Built for academics who value their time and their reputation

Reference checking is tedious but vital drudge work. The best advice is to be diligent and use an established citation management tool like Endnote / Zotero. But I know that's like telling someone to just eat their 5 portions of veg every day. This app is for those who just want to write in a flow state, add references (potential or used) as they go and then clean up later. Biblibuddy makes the clean up part lightning fast and painless.

Minimal AI, Maximum Rigour: How Biblibuddy Actually Works

Many academics assume that AI chatbots should be able to handle referencing checks in a single prompt, because it's "easy" and something an intern could do. But that assumption fundamentally misunderstands both how large language models work and the complexity of rigorous reference validation. This is not a single task. It's hundreds of micro-tasks, each of which has to be checked off systematically. No AI chatbot can do this reliably, as anybody who has tried knows. Asking a chatbot to one-shot this is like asking an intern to read your paper start to finish without taking notes and then from memory give you a list of which references matched or didn't and which ones might have typos. Same goes for online validation, that's a huge number of tasks each of which has to be logged. To do this rigorously the intern would have to note down each individual step and log results for a final accurate review. That's exactly what Biblibuddy does, each AI task is a micro-task as part of a controlled, orchestrated workflow.

The Scholarly APIs used

Biblibuddy take a multi-pronged approach to validate your references by querying several authoritative scholarly databases and catalogues, with web search as a fallback or when a bibliography entry was classified by the AI as non-scholarly (e.g. govt / consultancy report, media article etc.):

• OpenAlex: A comprehensive scholarly index covering millions of works across disciplines.
• CrossRef: The authoritative DOI registration agency, used to verify DOI-based references quickly and reliably.
• Semantic Scholar: Research tool from the Allen Institute, used as a backup when OpenAlex is uncertain.
• arXiv: The preprint repository for physics, mathematics, computer science, and related fields.
• Google Books: Used to verify book references and ISBNs.
• Web search: Used as a fallback for hard-to-verify references, such as grey literature, reports, or references with year mismatches in scholarly databases. If a bibliography entry includes a URL, the app checks that first including comparing the content to ensure it's actually what the right link

The Human-in-the-Loop Philosophy

I'm an emphatic human-in-the-loop fundamentalist when it comes to AI. The idea of letting an AI automatically edit a live draft document behind the scenes is insane to me. Even if you plead for it, I won't add any such feature.

Biblibuddy quickly produces a report which you use to fix issues yourself, maintaining full editorial control.

This approach respects the reality that you are the expert on your work. Biblibuddy provides evidence and flags potential issues that will have emerged from careless slip-ups rather than academic judgement. The report is structured to make corrections as easy as possible, with clear issue descriptions, evidence links, and actionable recommendations. But the final decisions i.e. what to fix, what to ignore, what to investigate further etc. are always yours.

Designed to be conservative and cautious

Biblibuddy is intentionally conservative. When in doubt, it flags items as possible rather than verified. Sometimes web search results simply don't have all the info to identify a perfect match (e.g. the publication year) - this app never accesses scholarly content from publishers, it simply searches and evaluates citation databases and web search results for the referencing metadata to identify the best match.

This means:

• Some perfectly valid references may be flagged as "possible" if the validation evidence is ambiguous (e.g., a book review in OpenAlex instead of the book itself, or a large year mismatch) even if to your eyes it looks like a slam dunk match.
• Author-aware validation prevents false positives. Biblibuddy won't mark a reference as verified if the title matches but the author doesn't. This might be annoying in cases where an author was added to a future version, but it's always better to err on the side of caution.
• Near-miss suggestions are surfaced explicitly for you, rather than silently auto-corrected.

This cautious approach reflects the reality that false confidence is worse than flagged uncertainty. If Biblibuddy isn't sure, it tells you, and you can then investigate and decide.