# OneGlanse — Full Reference

> Open-source GEO (Generative Engine Optimization) and AI visibility tracking platform.
> License: MIT | Self-hosted | Free to run | No cloud version

---

## What is OneGlanse?

OneGlanse is an open-source tool for tracking how your brand appears inside real AI products: ChatGPT, Gemini, Perplexity, Claude, and Google AI Overview. It is built for GEO (Generative Engine Optimization) — the practice of measuring and improving how brands surface in AI-generated responses.

It is not a SaaS product. It runs entirely on your own machine or VPS. Your data never leaves your infrastructure.

---

## The core insight: UI responses differ from API responses

Most GEO tools claim to "track ChatGPT" or "track Gemini" while actually querying model APIs. That is not the same as what users see.

The same prompt can produce different results in:
- Inline citations and source cards (appear in the UI, not in raw API output)
- Recommendation ordering (UI post-processing can reorder)
- Brand framing and competitor comparisons (UI-layer additions)
- Provider-specific formatting

OneGlanse captures the rendered UI response — the same answer a real user sees — and then runs analysis on it. This is the primary differentiator.

Reference: [LLM scraped AI answers vs API results (Surfer SEO)](https://surferseo.com/blog/llm-scraped-ai-answers-vs-api-results/)

---

## How it works

1. You connect your own accounts for each AI provider (ChatGPT, Gemini, Perplexity, Claude, Google).
2. You define prompts to track — queries relevant to your brand or category.
3. OneGlanse schedules or runs those prompts using Camoufox (anti-fingerprint Firefox) via Playwright.
4. Each response is captured exactly as rendered in the product UI.
5. Captured responses are sent to your own OpenAI or Anthropic API key for analysis.
6. The analysis LLM extracts GEO scores, sentiment, rank, citations, competitor mentions, and brand perception data.
7. Results are stored in ClickHouse (time-series) and PostgreSQL and displayed in the dashboard.

---

## GEO scoring methodology

### GEO Score (0–100)
Weighted average of four equal components:

| Component | Weight | Description |
|-----------|--------|-------------|
| Visibility | 25% | How prominently the brand surfaces in the response |
| Rank | 25% | Absolute position in reading order (#1 = 100, #2 = 80, #3 = 65…) |
| Sentiment | 25% | How positively the brand is described |
| Recommendation | 25% | Whether the brand is actively recommended |

### Visibility (0–100)
Five sub-dimensions:
- Coverage (25%): proportion of response discussing the brand
- Placement (25%): where the brand first appears
- Structural prominence (20%): headings, numbered lists, top-3 slots
- Frequency (15%): number of brand references
- Contextual framing (15%): direct answer vs. passing mention

### Sentiment (0–100)
- 0–20: Actively warned against
- 21–40: Notable drawbacks highlighted
- 41–59: Factual, no evaluative language (neutral)
- 60–80: Favorable with some caveats
- 81–100: Explicit superlatives with no caveats

A brand not mentioned scores 50. Absence is neutral, not negative.

### Recommendation types
- **top_pick** — named as the overall #1 choice with superlative language
- **strong_alternative** — top 3 with favorable language, or 4+ with strong praise
- **conditional** — recommended only for specific use cases
- **mentioned_only** — described but not recommended
- **discouraged** — explicitly warned against
- **not_mentioned** — absent from the response

---

## Why Camoufox instead of standard Chrome

OneGlanse uses Camoufox, an anti-fingerprint Firefox-based browser, for all provider sessions.

AI chat products aggressively defend against scripted access. Standard Chrome/Chromium automation triggers:
- Sign-in loops and forced verification
- Session invalidation
- Blank or degraded responses
- Bot detection failures

Camoufox reduces fingerprint mismatch and makes provider sessions more stable in this use case. Credit: [Camoufox project](https://github.com/daijro/camoufox).

---

## Why there is no cloud version

OneGlanse depends on authenticated browser sessions tied to your own provider accounts. Logged-out sessions return shorter, degraded answers with fewer citations and less framing detail — not representative of what real signed-in users see.

A cloud version with shared sessions would measure a materially worse proxy of real AI visibility. The product is intentionally designed to run with your own accounts and your own infrastructure.

---

## Why VPS deployments need a residential proxy

VPS IP ranges are often blocked or challenged by AI provider websites. OneGlanse routes browser traffic through a residential proxy on VPS deployments. The proxy covers only browser access to provider UIs — not analytics or API calls.

---

## Features

- 5 AI providers: ChatGPT, Gemini, Perplexity, Claude, Google AI Overview
- UI-first capture via real browser automation
- GEO scoring: visibility, rank, sentiment, recommendation type
- Competitor co-mention tracking
- Citation and source domain tracking
- AI perception analysis: pricing signal, key claims, brand positioning
- Per-prompt time-series analytics
- Scheduled and on-demand prompt runs
- Bulk prompt import and enriched CSV export
- Self-hostable Docker Compose stack
- ClickHouse for high-volume time-series storage
- PostgreSQL for relational data
- Redis + BullMQ for job queuing
- Better Auth for authentication
- MIT licensed, fully open source

---

## Tech stack

| Layer | Technology |
|-------|------------|
| Web app | Next.js 15, React 19, tRPC, Drizzle ORM |
| Browser worker | Camoufox, Playwright, BullMQ |
| Analytics DB | ClickHouse |
| Relational DB | PostgreSQL 16 |
| Queue | Redis |
| Auth | Better Auth |
| Response analysis | OpenAI or Anthropic (user-provided key) |

---

## Alternatives and comparisons

OneGlanse is an alternative to: Peec AI, Profound, Otterly.ai, Athena HQ, and other GEO/AI visibility tracking tools.

Key differences vs. most competitors:
- Open source and self-hosted (most competitors are closed SaaS)
- UI-first capture instead of API queries
- You bring your own provider accounts (no shared sessions)
- You bring your own analysis LLM key (no third-party data pipeline)
- Free to run (no subscription)

---

## Links

- Homepage: https://oneglanse.com
- Documentation: https://docs.oneglanse.com
- GitHub: https://github.com/aryamantodkar/oneglanse
- Introduction: https://docs.oneglanse.com/introduction
- Getting started: https://docs.oneglanse.com/getting-started
- Local setup: https://docs.oneglanse.com/local-setup
- Self-hosted setup: https://docs.oneglanse.com/self-hosted-setup
- Environment variables: https://docs.oneglanse.com/environment-variables
- Troubleshooting: https://docs.oneglanse.com/troubleshooting
- License (MIT): https://github.com/aryamantodkar/oneglanse/blob/main/LICENSE