Focal Interest publishes a filtered subset of the research indexed by OpenAlex, an open scholarly works database containing over 250 million works. The filtering is automated and operates at ingestion — before any article enters the publishing pipeline. This page documents the specific criteria used to decide what is and is not published.
These criteria are published in full because transparency about filtering decisions is part of what makes the site trustworthy.
The current published corpus contains 2,166 articles across 7 subject areas, with approximately 85% available as open access.
What this means for you
Every article on this site has passed the same automated filtering system before publication. That means when you read a summary here, the underlying work has already cleared checks for publication type, source credibility, language verifiability, abstract quality, and retraction status — among other criteria documented below.
The goal is not to show you everything that exists in the scholarly record. It is to show you a filtered subset where the source material meets a consistent, documented standard — so you can spend your attention on understanding the research rather than evaluating whether it belongs here.
What is verified before publication
Before an article enters the publishing pipeline, the system confirms the following: The work is classified as a journal article or review in the source database. A DOI or equivalent persistent identifier is present. The abstract is long enough to support a meaningful, verifiable summary. The publication date has arrived — pre-publication drafts are not summarized. The source language is English, confirmed through independent assessment of the abstract. The work has not been retracted according to CrossRef or Retraction Watch records. The publishing source is not on the maintained list of known predatory publishers.
Works that clear these checks and do not trigger additional soft-signal review proceed to automated summarization. The specific exclusion criteria and soft signals are documented in full below.
These standards form the foundation for the trust signals shown alongside each article.
What gets excluded automatically
The following categories of work are hard-excluded at ingestion. No work matching these criteria is published, regardless of its other characteristics.
The exclusions below share a common purpose: the pipeline can only publish content it is able to evaluate and summarise accurately. Works that fall into these categories are either not primary research, not verifiable through available metadata, or not compatible with the pipeline’s summarisation model. Each exclusion is applied automatically at ingestion — no human review is required, and no excluded work enters the publishing queue.
Work type exclusions
Works classified as editorials, errata, corrections, retractions, letters to the editor, book reviews, dissertations, or theses are excluded. These are not primary research publications.
Known predatory publishers
Works from publishers identified as predatory — operating journals that accept content indiscriminately without genuine peer review — are excluded via a maintained list of DOI prefixes. Each entry is documented with the evidence that justified its addition.
Title pattern exclusions
Works whose titles match known non-research patterns are excluded by automated title screening regardless of how the work is otherwise classified. Excluded patterns include titles beginning or ending with terms such as “book review,” “response to,” “correction,” “reply,” “commentary,” “letter,” “correspondence,” “retraction,” “erratum,” “corrigendum,” “comment on,” and similar markers.
Repository-only sources
Works sourced exclusively from preprint servers or institutional repositories, without a confirmed journal record, are excluded unless the work has accumulated at least one citation in the scholarly literature within one year of publication. This threshold is configurable and represents the default setting.
Future publication dates
Works whose official publication date has not yet arrived are excluded. A paper is not published until its stated publication date, regardless of whether it has been assigned a DOI or is accessible in a repository. This ensures summaries reflect the version of record rather than pre-publication drafts.
Retracted works
Works identified as retracted by CrossRef or the Retraction Watch database are excluded at ingestion and removed from publication if identified after publication. See the retraction monitoring section below.
Incomplete or unusable abstracts
Works without an abstract, or with abstracts that are too short to support a meaningful summary, are excluded.
Non-English works
Works whose primary language is not English are excluded. Focal Interest is an English-language publication serving English-language readers — AI-generated summaries of non-English abstracts cannot be verified by readers against the original work. Language is determined by a combination of source metadata and independent AI assessment at ingestion time.
What triggers additional automated review
Some works pass the hard exclusion checks but contain signals that warrant closer automated scrutiny before publication. These signals contribute to a scoring threshold — any single signal may not be enough to exclude a work, but several in combination will. This approach exists because scholarly databases may misclassify content types — a framework document, a thesis abstract, or a conference submission can appear in the same index as peer-reviewed journal articles. The soft-signal system is designed to catch what source classification misses.
Abstract self-identification
Works whose abstracts contain phrases indicating they are not primary research publications — including references to undergraduate or graduate thesis requirements, working paper series, policy briefs, conference proceedings, or circulated drafts — are flagged for exclusion regardless of how they are classified in the source database. Academic databases occasionally misclassify work types; abstract language provides a direct signal from the authors themselves.
Student voice indicators
Works whose abstracts reference thesis advisors, dissertation committees, course requirements, or capstone projects are flagged as likely student work.
Working papers and grey literature
Works whose abstracts identify themselves as discussion papers, NBER working papers, policy briefs, or pre-publication circulated drafts are flagged as grey literature.
Source quality signals
Works from sources without a confirmed ISSN — the standard identifier for peer-reviewed journals — combined with other indicators such as low citation counts or unverified source classification, accumulate as a combined signal toward exclusion. An absent ISSN is not by itself grounds for exclusion but increases scrutiny of the work’s other characteristics.
Single-page works from unverified sources
Works occupying a single journal page from sources without a confirmed ISSN are flagged as likely meeting abstracts or correspondence rather than primary research.
Conference papers
Works whose abstracts identify themselves as presented at or accepted to a conference, as proceedings papers, or as workshop submissions are flagged as likely conference papers rather than peer-reviewed journal articles. Note that top-venue conference papers in some fields — particularly computer science — are genuinely peer-reviewed and high quality; this signal is weighted accordingly and does not result in automatic exclusion.
Ongoing quality monitoring
Quality filtering does not stop at publication. Several processes run automatically after an article is live.
Retroactive exclusion review
When quality standards are updated or new exclusion criteria are introduced, an automated nightly scan audits the full corpus against current rules. Works that would not pass current ingestion standards are flagged in an internal review queue, where they can be actioned individually or held pending editorial review. This ensures the quality standard applies to the existing corpus, not just new content.
Retraction monitoring
All published works are checked against CrossRef and Retraction Watch on a weekly automated schedule. If a work is retracted after publication on this site, the corresponding article is updated or annotated to reflect its retracted status.
Data freshness
Published article metadata is periodically verified against OpenAlex to detect upstream changes, including classification updates, corrected author lists, or revised publication information.
Language validation
Language is assessed independently for a work’s title and abstract using a combination of character-based detection and AI model evaluation. These checks run as separate fields rather than a single combined assessment, which allows the system to identify cases where a title and abstract differ in language, or where one passes while the other does not.
When metadata indicates a non-English source but title and abstract validation return English, the system performs additional verification where possible to confirm the language of the full paper. Works where the body text is confirmed to be non-English are excluded.
This layered approach catches cases where source databases have incorrectly classified a non-English work as English — including works where the abstract is in English but the full paper body is not.
What these standards do not guarantee
Automated filtering substantially improves content quality but is not infallible. The pipeline operates on metadata and abstract text — it does not read full papers. A work that passes all ingestion criteria may still have methodological limitations, contested findings, or errors in the underlying research that are not detectable from the abstract alone. Summaries should be treated as research leads, not as endorsements of the underlying work’s conclusions.
For full details on how summaries are generated, see the Editorial Methodology page. For the site’s broader content policies, see the Editorial Policy page.
If you believe a published article should not meet these standards, please use the Contact page with a link to the article and a description of your concern.