Do the studies agree? If not, why might they differ? Did they examine different populations? Were the definitions different? Did one use a stronger research design? Was one study much larger? Did newer evidence use improved methods?

The researcher is not simply reporting individual trees.

They are describing the shape of the forest.

This is where research becomes more than information retrieval.

Contradictory Evidence Is Useful

Researchers are human.

Once an initial explanation begins to look convincing, evidence supporting it becomes satisfying.

Contradictory evidence becomes inconvenient.

That creates confirmation risk.

Good research deliberately searches for serious disagreement. A researcher might look for systematic reviews, critiques, replication studies, null results or evidence from populations where the expected effect does not appear.

The objective is not artificial balance.

One weak dissenting opinion should not be treated as equal to an overwhelming evidence base.

The goal is to determine whether the apparent conclusion survives contact with the strongest reasonable counterevidence.

Consensus Should Be Evaluated, Not Counted Mechanically

Suppose 20 papers support a conclusion and five do not.

It may be tempting to declare a four-to-one victory.

Research quality is not an election.

Perhaps the 20 supporting papers are small observational studies while the five contrary studies are substantially stronger. Or perhaps the opposite is true.

Evidence needs weighting according to method, relevance, sample, bias, precision and independence.

The number of papers matters less than the quality and structure of the total evidence.

Systematic Reviews Solve a Different Problem From Individual Studies

One study answers a specific research question under specific conditions.

A systematic review attempts to identify and evaluate a broader body of evidence using an explicit search and selection process.

A meta-analysis may go further by statistically combining compatible results.

These approaches can provide a stronger overview than selecting several individual studies informally.

But systematic reviews are not infallible.

Their conclusions depend on the quality of included studies, search coverage, eligibility criteria and analytical choices.

Research skill means understanding what each evidence type can and cannot provide.

Missing Evidence Requires Careful Interpretation

Not finding evidence does not automatically prove that something never happened.

The search might be incomplete.

Records may never have been created.

Archives may preserve some groups better than others.

Research with negative results may be less likely to be published.

At the same time, absence can become informative when evidence should reasonably exist.

If a company claims that a programme produced extraordinary results across thousands of participants but cannot provide any underlying data, the missing evidence matters.

The right question is therefore:

If this claim were true, what evidence would I reasonably expect to find?

That turns absence into something that can be analysed rather than merely assumed.

Research Scope Determines Research Quality

A research project can fail before the first search because the question is too large for the available resources.

“Why do democracies fail?” could support decades of scholarship.

A student with three weeks cannot responsibly settle it.

Narrowing the question is not intellectual cowardice.

It is what allows evidence to match the conclusion.

A smaller project asking how one institutional mechanism operated in a particular group of countries during a defined period may produce a far more defensible answer.

Good scope asks what can actually be investigated with the available time, data and expertise.

Research Must Distinguish What Is Known From What Is Inferred

Suppose three studies show an association between workplace flexibility and employee retention.

A researcher concludes that flexibility causes employees to remain with organisations.

That additional word may exceed the evidence.

Research writing needs to distinguish the observation from the interpretation.

Phrases such as “was associated with,” “evidence suggests,” “is consistent with,” “under these conditions” and “the available studies do not establish causation” are useful when they accurately reflect the evidence.

This is not weak writing.

It is calibrated writing.

Uncertainty Is a Research Result

People often imagine that successful research eliminates uncertainty.

Sometimes it does.

Often it reorganises uncertainty.

A researcher may conclude that one explanation is strongly supported, another remains plausible and a third is inconsistent with the available evidence.

That is useful knowledge.

A paper that ends with “the evidence is mixed because outcomes differ across populations and methods” can be much more valuable than one that manufactures a clear answer where the literature does not support one.

The responsibility of the researcher is not to make every question simple.

It is to describe the state of knowledge accurately.

Research Ethics Begins Before Citation

Ethical research includes giving proper credit, but the issue is broader than plagiarism.

Researchers make choices about what evidence to include, how participants are treated, how data are stored, how findings are framed and whether limitations are disclosed.

A technically accurate quotation can still be unethical if it is removed from context in a way that reverses the author's meaning.

A statistic can be real while being presented selectively to create a misleading impression.

Ethical research therefore includes fidelity to the evidence.

The objective is not merely to avoid copying another person's words.

It is to avoid making evidence appear to say something it does not say.

AI Changes Research, but Not the Standard of Evidence

Generative AI can accelerate parts of research.

