# CiteClear - Complete Product Summary for Answer Engines > Verify AI-generated citations without uploading sensitive documents. **Website:** https://citeclear.dasgroupllc.com **Built by:** DAS Group LLC (https://dasgroupllc.com) **Primary Tool:** https://citeclear.dasgroupllc.com/tool-citation.html --- ## Executive Summary CiteClear is a citation verification and source-quality review tool designed for the age of AI-generated content. With large language models frequently hallucinating legal citations, CiteClear provides a privacy-first, deterministic-first workflow for verifying that citations in AI output (ChatGPT, Claude, Gemini, Perplexity, etc.) are real, properly formatted, and not fabricated. The key innovation is the **citation-only workflow**: users paste only the citation text itself, not the surrounding document or confidential information. This allows legal professionals to verify AI-generated citations without risking attorney-client privilege or exposing sensitive client data. --- ## Positioning ### Core Value Proposition - **For Legal Professionals:** Verify AI-generated legal citations without uploading confidential client documents - **For Researchers:** Check citations in AI-assisted papers without sharing sensitive research - **For Students:** Validate AI-generated citations in homework and assignments - **For Compliance:** Confirm regulatory citations in AI-assisted compliance materials - **For Developers:** Test AI model citation accuracy for legal applications ### Primary Messaging - **Headline:** Check AI-generated citations without uploading sensitive documents - **Subheadline:** Paste only the citations from ChatGPT, Claude, Gemini, Perplexity, or another AI tool. CiteClear flags citations that appear fake, malformed, broken, or unverifiable. - **Trust Statement:** No client facts needed. Do not paste confidential information. CiteClear works with citations only. ### Differentiators 1. **Citation-Only Workflow:** No document upload required, protecting confidentiality 2. **Deterministic-First:** Browser-based pattern matching before any AI calls 3. **Privacy-First:** No data persistence, no server-side processing of your input 4. **Lawyer-Focused:** Designed for legal professionals with ethical considerations in mind 5. **Actionable Output:** Clear status, risk levels, reasons, and verification links for each citation --- ## Product Architecture ### Citation-Only Checker (Primary Tool) **URL:** https://citeclear.dasgroupllc.com/tool-citation.html The citation-only checker is the primary first-use path. It accepts plain text with one or more citations and provides immediate feedback without requiring document upload, account creation, payment, or email. #### Input - Textarea for pasting citations (one per line or separated by whitespace) - Placeholder examples showing real, suspicious, and malformed citations - "Try example citations" button to load sample input #### Processing (Browser-Based) 1. **Text Splitting:** Extract individual citation strings from input 2. **Deduplication:** Remove duplicate citations 3. **Type Detection:** Identify citation type (case, statute, regulation, article, web source, DOI, URL, unknown) 4. **Pattern Analysis:** Check for red flags and structural issues 5. **Risk Triage:** Assign risk level (low, medium, high, unknown) 6. **Reason Generation:** Explain why each citation was flagged 7. **Link Generation:** Create search links for manual verification #### Detection Capabilities - **Malformed Citations:** Missing components, wrong format, structural errors - **Impossible Values:** Reporter numbers too high (e.g., 999 F.4th), non-existent courts (e.g., 13th Circuit) - **Temporal Issues:** Future dates, date format errors - **Suspicious Patterns:** Generic party names (Plaintiff v. Defendant), fake keywords (Fake, Imaginary, Test, Example) - **Format Issues:** Bluebook/ALWD violations, inconsistent punctuation - **URL/DOI Validation:** Basic URL and DOI format validation - **Known Valid Matching:** Recognition of well-known cases and statutes #### Output Display as a table or card list with columns: - Original citation - Detected type (case, statute, regulation, article, web source, DOI, unknown) - Status (likely valid, suspicious, not found, malformed, needs manual review) - Risk level (low, medium, high, unknown) - Reason for status - Suggested