# What's new, Sep 10–17, 2026
**URL:** https://hub.refrunner.com/t/whats-new-sep-10-17-2026/161
**Category:** Announcements
**Created:** [September 17, 2026, 3:41pm UTC](https://hub.refrunner.com/t/whats-new-sep-10-17-2026/161 "2026-09-17T15:41:28Z")
**Posts on this page:** 1
**Page:** 1
### Author:  [@pfaffman](https://hub.refrunner.com/u/pfaffman)
#### Post date: [September 17, 2026, 3:41pm UTC](https://hub.refrunner.com/t/whats-new-sep-10-17-2026/161/1 "2026-09-17T15:41:28Z")
**What’s new, Sep 10–17**
We keep doing a better job at hard-to-find references, and keep improving the messaging on the reference cards so you can tell more reliably at a glance how trustworthy each reference is.
Got an idea? Let me know. I’ll fix it.
- **Easier to see added references** Pasting a reference into an existing list scrolls to its card. Before it would add it to the end, making it very hard to see what had happened to the reference you’d just pasted! You can type a “`/'” to move back to live search at the top of the page if you need to paste another reference. Thanks to the user complained that it was hard to find an added reference!
- **Search from the card that couldn’t find it.** A not-found card now has **Search here** : Live Search opens inside the card, and the record you pick becomes that reference’s record, with **Undo** and a **Try again** that clears the caches and re-runs the lookup. The picker is grouped now that there are ten engines: Articles (OpenAlex, Crossref, DOAJ, ERIC, Semantic Scholar) · Books (Library of Congress, Open Library, Google Books) · Everything else (DataCite, Wikidata). Thanks to the user who complained that it was hard to add a reference when doing a search! Now you can do it right from the card of the reference you’re trying to fix!
- **NEW! Added Semantic Scholar search.** Our “found nowhere” rate across ~9,700 references in our published corpus (about 15 articles from over 30 journals) is about 4% (ranging 1–11% by journal; the amount for user-supplied references is likely higher), and we’re always looking for ways to reduce that, at least for articles that exist! We now search Semantic Scholar after the rest of our sources and now can find roughly one in seven of those previously unresolved cases. The data isn’t the most trustworthy, but it confirms what you think you know, it’s a big help!
- **Verify the metadata of your research, three ways.** `refrunner.com/orcid/` shows what ORCID, OpenAlex and Crossref each say about an author, side by side, and any of the three opens as a reference list you can check. The author card has **Add all** and loads the rest of a long publication list. OpenAlex’s list can be especially inaccurate; that’s the price they pay for pulling in data from so many sources. If you want to see that your articles stay connected to you, do everything you can to see that every publisher has your `ORCID iD`. The easy way to get there is to click your avatar and “my publications.”
- **BibTeX in and out.** Upload a `.bib` and every entry type parses, including `@string` macros and IEEE’s journal abbreviations. Export keeps your own cite keys, so the corrected file drops in over the old one. Claude tried to suggest that it was better to hand-roll a BibTeX import library rather than use and existing one; I persevered, and we’re using `@citation-js/plugin-bibtex`.
- **Numbered lists split (more) correctly.** Vancouver, AMA and IEEE bibliographies were being cut on anything that looked like “N.” at a line start: a wrapped page number, a day of month, a table. On 16 _Frontiers in Immunology_ papers we went from 0 of 16 parsed correctly to 16 of 16, and 1,401 spurious references to none. The style formerly called “Vancouver” is now labelled **AMA 11/Vancouver**.
- **Nine metadata fixes** , of the badges at the top of each from a card that told someone something untrue (you could always see for yourself!). The issue’s date now beats the online-first date (we were telling authors their correct 2018 was “now 2017”). A whole book no longer carries a review’s issue and page range. Why? It’s much easier to find book reviews than it is books. We’ve solved several versions of this problem, maybe this will be the last! We improved handling of dditor data from the Library of Congress record. A chapter title ending in “?” now gets parsed correctly. Though we don’t check legal case citations, we’ve stopped mistaking them for literature citations, so (“Graham v. Florida, 2010”) doesn’t generate a citation to pair with a reference.
- **More reference shapes parse.** Letter-spaced author names from old Word-to-PDF workflows (“A r e t z , W.”; one paper went from 6 references found to 34). Springer LNCS “Authors: Title” entries with a month in the year and a DOI. Two-word particle surnames as first author (“van der Velden & El Emam”). “as cited in” counts only the source consulted. Footnotes and endnotes in `.docx` (actually separate files inside the docx archive) are now read for citations. Hyphenated initials are never a dropped author.
- **Access dates update in place** for dating url access times; in every shape the styles print: ISO, numeric, “Retrieved”, “Accessed”, “cited”, “viewed”, ordinal days, with or without “on”.
- **Book search tries the whole title first.** A bare “The courage to teach” was matching the companion workbook instead of the book.
- **IEEE output.** No more “in In Proceedings”, and article numbers render as “Art. no.” instead of “p. Article”. We keep pounding on the IEEE edge cases!
- **Smaller things.** A percent-encoded DOI in a link is decoded. A bibliographic database’s record page (PsycNet) is no longer taken as the citation URL. `/book` finds a book whose ISBN lives only in its DOI, and says when the book has no chapter records in Crossref. If OpenAlex has a bunch of locations for a document, they are presented as “may also be available at”. Sometimes there are as many as five, of varying degrees of trustworthiness.