Skip to main content

How does Keenious search and rank results?

Keenious performs searches based on your input. This explains what happens behind the scenes. We keep it updated as the approach evolves.

Updated yesterday

Why are we doing this?

We believe in transparency. Understanding how Keenious works helps you trust the results, steer the process, and notice why results might differ from what you expected.

What Keenious considers

Keenious builds an understanding of your information need from:

  • The words you type (prompts and follow-ups)

  • What’s currently displayed in the content frame (a paper or result list)

  • Any text you highlight

  • Uploaded PDFs (within size limits)

If your prompt is brief (e.g., “show me research on climate change”), our language model expands it to add helpful context so a meaningful search can run. You can always edit the query, add or remove filters, or take over completely.

The key factors that influence ranking

1) Semantic understanding of your request

Keenious interprets the meaning of what you’re asking (including entities and methodologies) to find papers truly about your subject—even if phrasing differs or the paper is in another language. This reduces missed-but-relevant results when wording varies.

2) Exact-term requirements (only when you say so)

If you explicitly state that certain terms must appear (e.g., “papers must mention PyMOL”), Keenious adds a precise term-matching clause to ensure those terms are present. Otherwise, strict term requirements are avoided so you don’t miss relevant papers that use alternate wording.

Balance: Papers that are strong on both semantic relevance and (when requested) exact-term evidence tend to rank higher.

3) Citation count

Citations provide a signal of visibility/influence. Helpful but imperfect: older papers naturally accumulate more citations, and practices vary by field.

4) Recency of publication

Newer work receives a gentle boost so current research is easier to find—without hiding foundational studies.

Additional practical signals

  • Abstract availability: Papers with abstracts are favored so you can assess relevance faster.
    ​

  • Format/DOI emphasis: Articles and reviews with DOIs get a boost; if strong matches are scarce, other work types from OpenAlex are included so useful material still appears.
    ​

Boosts vs. filters (you stay in control)

  • No automatic restrictive filters are applied. Instead, boosts influence ordering without hiding results.

  • Only automatic exclusion: retracted items are filtered out.

  • Everything else—years, document type, open access, language—is up to you and clearly adjustable in the interface.
    ​

Data fidelity & duplicates

Keenious mirrors OpenAlex faithfully. OpenAlex handles most versioning and de-duplication, but occasional duplicates can appear. We don’t rewrite or merge source records.

Personalization & privacy

There’s no personalization of ranking today. Any future experiments will be transparent and optional (opt-out available).

In summary

Keenious performs searches based on your input, gathers candidates through semantic understanding (and exact-term requirements when you request them), and ranks with a balanced blend of meaning, evidence, recency, and practical signals like abstracts—while filtering out only retractions. You retain full control at every step.

Did this answer your question?