Framer AI AgentsDesign and publish professional sites with AI
HallucinationBuster
AI research tool where YOU choose which publishers to trust
10 followers
AI research tool where YOU choose which publishers to trust
10 followers
ChatGPT fabricates 20-56% of academic citations. HallucinationBuster takes a different approach: it searches real academic databases, lets you pick which publishers and years to include, then processes ONLY your approved papers. AI-powered topic clustering, relevance scoring, and RAG chat, all grounded in verified sources with clickable DOIs. You control the sources. The AI never sees a paper you didn't approve. Free tier available, no credit card needed.
Hey Product Hunt Family!
I'm Priyabrata, and I built HallucinationBuster because I got burned by fake citations during my research work.
I was using ChatGPT to help with literature reviews and kept finding references that looked perfect with correct formatting, plausible author names, real-sounding journal titles, except the papers didn't exist. When I checked the DOIs, they went nowhere.
This isn't a rare problem. A 2025 study from Deakin University found that 56% of ChatGPT's academic citations contain errors or are completely fabricated. Another study found a 91% hallucination rate for Google Gemini on research queries. Even Big Four consulting firms aren't immune. KPMG delivered a government report on research integrity to Australia's NHMRC that contained a fabricated citation. Similarly, Deloitte was caught using AI in a $290,000 report to help the Australian government crack down on welfare fraud, after a researcher flagged hallucinations including non-existent references and a fake court quote.
So I built something different. HallucinationBuster doesn't ask AI to remember papers. It:
→ Searches real academic databases with 200M+ papers
→ Collects 1,000+ papers on your topic
→ Lets you filter by publisher (IEEE, Springer, Nature, etc.) and year range
→ Only THEN does AI process your selected papers for clustering topics, scoring relevance, and powering a RAG chat
The key difference: you control the source universe before AI touches anything. No tool I've found (Elicit, Consensus, Perplexity) gives you this level of pre-processing source control.
It's free to try: 2 queries/month, 50 papers per query. No credit card needed.
I'd genuinely love your feedback. What would make this more useful? What's missing?
🔗 https://hallucinationbuster.com
Finally, a tool that offers source control! I have been looking for an alternative to Elicit’s automated pull for a while and HallucinationBuster hits the mark. What sparked the decision to give the user more oversight at the start of the funnel?
@loan_dinh Great question! In recent days, during my lit review of my research, I kept running into the same problem of AI tools that would pull papers from everywhere without letting me control the source quality. In academia, where a paper is published matters as much as what it says. Different fields trust different publishers, and I wanted that decision to stay with the researcher, not the algorithm. So I designed the pipeline to let users filter by publisher and year before AI processing begins. This way the AI only works with sources you've already vetted.
I've just tried Hallucination Buster today. The app extracted all real references. Wow, what a difference! I produce a lot of papers and spend hours on literature reviews. Does this tool integrate with reference managers like Zotero or Mendeley, or is it a standalone check?
@bliuworld Thank you so much for trying it out! Really glad it made a difference for your workflow. Right now it’s standalone, but Zotero and Mendeley integration is something I’ve been thinking about. Being able to export directly into your reference manager would make a lot of sense, especially for heavy paper producers like yourself. I’ll prioritise that on the roadmap. In the meantime, the DOI links make it easy to add papers manually to your reference manager. Thanks for the suggestion!
I work in legal research, and verifying citations is one of the most tedious aspects of preparing reports. The publisher-filtering feature is very useful, since different fields rely on different trusted publishers. Does your system also allow filtering by specific journals, or only by publisher?
@katherine_parker3 Thanks for the feedback! Currently it filters by publisher, not individual journals. But that’s a great suggestion, especially for fields like legal research where specific journals carry much more weight than others. I’ll add journal-level filtering to the roadmap. In the meantime, the publisher filter should still help narrow things down significantly. Would love to hear how it works for your legal research use case if you give it a try!
I had a colleague submit a grant proposal where ChatGPT had generated three completely fake citations. The reviewers caught it. Something like this would have prevented that entirely.
@jason_tan001 That’s exactly the kind of situation that motivated me to build this. Getting caught with fake citations in a grant proposal is not just embarrassing, it can seriously damage your credibility with reviewers. With HallucinationBuster, every citation maps to a real paper with a clickable DOI, so that situation simply can’t happen. Hope your colleague’s future proposals go smoother!
Well, if Perplexity and Gemini still hallucinate citations, what’s the secret sauce? Is this just another GPT wrapper or real-time RAG against 200M+ live papers? How do you stop the AI from getting "lost in the middle" during clustering?