Is Your Research Idea
Actually Worth Pursuing?

ResearchChecker runs your question through a Cloudflare-hosted backend that calls Llama 3.3 70B on the Groq inference platform, with the model constrained to reason only over the real API results it receives. Three parallel PubMed searches (raw terms, MeSH-style quoted phrases, and a broader formulation) are deduplicated and expanded with related-article neighbors via the NCBI elink endpoint. Europe PMC is queried twice with SRC:PPR and SRC:CTX filters for preprints and Cochrane reviews. ClinicalTrials.gov v2 is queried for both the full trial set and the active/recruiting subset. Semantic Scholar provides citation-count-ranked papers and year-bucketed publication trend data. OpenAlex is queried for cross-validation. PROSPERO is checked via a targeted PubMed mention search. Every paper count and DOI in the final report comes from one of these live API responses — the model never invents citations.

To get useful output, write your idea the way you would write a PICO question: the more specific the population, intervention, comparator, and outcome, the tighter the search. "Meta-analysis of dapagliflozin versus placebo in HFpEF on all-cause mortality" produces a sharper report than "SGLT2 and heart failure." Read the Evidence Landscape and Gap Analysis sections critically — they are AI interpretations of real data, and the underlying data is sometimes thin or has known coverage gaps (Semantic Scholar can block browser requests; PROSPERO is not directly queried). Every cited PMID, DOI, and NCT ID is a clickable link, so verify anything the report claims is from a specific paper.

ResearchChecker is designed for the earliest, cheapest stage of a research project — the moment you are deciding whether an idea is worth the time. It is meant to gauge the validity and novelty of an idea before resources are committed, not to replace a librarian-led search, a statistician consult, or a careful read of the primary literature. Use it as a screening tool. Do not skip the manual verification.