03 · What You Need to Know
Why Local Research Literature Can Become Less Visible in AI-Assisted Discovery
AI Literature Discovery Is Not Necessarily a Comprehensive Search
Generative AI systems differ in how they obtain information about scholarly publications.
Some generate responses from patterns learned during training. Others retrieve documents from scholarly databases, search engines, publisher platforms, or connected repositories.
These approaches have different strengths and limitations.
A language model without access to current bibliographic sources cannot reliably establish whether a particular publication exists or whether a literature search is complete.
A retrieval-enabled assistant may identify current publications, but its coverage remains limited by the sources it can access and the procedures it uses.
Researchers should distinguish AI-generated bibliographic suggestions from a documented literature search.
A convincing list of references is not evidence that all relevant studies have been identified.
International Indexing Databases Do Not Cover Every Relevant Publication
Scholarly databases differ in their journal coverage, language representation, disciplinary scope, and inclusion criteria.
For example, Scopus and Web of Science index substantial bodies of scholarly literature, but neither contains every legitimate research publication.
Studies published in regional journals or institutional repositories may be absent from one or both databases.
Some publications may be discoverable through national research portals, university repositories, or other specialized sources.
Researchers should therefore select databases according to the review question rather than assuming that one index provides universal coverage.
Database inclusion is also not, by itself, proof that every indexed article is methodologically sound.
Local Journals May Be Less Visible in International Search Systems
Regional journals may publish research addressing national educational policies, institutional practices, local populations, or context-specific challenges.
Some have limited international indexing coverage, incomplete metadata, or relatively low citation visibility.
Consequently, they may be less likely to appear in AI-assisted recommendations that depend heavily on prominent scholarly sources.
For example, a Philippine educational technology study may be relevant to a local research question even if it has received relatively few international citations.
Researchers should evaluate the publication's relevance, methodology, and credibility rather than relying exclusively on journal visibility.
The objective is not to privilege local journals automatically but to avoid excluding relevant evidence merely because it is less internationally prominent.
Institutional Repositories Can Contain Relevant Research
Universities often maintain repositories containing theses, dissertations, working papers, technical reports, and other scholarly outputs.
These materials may address research questions that are insufficiently represented in international journal databases.
For example, an institutional thesis may examine the implementation of asynchronous learning within a particular university system.
Such evidence may help researchers understand the local context or identify additional publications.
However, repository inclusion does not necessarily indicate that a work has undergone journal peer review.
Researchers should distinguish publication types and evaluate their methodological quality before incorporating findings into a synthesis.
Government and Policy Documents May Be Essential Sources
Some research questions cannot be adequately answered through journal articles alone.
Studies examining educational policy, institutional governance, or regulatory implementation may require official government documents and administrative reports.
For example, research on Philippine higher education may need to consider relevant issuances from the Commission on Higher Education, official institutional policies, or national statistical publications.
An AI-generated literature review that focuses exclusively on journal articles may omit these sources.
Researchers should distinguish scholarly evidence from legal, regulatory, and administrative documents because they serve different evidentiary purposes.
Authoritative policy sources should be consulted directly when determining applicable requirements.
Language Differences Can Reduce Discoverability
Local research may use terminology in national or regional languages that differs from English-language scholarly expressions.
AI systems and retrieval platforms may not represent these languages equally.
Consequently, searches conducted only in English may overlook relevant studies.
For example, research involving Filipino educational experiences may use locally meaningful expressions that do not correspond directly to common international keywords.
Researchers should consider multilingual search terms where appropriate and verify translated concepts against the original language.
The broader problem of English-language conceptual bias in AI-assisted research may influence both discovery and interpretation.
Local Terminology May Differ From International Research Vocabulary
Researchers may describe similar phenomena using different terminology across countries or disciplines.
For example, one publication may use "distance education," another "open and distance learning," and another "online flexible learning."
These expressions may overlap without being fully interchangeable.
An AI system that searches only for one preferred phrase may overlook studies using alternative terminology.
Researchers should identify relevant synonyms, historical terms, local program descriptions, and institutional language.
Search expansion should preserve conceptual distinctions rather than assume that every related term refers to the same phenomenon.
Publication Metadata Can Affect Retrieval
Bibliographic discovery depends partly on titles, abstracts, keywords, author information, identifiers, and other metadata.
Incomplete or inconsistent metadata can make relevant studies harder to locate.
For example, a publication may have a locally formatted title, inconsistent institutional affiliation, or missing persistent identifier.
These problems may affect retrieval and citation matching.
Researchers should use alternative search fields, exact title searches, author searches, and repository browsing when appropriate.
Failure to retrieve a study through one search expression should not be interpreted as evidence of absence.
Citation-Based Ranking Can Favor Already Visible Research
Some scholarly discovery systems use citation relationships or other indicators of prominence to organize results.
Highly cited publications may receive greater visibility than newer or locally focused studies.
This can be useful when identifying influential research but may disadvantage relevant work with limited citation exposure.
For example, a recently published study from a provincial institution may address the research question directly while receiving relatively little attention in international citation networks.
Researchers should distinguish scholarly influence from contextual relevance.
Citation counts are not substitutes for evaluating whether a publication answers the review question.
AI May Favor Familiar Theories and Their Associated Literature
Generative AI may associate particular research topics with widely used theoretical frameworks.
For example, research concerning educational technology adoption may prompt recommendations involving technology acceptance models.
These frameworks can be appropriate, but repeated reliance on them may narrow the literature considered.
Studies using critical, institutional, cultural, or locally developed perspectives may receive less attention.
