Unkwn42 is the missed-literature check you run before you submit a climate or earth science paper. A plagiarism checker makes sure you didn't copy words; unkwn42 makes sure you didn't miss earlier work. You paste your abstract and reference list and it answers three questions: is my main claim already published, what earlier work am I missing, and is my reference list clean. Every paper cited is a real retrieved record, never generated.

How does Unkwn42 help researchers?

Before submitting, a researcher finds the papers a referee is likely to raise: earlier work making the same claim, work from a neighbouring field or under older terminology, and problems in the reference list such as retracted papers or preprints that have since been published. Before each check, the papers that cite the same sources are added, because that is where the missed paper usually is.

PhD students and postdocs in climate and earth science who are about to submit a journal paper. Later, the same check could run on every submission a journal receives.

Does Unkwn42 reject ideas that overlap with prior work?

No. It is not a pass or fail uniqueness test. Replication, extension, and doing a known study in a new place are legitimate research. The verdict has three states: Claimed (a specific paper already asserts the claim), Adjacent (the nearest work is distinguishable, with the delta stated), and Open (no paper unkwn42 holds asserts it). Even a Claimed result is information for a decision: what to match if replicating, or what to differentiate against if extending.

Will Unkwn42 invent or hallucinate citations?

No. Every paper cited is a real record retrieved from the literature, not generated. If the system cannot point to a specific paper, it does not make the claim. Grounding is verified in code before anything reaches you, and every result links to the source so you can check it yourself.

What do I need to give it?

Your abstract and your reference list, as pasted text, a .bib or .ris file, or a list of DOIs, and an email address for the sign-in link. The check takes about 10 minutes and emails you when it is ready.

Why not use a general AI research assistant or Google Scholar?

General research agents answer across all of science; Unkwn42 does one question in one field and goes deep. Keyword search also fails silently when you do not know the words: a paper that already did your work can sit fully indexed under a term your subfield stopped using, and no search you would think to run will surface it. Unkwn42 expands your idea across renamed and drifted terminology to retrieve exactly that prior work.

Your first check is free, one per email address: the answer on your main claim in full, how many earlier papers you are missing with the closest one in full, and how many reference problems were found. The full check is $19 per paper, with re-checks for 30 days. See https://www.unkwn42.com/pricing.

Retracted papers don't disappear from your search

Name: Unkwn42
Author: Unkwn42

1,369 of the 438,288 papers we hold carry a retraction, withdrawal, concern or correction notice. What that means for a literature search, and what to check.

Rohit Pawar · 22 October 2026 · 3 min read · Updated 23 October 2026

Key findings

Of the 438,288 climate and earth science papers in the unkwn42 corpus, 1,369 carry a notice: 87 are retracted, 2 withdrawn, 3 under an expression of concern and 1,267 corrected.
10 more records in the corpus are not papers at all but retraction notices that publishers filed as articles, so a search can return one as if it were prior work.
In biomedicine, only 722 (5.4%) of 13,252 citation contexts written after a paper's retraction mentioned the retraction (Hsiao and Schneider, 2021).

This week we checked every paper unkwn42 holds against the public record of retractions and corrections. It's the first time we've done it, and it's the kind of thing a tool that shows you prior work should have been doing from the start.

What we found

Of the 438,288 papers in our corpus, 1,369 carry at least one notice. Counting each paper once, by its most serious notice:

87 have been retracted.
2 have been withdrawn.
3 are under an expression of concern.
1,267 have been corrected.

There were also 10 records that aren't papers at all. They're retraction notices that publishers filed as if they were articles, with titles like "Retraction notice to..." or "Retraction Note:". A search that matches on words can return one of those as a piece of prior work.

These categories mean quite different things. A correction is the journal fixing something, often small, and it's a sign the record is working, not a mark against the paper. An expression of concern means the journal has questions about a paper but hasn't retracted it. A retraction means the journal no longer stands behind the paper, and a withdrawal is much the same, usually happening around publication. The 87 retracted papers are about two in every ten thousand that we hold. That's a small share, but it only takes one in your reference list.

Why a search won't tell you

A retracted paper doesn't leave the literature. It stays in databases, it keeps its citations, and it keeps collecting new ones. The best large study I know of comes from biomedicine. Hsiao and Schneider looked at 7,813 retracted papers indexed in PubMed and found they kept being cited after retraction. Of 13,252 places where a later paper cited one of them after the retraction, only 722, or 5.4%, mentioned that it had been retracted.

Researchers do care about this. But a citation can come from another paper's reference list, or from your own notes from a year ago, and nothing along the way tells you the status has changed. Unless the tool in front of you checks, the only way to know is to open each paper and look for a notice.

What you can check today

You don't need us for most of this.

The Retraction Watch database, the best-known list of retractions, was acquired by Crossref and made openly available in September 2023. Any tool can now check a reference list against it.

If you use Zotero, it already does that for your library. Since 2019 it has flagged retracted items using Retraction Watch's data and warns you if you try to cite one. When it launched, it worked for items with a DOI or a PubMed ID.

For the handful of papers your argument really rests on, open the publisher's page and look for a notice. Many publishers show a Crossmark button that lists any corrections or retractions issued since publication.

What we've done about it

Since 23 October, every paper the Terminal and Reviewer 2 show you carries its status, saved results included. A retracted paper stays visible, because a referee might still raise it, but it carries a clear label and a link to the notice. If a verdict rests on a retracted paper, the result says so right next to the verdict. The 10 notices filed as articles no longer come up as prior work, and the private reports we send to authors leave retracted and withdrawn papers out altogether. We check the whole corpus against the record again every week.

If a paper unkwn42 shows you turns out to be retracted and isn't marked, write to hello@unkwn42.com and I'll look at it.

How we counted

We matched papers by DOI against Crossref's records of retraction, withdrawal, concern and correction notices, which include the Retraction Watch database, and against OpenAlex's retraction flag. Where the sources disagreed, we counted the most serious status any of them gave. Counts are for our corpus of climate and earth science papers, not the whole literature, as of 22 October 2026.

Sources

Hsiao TK, Schneider J (2021). Continued use of retracted papers: Temporal trends in citations and (lack of) awareness of retractions shown in citation contexts in biomedicine. Quantitative Science Studies 2(4): 1144-1169. doi.org/10.1162/qss_a_00155
Crossref (2023). News: Crossref and Retraction Watch.
Zotero (2019). Retracted item notifications with Retraction Watch integration.