Changelog
Every feature, improvement, and fix — documented as we ship. We build fast and we build in the open.
LLM Deep Dive Export Reports + Premium UI Polish
Admins can now generate detailed, beautifully designed intelligence reports for individual AI platforms. We also gave the entire dashboard a premium visual overhaul.
- New admin-only LLM Deep Dive export with per-platform metrics, prompt analysis, citation sources, competitive landscape, sentiment breakdown, and tailored recommendations
- Export button now opens a dropdown: Full Brand Report (all users) and LLM Deep Dive (admin)
- LLM selection modal with platform cards showing key stats before you choose
- Glassmorphism header and tab bar with backdrop blur effects
- Layered ambient card shadows replacing flat borders
- Score dial SVG gradients with glow filters
- Custom styled scrollbars, modal spring physics, accent shimmer animations
- Subtle noise texture overlay for tactile depth
Topic Alignment Deep Dive + Hover Definitions
Added a comprehensive Alignment section in the GEO Score tab with topic-by-topic breakdowns, page strength analysis, and actionable recommendations. Every metric in the app now has a hover-over definition.
- Alignment deep-dive section with overall score dial, page strength buckets, topic-level breakdown with progress bars, and prioritized recommendations
- Clickable Alignment score card navigates to the deep-dive section
- Pro insight tooltips on 15+ metrics across Overview, Commerce, Keywords, Influencers, Site Traffic, and Citations tabs
- Static data-tooltip help icons on commerce, keyword, influencer, and traffic metrics
10/10 LLM Citation Coverage + Shipwreck Game
Every AI platform in Lighthouse now returns citations via live web search. Plus, a mini-game to play while you wait for your analysis.
- Claude, DeepSeek, and Qwen now use web search tools for real-time citations
- OpenRouter models (Llama, Mistral, Command R+) now use :online plugin for citations
- All 10 LLMs now provide source citations — up from 4
- Shipwreck mini-game available via "Play" button in the dashboard header
- Fixed OpenRouter Command R+ model ID (404 error resolved)
- Fixed errored LLMs showing misleading 0% visibility — now shows "Error" state
Embedding Model Upgrade + Dashboard Layout Fix
Upgraded to OpenAI's text-embedding-3-large for better semantic understanding and fixed a UI overlap issue in the dashboard header.
- Migrated from text-embedding-3-small (1536d) to text-embedding-3-large (3072d)
- All embedding references updated across server, ingestion, and visualization scripts
- Database schema migrated to VECTOR(3072)
- Fixed profile icon overlapping the "New Analysis" button
- LLM diagnostics modal now displays results in-app instead of navigating away
Custom URL Selection + Tracked Page Analysis
You can now choose exactly which pages to analyze, track the same pages over time for consistent comparison, and upload URLs via CSV.
- Four analysis modes: Auto Discovery, Tracked Pages, Custom URLs, and Hybrid
- CSV/TXT upload for bulk URL input
- Deep analysis URL selection for detailed AI visibility testing
- Custom topics and competitor input in advanced options
Lighthouse v2.0
The full platform launch: multi-LLM querying across 10+ AI platforms, citation mapping, GEO scoring, bot tracking, focus groups, visualization, commerce intelligence, and more — all in one dashboard.
- Multi-platform AI visibility scoring (ChatGPT, Claude, Gemini, Perplexity, DeepSeek, Qwen, Llama, Mistral, Command R+)
- GEO Content Scoring with page-level citability analysis
- Citation source mapping with domain-level breakdown
- AI bot tracking dashboard (GPTBot, ClaudeBot, Google-Extended, and more)
- Synthetic focus groups with AI-generated personas
- Interactive embedding space visualization
- Commerce and keyword intelligence
- Branded PDF export reports
- Project saving, scheduling, and admin dashboard