Citation Cleaner

The company appears to be a solo project named "Citation Cleaner", self-described as a tool for cleaning AI-generated Markdown locally in the browser. The author states it removes provider-specific citation artifacts from AI outputs and provides an audit trail of applied rules, with a Chrome extension and public API. It was submitted to the OpenAI 2026 hackathon.

What changed: The project evolved from a basic cleanup tool into one that uses deterministic rule pipelines and fingerprinting to identify AI sources and apply provider-specific rules only when evidence supports them. It now includes an explainability layer (Citation Fingerprint) and supports multiple export formats.

The single most important open question: Is there any evidence of user adoption, revenue or traction beyond the author's own description? The project is presented as a hackathon submission with no indication of commercial use or market validation.

• The description states Citation Cleaner is a tool that "turns AI-generated Markdown into review-ready copy entirely in the browser."
• It claims to "protect code, removes deterministic citation and export artifacts, shows an inline diff, and reports every applied rule."
• A secondary component, Citation Fingerprint, is described as a "build week extension" that "detects concrete provider signatures, reports confidence and matching evidence, and only enables provider-specific rules when the fingerprint supports them."
• The project uses Next.js, React, TypeScript, Vitest, and Manifest V3 Chrome extension.
• It includes a public API that returns structured detection receipts.
• The tool is described as deterministic: "a deterministic rule pipeline" that "protects Markdown code, resolves the source fingerprint, applies rules allowed by source and intensity, restores code, and returns output plus an audit trail."

Not evidenced: No evidence of actual product usage, customer feedback, or commercial deployment beyond the author's own account.

• The author states the tool was inspired by "AI answers are easy to generate and surprisingly hard to publish" and that existing cleanup workflows either upload sensitive drafts or use broad regexes.
• It positions itself as a solution that "protects code, removes deterministic citation and export artifacts, shows an inline diff, and reports every applied rule."
• The evolution from basic cleanup to explainable automation is described: "Auto mode now requires evidence before applying provider-only rules" and "Every source decision exposes confidence and concrete matching signals."
• The tool is positioned as privacy-preserving ("entirely in the browser") while also offering a public API.
• It claims to support "nine languages" and that "the same detector powers the browser UI and public API."

Inference: The positioning appears to have evolved from a simple text cleanup utility into a more sophisticated, explainable automation tool with privacy features.

• The description states it is for users who "copy from ChatGPT, Claude, Gemini, Perplexity, or AI Overviews" and want to clean up AI-generated Markdown.
• It targets users who need "review-ready copy" and are concerned about "citation markers, hidden footnotes, source trails, tracking links, and provider-specific export artifacts."
• The tool is described as supporting "nine languages," suggesting a global audience.

Not evidenced: No evidence of specific customer segments, personas, or market research. No indication of whether the tool targets individual users, content creators, or enterprise clients.

• The description states there is a "public API" and that it returns "the same structured detection receipt."
• It mentions a "live app" and "Try Perplexity" functionality, suggesting a public-facing interface.
• No pricing information, monetization strategy, or business model details are provided.

Inference: The tool may be offered as a freemium or open-source utility with optional paid API access, but this is not stated.

• The project uses Next.js, React, TypeScript, Vitest, and Manifest V3 Chrome extension.
• It implements a "deterministic rule pipeline" that protects code, resolves source fingerprint, applies rules, restores code, and returns output plus audit trail.
• The tool is described as supporting "nine languages" and having an "audit trail" that is not English-only.
• It includes a browser extension (Citation Fingerprint) and a public API.
• The authors state they used Codex running GPT-5.6 to design and implement the fingerprint contract.

Not evidenced: No evidence of scalability, performance metrics, or production deployment details beyond the author's own account.

• The project was submitted to the OpenAI 2026 hackathon.
• It includes a "live app" and "Try Perplexity" functionality.
• The authors mention "user-contributed fingerprint fixtures," "signed rule-pack releases," and "downloadable cleanup receipt for editorial review systems" as future features, suggesting ongoing development.

Not evidenced: No evidence of user adoption, revenue, or market traction beyond the author's own description. No data on usage, retention, or customer feedback.

• The description mentions that existing cleanup workflows either upload sensitive drafts or use broad regexes.
• It positions itself as an alternative to tools that "silently remove real content."
• The tool is described as being able to distinguish between provider-specific and generic citation syntax.

Not evidenced: No evidence of direct competitors, market size, or competitive positioning beyond the author's own claims.

• The project is a solo effort (1 person team) with no indication of commercial traction.
• It is presented as a hackathon submission, suggesting it may not be fully mature for production use.
• No evidence of monetization strategy or revenue model.
• The tool claims to be "local" but also exposes a public API — this raises questions about data privacy and compliance.
• The project has no third-party validation or user feedback.

Inference: The lack of traction, revenue, or market validation is a key risk. The solo team size may limit scalability and long-term development.

1. What is the actual use case for this tool? Is it being used by individuals or organizations?
2. How does the tool handle edge cases in citation formats, especially across different AI providers?
3. What are the plans for monetization and long-term sustainability?
4. Are there any known limitations or blind spots in the fingerprinting logic?
5. How is the tool tested, and what quality gates are in place for updates or new features?