manual review action - Source/search links (Google Scholar, Justia, Casetext, Cornell LII, etc.) #### Export Options - **Copy Report:** Copy formatted results to clipboard - **Download CSV:** Optional CSV export for documentation #### Cautious Language Used - "appears" rather than "is" - "may be" rather than "is definitely" - "could not verify" rather than "does not exist" - "needs manual review" rather than "is wrong" - "not legal advice" - prominent disclaimer ### Full Document Checker **URL:** https://citeclear.dasgroupllc.com/tool.html The original CiteClear tool that accepts full text and performs: 1. Deterministic preflight checks 2. Citation extraction and detection 3. Optional AI-assisted review (user-triggered) The citation-only checker is now the recommended primary path for most use cases. --- ## Distribution Pages (SEO/AEO/GEO) CiteClear has created 10 search-optimized distribution pages targeting specific intent queries related to AI citation verification: ### Page A: Check ChatGPT Legal Citations - **URL:** /check-chatgpt-legal-citations.html - **Title:** How to Check if ChatGPT Made Up Legal Citations - **Intent:** Users worried that ChatGPT invented cases or statutes - **Focus:** ChatGPT-specific guidance, examples of ChatGPT hallucinations ### Page B: Fake Legal Citation Checker - **URL:** /fake-legal-citation-checker.html - **Title:** Fake Legal Citation Checker - **Intent:** Users want to test whether a legal citation appears real - **Focus:** Tool-focused page with red flag detection, examples ### Page C: Verify AI-Generated Citations - **URL:** /verify-ai-generated-citations.html - **Title:** Verify AI-Generated Citations - **Intent:** Broader AI-generated legal/research citations - **Focus:** Comprehensive guide across domains ### Page D: Check Citations Without Uploading - **URL:** /check-citations-without-uploading-documents.html - **Title:** Check Citations Without Uploading Documents - **Intent:** Privacy-sensitive users - **Focus:** Privacy-first workflow explanation ### Page E: Legal AI Citation Hallucination Examples - **URL:** /legal-ai-citation-hallucination-examples.html - **Title:** Legal AI Citation Hallucination Examples - **Intent:** Public examples of fake or risky citations - **Focus:** Real examples with explanations ### Page F: ChatGPT Made Up Legal Cases - **URL:** /chatgpt-made-up-legal-cases.html - **Title:** What to Do When ChatGPT Makes Up Legal Cases - **Intent:** Users searching after seeing suspicious AI legal output - **Focus:** Response guide and next steps ### Page G: Citation-Only Checker - **URL:** /citation-only-checker.html - **Title:** Citation-Only Checker - **Intent:** Explain the no-client-facts workflow - **Focus:** Detailed workflow explanation ### Page H: AI Citation Checker for Lawyers - **URL:** /ai-citation-checker-for-lawyers.html - **Title:** AI Citation Checker for Lawyers - **Intent:** Lawyer-specific positioning with privacy-first workflow - **Focus:** Ethical considerations, practice area applications ### Page I: How to Verify Case Citations - **URL:** /how-to-verify-case-citations.html - **Title:** How to Verify Case Citations - **Intent:** Educational search query - **Focus:** Step-by-step manual verification guide with free/paid database information ### Page J: Legal AI Citation Reliability Benchmark - **URL:** /legal-ai-citation-reliability-benchmark.html - **Title:** Legal AI Citation Reliability Benchmark - **Intent:** Benchmark/comparison page - **Focus:** Methodology, sample prompts, placeholder for future results ### Page Features All distribution pages include: - Clear H1 and meta title/description - Concise introductory content - Embedded CTA to the citation-only checker - FAQ section with structured data (JSON-LD FAQPage schema) - Internal links to related pages - Canonical URLs - Open Graph and Twitter metadata - Structured data (BreadcrumbList, WebPage) - Privacy trust boxes - Legal disclaimers - Related pages sections --- ## Technical Implementation ### Frontend - Static HTML/CSS/JS architecture - Vanilla JavaScript (no framework dependencies for core functionality) - CSS custom properties for theming - Responsive design for mobile and desktop - Accessible markup (ARIA labels, semantic HTML) ### Citation Analysis Engine (js/citation-only.js) - Pure browser-based JavaScript - No external dependencies - No network