Researchers should examine whether AI favors dominant scientific theories in ways that limit relevant evidence discovery.
Theoretical diversity should be pursued according to the research question rather than as an arbitrary numerical requirement.
AI Can Confuse Missing Literature With a Research Gap
A research gap is a defensible claim about limitations or unanswered questions in the available body of relevant evidence.
It is not established merely because an AI system fails to recommend publications on a topic.
For example, an AI assistant may suggest that no Philippine studies examine generative AI literacy among university students.
Before accepting that claim, researchers should search relevant international and local databases, repositories, and other appropriate sources.
They should also examine related terminology and publication dates.
The absence of evidence from one AI response is not evidence that the research itself is absent.
AI Can Produce Incorrect Claims About Local Research
Generative AI may fabricate references, misattribute findings, or describe publications inaccurately.
These errors can be particularly difficult to detect when researchers are unfamiliar with local journals or institutions.
For example, a model may generate a plausible article title involving a Philippine university that does not correspond to an actual publication.
It may also associate a genuine article with findings that the authors did not report.
Researchers should verify bibliographic records and consequential claims against original sources.
A plausible citation is not evidence that the cited publication exists or supports the associated statement.
Local Literature Is Not Automatically More Relevant
Geographic proximity can improve contextual relevance, but it does not establish scientific quality.
A local publication may be methodologically weak or address a substantially different research question.
Conversely, an international study may provide strong theoretical or methodological evidence applicable to the local context.
Researchers should evaluate all sources according to transparent eligibility and quality criteria.
Including local literature should improve the completeness and contextual suitability of the evidence base rather than function as a geographic quota.
Grey Literature Can Be Useful but Requires Critical Appraisal
Grey literature includes research and information distributed outside conventional commercial or academic publishing channels.
Examples may include institutional reports, policy documents, working papers, and selected conference materials.
These sources can be valuable when studying emerging technologies, local implementation, or policy development.
However, their review procedures and methodological transparency may vary.
Researchers should evaluate authorship, provenance, methods, relevance, and potential conflicts of interest.
The inclusion of grey literature should follow the purpose and eligibility criteria of the review.
Searching Local Sources Can Reduce Evidence Selection Bias
Researchers can supplement AI recommendations through independent searches of relevant national and institutional sources.
For Philippine educational research, this may include university repositories, official education agency publications, and locally relevant journals.
Search procedures should be documented sufficiently to explain how evidence was identified.
Researchers should also consider citation chasing, reference-list screening, and consultation with relevant disciplinary specialists.
These approaches may identify studies missed by initial AI recommendations.
However, no single supplementary search guarantees complete coverage.
AI Can Help Generate Search Terms Without Determining Search Completeness
Generative AI may be useful for identifying synonyms, related constructs, and alternative expressions.
For example, a researcher examining asynchronous learning may ask for terminology associated with distance education, flexible learning, and online course delivery.
The suggestions can help develop a broader search strategy.
However, researchers should verify that the terms are conceptually appropriate and avoid adding irrelevant expressions merely to increase retrieval volume.
AI-generated keywords should support, rather than replace, a transparent search methodology.
Systematic Reviews Require More Than AI-Generated Recommendations
A systematic review requires a search strategy appropriate to its research question and reporting framework.
Researchers should specify eligible sources, search expressions, inclusion criteria, screening procedures, and relevant limitations.
AI may assist with selected stages, but its outputs should be evaluated and documented.
For example, AI-assisted screening may require validation to determine whether relevant local studies are disproportionately excluded.
Researchers should not describe an AI-generated reference list as a systematic search merely because it contains many publications.
Applicable review reporting guidance, such as PRISMA 2020, should be considered where appropriate.
Local Research Coverage Can Be Audited
Researchers may evaluate whether an AI discovery tool systematically overlooks relevant local publications.
One approach is to develop an independently verified reference set of eligible studies.
The researcher can then determine which publications the AI system retrieves under specified search conditions.
Relevant comparisons may involve language, publication source, country, publication year, or disciplinary area.
However, the reference set itself may be incomplete, and omission rates should be interpreted within its coverage.
The procedures for testing AI-assisted research workflows for bias can be adapted to literature discovery tasks.
Researchers Should Distinguish Retrieval Failure From Geographic Bias
A missing publication may result from several mechanisms.
The article may not be indexed, its metadata may be incomplete, the search expression may be unsuitable, or the retrieval system may rank other publications more highly.
Geographic bias is one possible explanation but should not be assumed from an isolated omission.
Researchers should investigate whether relevant studies from particular countries are systematically underrepresented under comparable retrieval conditions.
The distinction matters because different causes require different corrective procedures.
Missing Local Literature Can Distort Research Conclusions
Literature selection influences theoretical framing, research gap identification, and interpretation.
For example, a review dominated by studies from highly resourced universities may underestimate the importance of connectivity constraints or employment obligations in other settings.
If locally relevant evidence is omitted, the resulting synthesis may present an incomplete account of the phenomenon.
Researchers should examine whether additional studies materially change the interpretation.
The concern relates to biased AI contributions affecting research conclusions.
Omissions should be evaluated according to their substantive importance rather than counted without considering relevance.
Researchers Should Document Search Coverage and Limitations
Research reports should identify the databases, repositories, search terms, language restrictions, and relevant dates used to locate literature.
Where AI contributes to discovery, researchers should document its role and the procedures used to verify recommendations.
They should distinguish sources retrieved independently from those suggested by AI.
Where relevant local evidence may remain inaccessible, the limitation should be acknowledged.
Transparent documentation helps readers evaluate whether claims about the state of the literature are adequately supported.
