Consensus Adds Four Reusable Research Skills
The new workflows cover literature review, citation trails, source finding and citation checks while leaving study quality and evidentiary weight to the user.
Edited by Tyronne Panaino
Consensus introduced Skills on October 5 as four ready-to-run workflows for recurring research tasks. Literature Review, Citation Snowballing, Reverse Citation Search and Verify Citations move different parts of evidence discovery and checking into named product actions. The change is aimed at students, researchers and writers who repeatedly move between finding papers, tracing a field and testing whether a source actually supports a sentence.
The useful distinction is that these are not four versions of the same search. Two help a reader map or expand a body of research; two begin from a claim or draft and work backward toward evidence. Consensus also makes an important limitation explicit: retrieved passages and product verdicts are inputs to review, not substitutes for deciding whether a study is relevant or methodologically sound.
Literature Review starts from a question
The Consensus Skills announcement describes Literature Review as the workflow for building a structured overview from a research question. Its value is organizational. A user can begin with a topic and receive a map of relevant work rather than manually turning one search into many separate searches.
That output should be treated as a reading plan, not a settled account of a field. Coverage depends on the papers available to the system, the way the initial question is framed and the criteria used to group evidence. A responsible review still records inclusion choices, dates, populations and outcome measures so another reader can understand what the overview represents.
Citation Snowballing follows the research graph
Citation Snowballing starts from a paper and follows both its references and newer work that cites it. This is a different reader intent from a broad review. It helps trace the lineage around one result, find replications or extensions and identify where later papers changed the interpretation.
The workflow can reduce the mechanical work of moving backward and forward through citations. It cannot decide that every citing paper is supportive, independent or high quality. Citation links show a relationship between documents; researchers still need to read how the later work uses, challenges or merely mentions the starting paper.
Reverse Citation Search starts from the claim
Reverse Citation Search takes a sentence that needs support and looks for relevant research. That reverses the usual workflow: instead of searching a topic and later deciding what to write, the user presents the proposed claim and asks what evidence could justify, narrow or contradict it.
This can be particularly useful for drafts containing placeholder citations. The risk is confirmation bias if the user treats the first matching paper as validation. A stronger process asks whether the population, intervention, comparison and outcome actually match the sentence, then rewrites the claim when the evidence is narrower or mixed.
Verify Citations checks the draft-to-source link
Verify Citations examines a draft section and its citations, then assesses whether each source supports, contradicts or does not address the attached claim. That targets a common failure that ordinary reference managers do not solve: a citation can be correctly formatted and still fail to support the statement beside it.
The product's assessment is best used as triage. A flagged mismatch tells the writer where to read the underlying passage and methods. It is not an adjudication of truth, and an apparently supportive passage may still come from a weak design, an inapplicable population or a result that has not held up elsewhere.
The human checkpoint remains central
Together, the four workflows separate discovery, graph navigation, claim sourcing and citation audit. That separation can make research work more inspectable because the user can choose the stage they are actually trying to complete. It also creates clearer review questions: what was searched, which papers were followed, why a source was attached and whether the cited passage matches the prose.
The next useful evidence will be independent testing across disciplines and document types. Evaluators should measure recall of important papers, false support judgments, contradictory evidence found, time saved and the amount of human correction required. Without that evidence, the launch establishes product scope, not accuracy across every field.
Status
Confirmed. Consensus's first-party announcement supports the four named Skills and their intended research tasks. Internal confidence is medium because only the company's description was fetched and no independent evaluation of coverage or citation judgments was available in this run.
Sources
Update note: Last reviewed 2026-10-12. We will revise this post if independent evaluations clarify coverage, citation-checking accuracy or correction burden.
Sources
Drafted with AI assistance from source briefs; reviewed for citation completeness and label accuracy.