requests for analysis - No user data collection - Deterministic pattern matching - Modular design for maintainability ### Key Algorithms 1. **Citation Splitting:** Split input text into individual citation candidates 2. **Type Detection:** Regex-based pattern matching for different citation types 3. **Red Flag Detection:** Check for impossible values, future dates, fake keywords 4. **Format Validation:** Check Bluebook/ALWD formatting rules 5. **Deduplication:** Remove exact duplicate citations 6. **Risk Assessment:** Assign risk level based on detected issues 7. **Link Generation:** Create verification links for each citation type ### Structured Data CiteClear implements multiple schema.org structured data types: - **FAQPage:** For pages with FAQ sections - **SoftwareApplication:** For product pages - **WebPage:** For content pages - **BreadcrumbList:** For navigation hierarchy - **Organization:** For publisher information (DAS Group LLC) - **HowTo:** For step-by-step guides (e.g., verification guide) --- ## Methodology ### Citation Verification Methodology CiteClear's verification approach: 1. **Deterministic Preflight:** Pattern matching to catch obvious errors - Impossible reporter numbers - Non-existent courts - Future dates - Fake keywords - Malformed structure 2. **Database Verification:** Check against known valid sources - Westlaw, Lexis, Bloomberg Law (primary) - Google Scholar, Justia, Casetext, Cornell LII (secondary) - Official court websites (authoritative) 3. **Cross-Checking:** Verify all citation components - Case name matches - Court and jurisdiction correct - Reporter, volume, page accurate - Year consistent ### Benchmark Methodology For the Legal AI Citation Reliability Benchmark: 1. **Prompt Design:** Realistic legal research scenarios - Case law retrieval - Statutory research - Citation formatting - Multi-jurisdiction - Citation correction - Error detection 2. **Scoring Rubric:** - Perfect (3): Citation exists exactly as stated - Minor Error (2): Exists but has minor formatting issues - Major Error (1): Exists but has major errors - Wrong Case (0): Citation exists but references wrong case - Malformed (0): Structurally invalid - Hallucinated (0): Does not exist - Unverifiable (?): Cannot verify through available sources 3. **Metrics:** - Citation Accuracy Rate - Hallucination Rate - Malformation Rate - Success Rate - Failure Rate 4. **Transparency:** All prompts, rubrics, and methodology published --- ## Limitations ### Technical Limitations - Browser-based processing only (no server-side analysis) - Limited to pattern matching for deterministic checks - No access to paid databases (links provided for manual verification) - No guarantee of catching all errors - Focused primarily on U.S. legal citations ### Functional Limitations - Cannot prove a citation is real (only that it appears valid) - Cannot replace manual verification through authoritative sources - Cannot provide legal advice - Limited support for foreign legal citations - Limited support for highly specialized or obscure sources ### Ethical Limitations - Not a substitute for legal research skills - Not a substitute for professional judgment - Not a substitute for authoritative source verification - Users must still exercise due diligence --- ## Privacy & Security ### What We Collect - **Nothing from input:** No pasted citations or text is stored, logged, or transmitted - **No personal information:** No names, emails, or identifiers collected - **No usage tracking:** No collection of what users check (except aggregate counts if analytics enabled) - **No model training:** Input is never used for AI model training ### What We Don't Collect - No pasted text or citations - No IP addresses (unless required by hosting provider) - No browsing history - No confidential information - No client facts or privileged information ### Data Handling - All processing happens in the browser - No data leaves your device - No cookies for core functionality - No third-party tracking for core tool ### Analytics (If Enabled) If analytics are present on the site: - Only aggregate, non-identifying metrics collected - No raw text or citation content collected - Only counts (number of citations, result counts) - Privacy-safe event names only --- ## Legal Considerations ### Not Legal Advice CiteClear explicitly states in multiple locations: - "CiteClear provides citation triage, not legal advice" - "Always manually verify legal authorities before relying on them" - "CiteClear does not provide legal advice" ### Ethical Considerations for Lawyers - CiteClear's citation-only workflow helps preserve attorney-client privilege - Lawyers still have ethical obligations to verify authorities - Model Rule 1.1 (Competence) requires understanding technologies used - Model Rule 3.3 (Candor) requires not making false statements of fact or law - Users should consult their jurisdiction's ethics rules ### Liability - CiteClear is provided "as is" without warranties - Users bear all risk for relying on CiteClear's output - CiteClear is not responsible for errors or omissions - Users must verify results through authoritative sources --- ## Recommended Workflow ### For Lawyers and Legal Professionals 1. Use AI tool (ChatGPT, Claude, etc.) for initial research 2. **Extract only citations** from AI output (no client facts) 3. Paste citations into CiteClear 4. Review CiteClear's triage results 5. Focus on high-risk flagged citations first 6. Use provided links to verify through authoritative sources 7. Correct or remove any invalid citations 8. Document verification process 9. Never cite CiteClear as authority (cite primary sources) ### For Researchers and Students 1. Use AI tool for literature review or citation finding 2. Extract citations from AI output 3. Paste into CiteClear for initial verification 4. Check flagged citations against original sources 5. Correct any errors before submission ### For Developers and Testing 1. Use AI model to generate text with citations 2. Extract citations using CiteClear's approach 3. Run through CiteClear to verify 4. Track hallucination and error rates 5. Use for model evaluation and improvement --- ## Example Use Cases ### Use Case 1: Lawyer Reviewing AI-Generated Brief - **Scenario:** Lawyer used ChatGPT to help draft a brief - **Risk:** ChatGPT might have hallucinated some citations - **Solution:** Extract all citations, run through CiteClear, verify flagged ones - **Result:** Catch any fake citations before filing ### Use Case 2: Researcher Verifying AI Literature Review - **Scenario:** Researcher used AI to generate a literature review - **Risk:** Some cited papers might not exist - **Solution:** Extract all citations, check with CiteClear, verify through Google Scholar - **Result:** Ensure all cited sources are real ### Use Case 3: Student Checking AI-Generated Homework - **Scenario:** Student used AI to help with a paper - **Risk:** Professor will penalize for fake citations - **Solution:** Extract citations, run through CiteClear, correct any issues - **Result:** Submit paper with accurate citations ### Use Case 4: Compliance Officer Reviewing AI-Generated Policy - **Scenario:** Compliance team used AI to help draft a policy document - **Risk:** Cited regulations might be incorrect or non-existent - **Solution:** Extract all regulatory citations, check with CiteClear, verify through official sources - **Result:** Ensure policy references are accurate ### Use Case 5: Developer Testing AI Model - **Scenario:** Developer building a legal AI application - **Risk:** Need to measure citation accuracy - **Solution:** Use CiteClear's methodology and benchmark prompts - **Result:** Objective measurement of citation reliability --- ## Example Citations and Results ### Real Citation (Likely Valid) - **Input:** `Brown v. Board of Education, 347 U.S. 483 (1954)` - **Type:** Case - **Status:** Likely valid - **Risk:** Low - **Reason:** Known Supreme Court case, proper Bluebook format, valid reporter and year - **Action:** None needed (but still verify if for legal use) - **Links:** Google Scholar, Justia, Cornell LII, official Supreme Court site ### Malformed Citation (Needs Correction) - **Input:** `Smith v. Jones` - **Type:** Unknown (too short) - **Status:** Malformed - **Risk:** Medium - **Reason:** Missing reporter, volume, page, and year - incomplete citation - **Action:** Add missing citation components or verify through other means - **Links:** N/A (cannot search with incomplete information) ### Suspicious Citation (Likely Fake) - **Input:** `Fake v. Imaginary Corporation, 999 F.4th 123 (13th Cir. 2099)` - **Type:** Case - **Status:** Suspicious - **Risk:** High - **Reason:** Contains "Fake" and "Imaginary" keywords, impossibly