It can help generate possible search terms, explain unfamiliar terminology, suggest ways to narrow a question, summarise supplied documents or help organise notes.

Those capabilities can be useful.

But an AI answer is not automatically a source.

UNESCO's guidance on generative AI in education and research warns that GenAI outputs should be treated critically and describes the technology as fast but potentially unreliable as a source of information. UNESCO specifically recommends that learners and researchers critique generated responses rather than treating them as authoritative.

The central research rule therefore remains unchanged:

important claims need traceable evidence.

AI Can Help Generate Leads Without Becoming the Evidence

Suppose an AI system tells you that a particular study exists.

That can be a useful lead.

The next step is to find the actual study.

Confirm the title, authors, publication, date and findings. Read enough of the source to determine whether it supports the claim being made.

Do not cite a paper merely because an AI system described it.

Generative systems can produce incorrect references, combine details from different publications or summarise a real study inaccurately.

Research requires returning to the evidence itself.

AI Summaries Can Hide Important Qualifications

A 50-page report may contain limitations that disappear in a five-sentence summary.

Perhaps the study examined only one country.

Perhaps the sample was small.

Perhaps the authors found an association rather than causation.

Perhaps the overall result was statistically uncertain.

A generated summary can be useful for orientation, but researchers should inspect original material before relying on consequential claims.

This principle applies to human-written summaries too.

Compression always creates the possibility that context will disappear.

Research Skills Matter More When Search Becomes Easier

Modern tools have made information retrieval extraordinarily easy.

That does not automatically make research easier.

When millions of results can be produced almost instantly, the limiting skill shifts from finding something to deciding what deserves to be used.

A 2025 study examining first-year university students and faculty found a notable gap between faculty expectations and students' research preparation. Students expressed greater confidence in mechanical tasks such as finding full text than in more analytical research skills.

That distinction matters.

Access to information is not the same as information literacy.

Research Guides and Checklists Have Limits

Universities frequently provide research guides explaining databases, citation systems and evaluation methods.

These tools can be helpful, but they do not automatically produce sophisticated researchers.

A 2025 scoping review examined 1,724 publications about academic library research guides and included 61 studies that met its criteria. The authors concluded that the evidence base for the guides' effectiveness in developing information-literacy skills remained limited and was often exploratory or correlational.

The lesson is not that research guides are useless.

It is that research skill develops through practice, feedback and repeated application, not simply by reading instructions once.

Research Is Learned by Doing Research

A student can memorise that sources should be evaluated for relevance, authority and evidence.

The difficult part arrives when two credible sources disagree.

Which one should receive more weight?

That requires judgment.

Likewise, someone can learn Boolean operators in five minutes but still struggle to identify the concepts required for a sophisticated search.

Research is therefore procedural.

People improve by asking questions, searching badly, discovering why the search failed, revising terminology, comparing evidence, receiving feedback and trying again.

The mistakes become part of the education when the process makes them visible.

A Practical Research Process Begins With the Question

A strong project begins by writing the question in one or two clear sentences.

Then ask what would count as a satisfactory answer.

What time period matters? Which population or geography? What type of evidence could realistically establish the claim?

The next stage is orientation. Learn the terminology of the field, identify major organisations and discover which databases or publication types experts use.

Then search broadly enough to understand the landscape before deliberately narrowing the evidence base.

Evaluate sources as they appear, record them carefully and compare them rather than collecting them passively.

As the evidence develops, revisit the original question.

If it contains an unsupported assumption, revise it.

The final stage is synthesis: answer the question using the evidence as a whole and state clearly what remains uncertain.

That sequence is better understood as a loop than a straight line.

Researchers frequently move backward.

Good Research Knows When to Stop Searching

Searching forever is not rigor.

At some point, additional sources stop materially changing the answer.

Researchers need a stopping rule appropriate to the project.

A systematic review may have a formal protocol defining databases, search dates and inclusion criteria.

A shorter explanatory article may stop when recent authoritative evidence, important primary material and credible competing interpretations have been adequately represented.

The important point is that stopping should not occur merely because the researcher found evidence they liked.

It should occur when the evidence base is sufficient for the claim being made.

Good Research Also Knows When Not to Conclude

Sometimes the responsible result is:

We do not know yet.

Perhaps the studies are too small.

Perhaps available data are too old.

Perhaps measurements are inconsistent.

Perhaps the event is too recent for reliable evidence to exist.

The pressure to produce a definitive answer can tempt writers into pretending uncertainty has disappeared.

