The CRAAP test is a five-criterion checklist — Currency, Relevance, Authority, Accuracy, Purpose — for evaluating whether a source is credible enough to use. It was built for library instruction and works well as a first pass on any unfamiliar source. It was not built for the specific problems that come up when the source is a journal article: whether the venue is indexed or predatory, whether the paper has since been retracted, whether a conflict of interest changes how its findings should be weighted, and whether an AI tool surfaced a citation that does not actually exist. This guide covers the standard framework, extends it to those scholarly-specific checks, and is honest about where the checklist approach itself breaks down — because a page that just recites the acronym does not actually help anyone evaluate a source.
What is the CRAAP test?
The CRAAP test was developed by Sarah Blakeslee, a librarian at California State University, Chico, in 2004, as a mnemonic for a first-year information-literacy workshop. It has since become one of the most widely taught source-evaluation frameworks in academic libraries, precisely because the acronym is easy to remember: Currency, Relevance, Authority, Accuracy, Purpose. It asks five plain questions of any source — a webpage, a book, a news article, or a journal article — before you rely on it.
The test’s popularity is also its limitation. It was designed to be applied quickly, by undergraduates, to a wide range of source types with no domain expertise assumed. Applied to scholarly literature by a researcher, research administrator, or reviewer, each of its five criteria needs to go further than the general-purpose version does — and, as covered later in this guide, the checklist format itself has real, documented weaknesses that a researcher should know about before treating it as sufficient on its own.
The five CRAAP criteria, expanded for scholarly sources
Currency
When was the source published, and has it been updated or superseded since? For most topics this is straightforward — a 2009 review of a fast-moving field has probably been overtaken by more recent primary literature. But currency cuts both ways in scholarly work: a foundational paper being old is not itself a problem if you are citing it for a well-established finding, a method that has not changed, or as a historical primary source. The real question is not “how old is this” but “has anything material changed since this was published that would change its conclusions” — which for an empirical claim usually means checking whether the finding has since failed to replicate or been revised by later work, not just checking a date.
Relevance
Does the source actually address your research question, at the right level and for the right audience? A source can be current, authoritative, and accurate, and still be the wrong source — a general audience explainer cited where the primary study is needed, or a study in a different population/context than the one your claim is actually about. Relevance also means checking you are reading the primary source rather than a secondary summary of it: citation-chain decay, where a claim gets passed from paper to paper via “as cited in” references without anyone checking the original, is a real and well-documented problem in the literature. If you cannot locate and read the original source, that is a reason for caution about the claim, not a reason to cite the intermediary as if it were the original. CASRAI’s guide to primary vs. secondary sources in research covers this distinction in depth — including why the classification depends on your research question rather than being fixed to the document itself.
Authority
Who wrote this, and what is their basis for authority on the specific claim being made? For a scholarly source, authority is not just “does the author have a PhD” — it includes institutional affiliation, whether the work went through peer review and what kind (the rigor of peer review varies substantially by venue and model), whether the journal or publisher is indexed by a recognized body, and whether the author has domain-relevant expertise for this specific claim (a credentialed expert in one field writing outside it is not automatically authoritative there). Authority also extends to the publishing venue itself — see the “applying CRAAP to scholarly sources” section below, since this is where a generic CRAAP treatment usually stops short of what a researcher actually needs to check.
Accuracy
Is the information correct, verifiable, and supported by evidence — and can you actually check that? For a scholarly source, accuracy checks include whether the methods and data support the stated conclusions, whether the work has been independently replicated or corroborated, whether the underlying data or code are available for inspection, and whether the source has since been corrected or retracted. Accuracy is the criterion most damaged by treating CRAAP as a source-in-isolation exercise — verifying accuracy properly usually means looking outward, at what other sources say about this one, not just reading the source itself more carefully. That distinction is the core of the lateral-reading critique covered later in this guide.
Purpose
Why was this source created, and does that purpose introduce bias? Is it reporting original research, advocating a position, selling a product or service, or written to satisfy a funder or sponsor’s interests? For scholarly sources this means reading the conflict of interest and funding disclosure statements, not skipping them — a study funded by an entity with a direct stake in its outcome is not automatically wrong, but the disclosure changes how much independent scrutiny the claim deserves. Purpose also covers the venue’s business model: a “publish quickly for a fee” incentive structure, as opposed to a subscription or genuinely open-access model, changes what selection pressure the venue is optimizing for.