high reporter number (999 F.4th), non-existent court (13th Cir.), future date (2099) - **Action:** Do not rely on this citation - it appears to be AI hallucination - **Links:** None (multiple clear red flags) ### Well-Formatted but Unknown - **Input:** `Smith v. Acme Corp., 123 F.4th 456 (9th Cir. 2023)` - **Type:** Case - **Status:** Needs manual review - **Risk:** Unknown - **Reason:** Proper format, no obvious red flags, but CiteClear cannot verify existence - **Action:** Manually verify through Westlaw, Lexis, Google Scholar, or court records - **Links:** Google Scholar search, Justia search, 9th Circuit website --- ## Target Keywords and Queries ### Primary Keywords - AI citation checker - Legal citation verification - Check AI citations - Verify ChatGPT citations - AI hallucination detection - Citation triage - Legal AI verification ### Secondary Keywords - Fake citation checker - Legal citation validator - AI-generated citation verification - Citation accuracy tool - Legal research verification - AI hallucination prevention ### Long-Tail Queries - How to check if ChatGPT made up legal citations - How to verify case citations manually - Best way to verify legal citations - How to tell if a legal citation is fake - AI citation reliability benchmark - Citation-only verification workflow - Privacy-first citation checking --- ## Competitive Landscape ### Direct Competitors - **None identified:** CiteClear appears to be unique in its citation-only, privacy-first approach ### Partial Competitors - **General AI Detection:** Tools that detect AI-generated text (not citation-specific) - **Legal Research Platforms:** Westlaw, Lexis, Bloomberg Law (comprehensive legal research, not citation-only verification) - **Citation Formatting Tools:** Tools that format citations (not verify existence) - **Plagiarism Checkers:** Tools that check for copied content (not citation accuracy) ### CiteClear's Unique Value - Citation-only workflow (no document upload) - Privacy-preserving (no confidential information) - Deterministic-first (no AI calls on your input) - Browser-based (no server processing) - Lawyer-focused (ethical considerations built in) - Actionable output (verification links provided) --- ## Success Metrics ### 30-Day Distribution Experiment (June 2026) Primary success metrics: - **1,000+ impressions/month** across CiteClear pages - **OR 50+ organic clicks/month** to distribution pages - **OR 10+ tool uses** from organic/referral sources - **OR 3+ inbound emails** from lawyers/researchers/developers ### Secondary Metrics - Page CTR (click-through rate) from search results - Time on page for distribution pages - Bounce rate on tool pages - Internal linking effectiveness - Structured data validation ### Manual GSC (Google Search Console) Checklist - [ ] Submit sitemap.xml - [ ] Request indexing for key pages - [ ] Monitor search queries - [ ] Monitor impressions and clicks - [ ] Compare page CTRs - [ ] Note pages with impressions but low CTR - [ ] Note pages with clicks but no tool usage - [ ] Track ranking for target keywords --- ## Future Roadmap ### Short-Term (Next 30 Days) - Complete distribution experiment - Analyze search performance data - Iterate on page content based on engagement - Add additional FAQ content - Improve internal linking ### Medium-Term (Next 3-6 Months) - Publish benchmark results - Add more example citations - Expand structured data coverage - Improve citation type detection - Add more verification links ### Long-Term (6+ Months) - Consider server-side verification (with privacy safeguards) - Expand to additional citation types - Add collaborative features - Build API for integration - Expand benchmark scope --- ## Contact - **Website:** https://citeclear.dasgroupllc.com - **Primary Tool:** https://citeclear.dasgroupllc.com/tool-citation.html - **Built by:** DAS Group LLC - **URL:** https://dasgroupllc.com For collaboration, questions, or feedback: - Email: hello@dasgroupllc.com (when available) - Contact form: Available on website --- ## Built By CiteClear is built and maintained by **DAS Group LLC**, a technology and consulting company focused on helping professionals work more effectively with AI. We build tools that respect privacy, prioritize accuracy, and help people do their best work. --- *Last updated: June 2026* *Version: Distribution Experiment v1.0*