Research quality often becomes visible precisely at this point.

A good researcher knows the difference between a question that has been answered and one that merely has information surrounding it.

Common Research Mistakes

One common mistake is beginning with the conclusion and searching only for evidence that supports it. Another is treating the first page of search results as a representative evidence base. Researchers may also confuse repeated reporting with independent confirmation, cite a prestigious source that does not actually address the question or use language of causation when the underlying evidence shows only association.

Another frequent mistake is losing the source trail during note-taking. A fact is copied without its page number, a quotation becomes mixed with a paraphrase, or a statistic circulates through several secondary sources until no one checks the original dataset.

AI introduces a newer version of the same problem: a generated answer may look well researched because it is fluent and includes references, yet those references still need independent verification.

These errors differ on the surface.

They share one underlying problem: the path between evidence and conclusion has become unclear.

How to Know Whether Your Research Is Strong

A strong research project should survive several questions.

Can you state exactly what you were trying to find out? Can you explain why the evidence you used was appropriate to that question? Can another person trace your important claims back to their sources? Did you investigate meaningful contradictory evidence? Do you understand the limitations of the methods you relied on? Did your research question change when the evidence required it?

Most importantly, could you explain what evidence would cause you to change your conclusion?

If the answer is nothing, the process has probably stopped being research.

A conclusion protected from all possible revision is a belief being defended, not a question being investigated.

Research Is Not a Competition to Sound Certain

Academic and public writing often reward confident language.

Evidence does not always deserve it.

One study should not become “research proves.” A small sample should not become “people generally.” A correlation should not automatically become “causes.”

A disciplined researcher calibrates the wording of the conclusion.

If evidence is strong, say so.

If evidence is mixed, say that.

If an important limitation restricts generalisation, explain it.

Precision about uncertainty strengthens credibility because it allows readers to see exactly what the evidence can bear.

The Best Researchers Are Revisable

Strong researchers are not people who begin every project without assumptions.

That would be impossible.

They are people whose method gives those assumptions a chance to fail.

A source contradicts the original belief. A dataset produces an unexpected result. A definition turns out to be too broad. A supposed consensus disappears after stronger studies are found.

The researcher changes course.

This capacity for revision is one of the most important differences between research and advocacy conducted backwards.

The researcher may begin with a hypothesis.

They should not be required to end with it.

The ACRL Framework Is Itself Being Revised

There is an interesting contemporary example of the same principle.

ACRL's Framework for Information Literacy for Higher Education was adopted in 2016 and remains the organisation's current adopted Framework. Its six influential concepts include Research as Inquiry, Searching as Strategic Exploration, Authority Is Constructed and Contextual, Information Creation as a Process, Information Has Value and Scholarship as Conversation.

But the information environment has changed substantially since 2016.

Generative AI, algorithmic information systems and new digital publishing practices have altered how researchers discover and evaluate information. ACRL therefore released a first draft of a revised Framework in April 2026 for consultation. As of August 2026, the existing 2016 Framework remains the adopted version rather than the draft revision.

That is research culture working at the institutional level.

Frameworks themselves have to remain revisable when the environment changes.

Frequently Asked Questions

What are research skills?

Research skills are the abilities required to turn a question into a well-supported answer. They include question formulation, search strategy, source evaluation, note-taking, comparison of evidence, synthesis, citation, ethical use of information and communicating uncertainty appropriately.

What is the first step in research?

The first meaningful step is defining the question. Searching before the problem has been framed often produces large amounts of information without a clear principle for determining relevance.

What makes a good research question?

A good research question is specific enough to investigate, important enough to justify investigation and appropriately scoped for the available time and evidence. It should make clear what phenomenon, population, relationship or problem is being examined.

What is the difference between a research topic and a research question?

A topic identifies an area, such as climate change or artificial intelligence. A research question specifies what the researcher wants to determine about that topic.

Why does a research question sometimes change?

New evidence can reveal that the original question was too broad, used the wrong terminology or contained an unsupported assumption. Revision is therefore a normal part of research rather than evidence of failure.

What is information literacy?

Information literacy broadly concerns the ability to find, evaluate, understand, use and create information appropriately. ACRL's Framework treats it as a complex set of practices rather than a simple checklist of search skills.

What does “Research as Inquiry” mean?

It means research is an iterative process driven by questions. Answers can lead to new questions, and investigation may require the researcher to revise the original problem.

What does “Searching as Strategic Exploration” mean?

