paipers.ai

Part 2 of our series on LLM hallucination and the future of credible AI writing

In our previous post, we explored one of the most consequential and underappreciated flaws in AI-generated content: fabricated references. From a Government health report citing studies that don't exist, to a policy document built on invented legal judgments, the evidence is damning. And as research presented at ISMPP 2026 demonstrated in quantitative terms, there are not edge cases. It is a near-universal feature of how large language models work.

To recap why: LLMs generate text by predicting what comes next based on statistical patterns, not by retrieving verified facts. When asked to support a claim with a citation, they will produce something that looks like a reference, with the correct format, plausible journal, realistic author names and titles, but which may not actually exist. They have no mechanism to know the difference. The result is authoritative-sounding fiction.

The question is: what do you do about it?

A Platform Built Around the Problem

PAIPERS was designed from the ground up with this flaw in mind. It is a cloud-based content generation platform that combines the speed and fluency of LLM-generated writing with a robust, proprietary reference system; so that every citation you publish is real, verifiable, and relevant.

The workflow is intuitive. Users build their documents section by section, generating AI-assisted text for each part, then editing it manually as needed. When it's time to add references, rather than trusting the model to conjure them up, PAIPERS connects you to a database of more than 47 million peer-reviewed articles. The finished document can be exported directly to Microsoft Word or Google Docs, ready for submission or publication.

SmartCite®: Where the Real Difference Is Made

The heart of the platform is SmartCite®, our proprietary reference discovery technology. Here's how it works in practice: you highlight a passage of text in your document, hit search, and it automatically finds real, published, peer-reviewed references that support what you've written. No prompting, no guesswork, no fabrication.

Crucially, SmartCite® gives you control. Each returned reference comes with a hyperlink to the original abstract so you can quickly assess its relevance before committing to it. If the initial results don't fully support the specific point you're making, you can refine and re-run the search until you have exactly what you need.

Under the hood, SmartCite® operates through a sophisticated multi-stage pipeline designed to maximise both accuracy and relevance. It begins with a search that understands the meaning of your text, not just its keywords. This is layered with a hybrid re-ranking process that also rewards precise terminology matching, particularly important in technical writing, where exact terms matter. If the initial search doesn't return enough high-quality results, the system progresses through additional intelligent search stages that systematically broaden and vary the query using domain-specific knowledge, drawing directly from authoritative external literature databases. The architecture makes hallucination structurally impossible; every result returned is a real, indexed, published paper. The pipeline is fast, deterministic, and built to give you the right references for your specific text, every time.

Why This Matters

The academic and professional publishing world is beginning to reckon seriously with the reference hallucination problem. Careers, policy decisions, and patient outcomes can rest on the integrity of a citation. A tool that lets you write with AI efficiency while guaranteeing the integrity of your references isn't a luxury, it's rapidly becoming a necessity.

PAIPERS is that tool.

Ready to write with confidence? Sign up and experience SmartCite® for yourself. Stop hoping your references are real. Start knowing they are.