Built for grant writers

The 12-minute reviewer test

When a programme officer or study-section reviewer reads your novelty claim ('to our knowledge, no prior work addresses X…'), the first instinct is to test it. They spend ~12 minutes — one search session in Scholar, Semantic Scholar, or OpenAlex. If they find a paper that refutes the claim, your whole aim is compromised.

This is a literature-completeness problem dressed up as a rhetorical one, and we attack it from two sides. The research-gap search runs your aim against OpenAlex live — the field overview and its most-cited and newest papers — and against a 4.5M-paper local library, where the 120K+ open research gaps are mined from 100,000+ papers, mostly drawn from each paper's own stated research gap, future-work, challenge and limitation notes, and its abstract. On the full draft, AI Review's 8 specialist agents include a Literature Coverage agent that evaluates your citations for completeness and missing seminal references, with a live OpenAlex snapshot of the field for context.

Three grant-writing workflows

Three ways to use it on a grant:

• Pre-submission aim check — paste your specific aim into /research-gaps?topic=<your-aim>. The search, which costs no credits, gives back the field overview and every paper it found, in one list you can filter (year, journal quartile, indexing, venue), sort (most cited, newest and more) and download as RIS or BibTeX. Unlocking the topic compares the papers on it field by field and writes the open questions out of that comparison: the sanity check on whether your novelty claim still holds. 20 minutes per aim.
• Preliminary data audit — submit your preliminary-data draft as a pseudo-manuscript to the 8-agent pipeline. The Methodology agent flags study-design weaknesses; the Originality + Prior Publication pair finds overlap with your own prior work and your competitors'.
• Post-rejection rescue — if a grant is rejected on novelty or insufficient preliminary evidence, re-run the specific aims against current literature (research moves fast; a gap that existed 18 months ago may have closed). Revise aims accordingly.

What the Research Gap Finder adds over Scholar

Google Scholar ranks by citation count and recency. OpenAlex has a better canonical citation graph but no native 'show me the gap' query. We layer:

• Research gaps written from the filled comparison table once you unlock the topic — each one points back at a cell you can read, not at a summary you have to trust.
• Supporting-gap evidence: the open-gap index behind it holds 120K+ gaps extracted from 100,000+ real papers in a 4.5M-paper local library, mostly drawn from each paper's own stated research gap, future-work, challenge and limitation notes, and its abstract.
• Field overview, queried live from OpenAlex: total works, top venues, the year range and the peak year. Lets you calibrate saturation before committing a grant cycle.
• Both foundational (most-cited) and recent (newest-first) literature from the same list — sort it either way, so you can cite both classes appropriately in the aim.

Study sections and reporting standards

The Methodology agent is prompted with the CONSORT, STROBE and PRISMA reporting checklists. NIH study sections increasingly apply these standards to preliminary-data sections. We flag missing reporting elements (power calculations, randomisation, blinding, etc.) before a reviewer does.

For large NSF/NIH mechanisms (R01, R21, U01, DP1–DP5), the full 8-agent report on your preliminary-data draft takes ~15 minutes and 30 credits, and reads as a 'mock study section' on the quantitative rigor of your prelim.

Frequently asked questions