It refers to the idea that searching is nonlinear and requires adaptation. Researchers refine terminology, choose different tools and change strategies as their understanding develops.

How do I know whether a source is reliable?

Reliability depends on the question. Examine who produced the information, their expertise, the method used, the evidence provided, the source's limitations, its date and whether independent evidence supports the claim.

Are primary sources always better than secondary sources?

No. Primary sources provide direct evidence, but a strong systematic review or expert synthesis may provide a better overview of an entire evidence base. Source type should be matched to the question.

Is Google acceptable for research?

Yes, for many purposes. Google can be excellent for locating official documents, organisations and publicly accessible material. More specialised questions may require academic databases, archives, legal databases or discipline-specific search tools.

Is Google Scholar enough for academic research?

It can be very useful, but no single search system covers every relevant source or provides the best controls for every discipline. Important projects often benefit from multiple search tools.

How many sources do I need?

There is no universal number. The appropriate evidence base depends on the complexity of the question, type of project and degree of certainty required. Ten weak sources do not become stronger than three highly relevant, independent and well-designed sources simply because there are more of them.

What is source triangulation?

Triangulation involves comparing evidence from multiple sources, methods or perspectives to determine whether a conclusion receives independent support. It is particularly useful when no single source can answer the entire question.

What is citation chaining?

Citation chaining involves using the references in a useful source to find earlier research and looking at later works that cite it. It helps researchers trace how a scholarly conversation developed.

What is synthesis in research?

Synthesis means combining and comparing evidence to explain the overall pattern. It goes beyond summarising studies individually by examining agreement, disagreement, methodological differences and gaps.

How do I avoid confirmation bias in research?

Actively search for serious evidence against your preferred explanation. Examine critiques, null findings, replication attempts and alternative interpretations rather than collecting only supportive material.

Can AI be used for research?

AI can assist with tasks such as brainstorming search terms, explaining unfamiliar terminology, organising supplied information and generating possible questions. Important factual claims should still be traced to reliable sources, and AI-generated references should be independently verified. UNESCO recommends critical scrutiny of GenAI outputs in education and research.

Can I cite an AI answer as evidence?

Citation rules differ by institution and publication, but for factual research the stronger practice is usually to locate and cite the underlying evidence rather than treating an AI-generated synthesis as the original source of the factual claim.

How do I know when research is finished?

Research is sufficiently complete when the evidence base is appropriate for the scope of the question, major relevant perspectives and counterevidence have been considered, additional searching produces little material change and the resulting conclusion can be stated with defensible confidence.

Why is uncertainty important in research?

Because evidence varies in strength. Stating uncertainty prevents conclusions from becoming stronger than the methods and data justify.

Research Is a Process of Building an Evidence Chain

At its simplest, good research can be understood as a chain.

A question determines what evidence is relevant. Search methods determine which evidence becomes visible. Source evaluation determines how much weight that evidence deserves. Comparison determines whether individual findings form a broader pattern. Synthesis connects the pattern to an answer. Citation allows other people to inspect the chain.

A weakness anywhere can affect the conclusion.

A brilliant search cannot rescue a badly framed question. Prestigious sources cannot rescue an inference they do not actually support. A perfectly formatted bibliography cannot rescue selective evidence.

Research quality lies in the connections.

The Central Idea

Research is the disciplined transformation of uncertainty into a conclusion that deserves more confidence than the researcher had at the beginning.

That transformation requires much more than finding information.

The question has to be framed carefully. The vocabulary of the field has to be learned. Search strategies need to change as understanding develops. Evidence has to be selected according to the type of question being asked. Sources must be evaluated for relevance, method and authority rather than prestige alone.

Researchers need to preserve a source trail, distinguish quotation from interpretation, recognise when apparently independent sources originate from the same evidence and compare disagreement rather than hiding it.

They must also allow evidence to change the question.

Generative AI can accelerate parts of this process, but it does not remove the need for source verification. If anything, an environment in which plausible answers can be produced instantly makes the distinction between retrieving an answer and establishing an answer more important. UNESCO's guidance reflects this by urging researchers and learners to treat generative AI outputs critically rather than as inherently authoritative information.

The strongest research skill is therefore not knowing every database, memorising every citation style or finding the largest number of sources.

It is revisability.

A strong researcher can begin with an incomplete question, discover better terminology, find evidence that challenges the initial assumption and emerge with a conclusion different from the one they expected.

That is not losing control of the research.

That is what research is for.