Applying CRAAP to scholarly sources specifically
This is the part a generic library-guide treatment of CRAAP usually does not cover in depth, and it is where the checklist earns its keep for research work. Five checks worth running on any journal article before you rely on it:
Check whether the journal is indexed — and where
Whether a journal appears in a recognized index is a meaningful, checkable signal about editorial vetting. For open-access journals, check the Directory of Open Access Journals (DOAJ), which applies its own inclusion criteria and periodically removes journals that stop meeting them. More broadly, check whether the journal is indexed in Scopus or in Clarivate’s Web of Science Master Journal List. Absence from any recognized index is not automatic disqualification — some legitimate, newer, or regional journals are genuinely not yet indexed — but it is a reason to look harder at the other criteria below rather than take the venue’s own claims about itself at face value.
Screen for predatory and hijacked journals
A predatory journal charges publication fees while skipping or faking the peer review and editorial vetting it claims to provide. A related and increasingly common problem is journal hijacking, where fraudulent actors clone the name, ISSN, and branding of a legitimate journal to solicit and publish papers the real journal never saw. Warning signs include: aggressive solicitation emails, implausibly fast peer review turnaround for the field, a fee structure that is not clearly disclosed until after acceptance, an editorial board that cannot be verified independently, and a website with only superficial resemblance to the journal it claims to be. CASRAI’s guide on how to identify predatory journals and publishers covers the specific indicators and verification steps in full.
Check for retractions and post-publication concerns
A source’s credibility can change after publication. Before relying on a specific paper, check whether it has been retracted, corrected, or flagged with an expression of concern. Retraction Watch maintains a searchable database of retractions across the literature, and PubPeer is a post-publication peer review platform where researchers raise methodological or integrity concerns about published papers — including ones that have not (yet) been formally retracted. Neither database is exhaustive, and a PubPeer comment thread is not itself a finding of misconduct, but both are free, fast checks that a purely in-text reading of the source cannot surface on its own.
Read the conflict of interest and funding statements
Nearly every reputable journal now requires authors to disclose funding sources and competing interests, and most papers state these plainly near the end of the article. Read them. A disclosed conflict does not automatically invalidate a finding, but it is directly relevant to the Purpose criterion above, and its absence in a venue that should require it is itself a signal worth noting.
Check whether data and code are available
Whether a paper makes its underlying data and analysis code available for independent inspection — through a repository, supplementary materials, or a stated data-availability statement — is a strong, checkable proxy for how seriously the authors and venue take verifiability. Its absence does not mean the work is wrong, but its presence lets you (or anyone) actually check the Accuracy criterion above rather than take it on trust.
The limits of CRAAP — and why information-literacy research has moved past it
CRAAP’s core limitation is structural, not a matter of applying it more carefully: it is a checklist applied to a source in isolation. You read the source, ask five questions of it, and reach a verdict — all without ever leaving the page. Information-literacy research over the past decade has found this approach measurably weaker than an alternative called lateral reading: instead of scrutinizing a source’s own content and design cues for longer, you open new tabs and check what other, independent sources say about this source, its publisher, and its author. Professional fact-checkers were found to do this reflexively and effectively; students trained on checklist-style vertical evaluation, working within a single page, performed markedly worse at distinguishing credible from non-credible sources — sometimes rating a well-produced fake source as trustworthy specifically because it presented as authoritative on its own page.
The practical implication for research work is not “discard CRAAP” — its five questions are still the right things to ask. It is that CRAAP should never be the last step. A source that reads as professional, current, and well-cited within its own page can still be a predatory venue with a convincing website, an author with an undisclosed conflict, or a paper that has since been quietly retracted — none of which vertical, in-page reading alone will surface. The scholarly-specific checks in the section above (journal indexing, retraction databases, COI disclosures) are, in effect, lateral reading applied to the research-literature context: they all involve leaving the source to check it against independent, external information.
Alternatives: SIFT and lateral reading
SIFT is a shorter, action-oriented alternative built directly around lateral reading, commonly associated with digital literacy researcher Mike Caulfield. Its four moves:
- Stop — before reading further or sharing a claim, pause and check whether you actually know the source and whether it is worth your time.
- Investigate the source — spend a minute establishing who is behind the source and what their expertise, agenda, and track record are, before evaluating the content itself.
- Find better coverage — instead of relying on your own assessment of a single source, check what other, more authoritative sources say about the same claim.
- Trace claims, quotes, and media to the original context — if a claim, statistic, or quote has been passed along through secondary reporting, follow it back to where it originated before citing it.
For scholarly source evaluation specifically, SIFT and the scholarly-specific CRAAP checks above converge on the same underlying discipline: verify externally, don’t just read carefully. Neither framework is a substitute for domain expertise, and neither replaces actually reading the methods section — but both correct for the same failure mode, which is trusting a source’s own presentation of itself.
Evaluating AI-generated and AI-surfaced sources
Generative AI tools introduce a source-evaluation problem that predates CRAAP and SIFT alike: fabricated citations. Large language models can produce references that look completely plausible — a real-sounding author, a real journal name, a real-looking DOI or year — that do not correspond to any actual publication, or that misattribute a real claim to the wrong paper. This is a known, well-documented failure mode of current-generation models, not an occasional glitch, and it is distinct from an AI tool simply being wrong about a factual claim: a fabricated citation can look, at a glance, exactly like a correctly formatted real one.
The practical rule is unconditional: every citation an AI tool surfaces — whether it generated the reference itself or is summarizing/quoting a source it found — must be independently verified at the original source before you cite it or rely on it. That means locating the actual paper (via the publisher’s site, a DOI resolver, or a database like Scopus or Web of Science), confirming it says what the AI claimed it says, and running it through the same evaluation above — indexing, retraction status, COI, data availability. A citation is not verified because the DOI resolves; it is verified because you read the actual source and confirmed the claim matches. See CASRAI’s guidance on AI-assisted writing for the broader disclosure and verification norms that apply when AI tools are used anywhere in the research or writing process, and the guide on quoting and citing sources correctly for citation mechanics once a source is verified.
A practical source-evaluation checklist
- Read the source itself: apply Currency, Relevance, Authority, Accuracy, Purpose.
- Leave the source: search independently for the author, publisher, and journal before trusting the source’s own self-description.
- Check journal indexing (DOAJ for open access, Scopus, Web of Science Master Journal List) and screen for predatory or hijacked-journal warning signs.
- Check retraction and post-publication-review status (Retraction Watch, PubPeer) for the specific paper.
- Read the conflict-of-interest and funding-disclosure statements.
- Check whether underlying data and code are available for independent inspection.
- If the source or claim reached you via an AI tool, verify the citation at the original source before using it — never cite an AI-surfaced reference you have not personally located and read.
- If you are relying on a summary or “as cited in” reference rather than the original, either locate the original or flag explicitly that you have not verified it firsthand.
What to do when sources conflict
Two credible-looking sources disagreeing is common in an active research area and is not, by itself, a sign that one of them is untrustworthy. Before concluding either source is wrong:
- Check whether they are actually answering the same question — different populations, methods, timeframes, or definitions can produce genuinely different, non-contradictory findings.
- Check publication dates and citation relationships — a later source may have already engaged with and addressed the earlier one, or vice versa; check whether one cites and responds to the other.
- Weigh study design and evidence strength rather than defaulting to whichever source is more recent, more prestigious-sounding, or easier to find — a well-designed older study can outweigh a weaker recent one.
- Check whether either source has since been corrected, retracted, or subject to a documented replication failure.
- Where the disagreement remains genuinely unresolved in the field, say so directly in your own writing rather than silently picking a side — representing an active disagreement as settled is itself a credibility problem.
Frequently asked questions
What does CRAAP stand for?
Currency, Relevance, Authority, Accuracy, and Purpose — five questions to ask of any source before relying on it.
Who created the CRAAP test?
Sarah Blakeslee, a librarian at California State University, Chico, developed it in 2004 for a first-year information-literacy workshop.
Is the CRAAP test still useful for research, or is it outdated?
Its five questions are still useful, but on their own they are not sufficient for scholarly work. Information-literacy research shows checklist-style, in-page evaluation is measurably weaker than lateral reading — leaving the source to check it against independent, external information. Use CRAAP’s questions alongside the scholarly-specific checks (journal indexing, retractions, COI) covered in this guide, not instead of them.
What is lateral reading and how is it different from CRAAP?
Lateral reading means opening new tabs to check what independent sources say about a source, its author, and its publisher, instead of evaluating the source only by reading it more carefully on its own page. It is an approach, not a checklist, and research on professional fact-checkers’ behavior found it more effective than vertical, in-page evaluation.
What is the SIFT method?
A four-step alternative to checklist source evaluation: Stop, Investigate the source, Find better coverage, and Trace claims to their original context. It is built around the same lateral-reading discipline as the scholarly-specific checks in this guide.
How do I check if a journal is predatory?
Check whether it is indexed in DOAJ (for open access), Scopus, or Web of Science’s Master Journal List; verify the editorial board independently; and watch for aggressive solicitation, undisclosed fees, and implausibly fast review. See CASRAI’s full guide on identifying predatory journals and publishers.
Can I trust a citation an AI chatbot gives me?
Not without checking it. AI tools can fabricate plausible-looking citations — real-sounding authors, journals, and DOIs that do not correspond to an actual publication. Always locate and read the actual source before citing it.
What should I do when two credible sources disagree?
Check whether they are actually answering the same question, check whether either has been corrected or retracted, weigh study design rather than recency alone, and if the disagreement is genuinely unresolved in the field, say so in your own writing rather than picking a side.
