See what AI engines actually say about your brand.

Most GEO tools score your page against heuristics — "does your content look citable?"

XanLens doesn't guess. It asks ChatGPT, Gemini, Grok, DeepSeek, Claude, Llama, and Qwen directly: "Do you know this brand?" Then shows you exactly what each one says.

Same brand. Seven different answers. That's the insight.

1. You enter a URL
2. XanLens auto-detects your brand, industry, competitors, and keywords by scraping your site
3. Generates 30+ real prompts — the kind your customers actually ask AI ("best CRM for startups", "alternatives to Salesforce")
4. Queries 7 AI engines simultaneously — not heuristics, real queries to real models
5. Scores you 0-100 — knowledge score, discoverability score, citation quality
6. Generates content fixes — specific, actionable changes to improve your AI visibility

Self-host (free): Clone the repo, add your own API keys, run it. Full control, no limits.

Don't want to self-host? xanlens.com — hosted version, pay per audit, zero setup.

git clone https://github.com/FayAndXan/xanlens.git
cd xanlens
cp .env.example .env.local
# Add your API keys (Gemini required, others optional)
npm install
npm run dev

Open localhost:3000 and audit your first brand.

Gemini is required — it powers both knowledge scoring and grounded search (discoverability), plus the LLM judge that verifies brand accuracy. Get a free key at ai.google.dev. Add other engines for the full multi-engine picture.

XanLens generates prompts in three categories:

• Category queries (65%) — "Best [your industry]", "How to [your use case]" — the queries your customers actually ask
• Competitor queries (30%) — "Alternatives to [competitor]", "[you] vs [competitor]" — comparison intent
• Brand queries (5%) — "What is [your brand]?" — baseline awareness check

Each engine response is scored by an LLM judge that verifies:

1. Is the response actually about YOUR brand (not a name collision)?
2. Does it accurately describe what you do?
3. Are you recommended, mentioned, or cited?

Final score combines: knowledge (k), discoverability (d), and citation quality (c).

Next.js 15 app
├── /api/v1/audit/run     — Combined detect + submit + execute
├── /api/v1/audit/status  — Poll results + get agent instructions
├── /api/v1/audit/fixes   — AI-generated content fixes
├── /app/lib/
│   ├── engine-config.ts  — All 7 engine configs (add your own here)
│   ├── prompts.ts        — Prompt generation from brand data
│   ├── auto-detect.ts    — Website scraping + brand detection
│   └── scoring.ts        — LLM judge + score computation
├── Turso/LibSQL          — Audit history + results storage
└── Upstash Redis         — Caching + rate limiting

# Start an audit
curl -X POST http://localhost:3000/api/v1/audit/run \
  -H "Content-Type: application/json" \
  -d '{"website": "yourbrand.com"}'

# Returns: { job_id, poll_url }

# Poll for results
curl http://localhost:3000/api/v1/audit/status?id=JOB_ID

# Returns: scores, engine responses, content fixes

XanLens has a 3-step agent flow:

1. POST /api/v1/audit/run — Score the brand
2. GET /api/v1/audit/status — Get results + agent_instructions
3. POST /api/v1/audit/fixes/push — Push drafted content fixes

Any AI agent can audit a brand, get actionable instructions, draft fixes, and push them back.

• Node.js 18+
• A Turso database (free tier works)
• An Upstash Redis instance (free tier works)
• At least one AI engine API key

• Framework: Next.js 15 (App Router)
• Database: Turso (LibSQL)
• Cache: Upstash Redis
• AI Engines: Gemini, GPT-4o, DeepSeek, Claude, Grok, Llama, Qwen
• Payments: Coinbase Smart Wallet (USDC on Base) — optional
• Hosting: Vercel (or any Node.js host)

• 63% of B2B software decisions now start with an AI query
• AI referral traffic grew 527% year-over-year (Previsible 2025)
• 28.3% of ChatGPT's most-cited pages have zero Google visibility
• Same brand, 46x citation gap — 0.59% on ChatGPT vs 27% on Grok
• GEO market: $886M today → $7.3B by 2031

Traditional SEO tools don't show this. XanLens does.

PRs welcome. The biggest impact areas:

1. Add new engines — extend engine-config.ts
2. Improve prompts — better category/competitor prompt templates
3. Better scoring — improve the LLM judge rubric
4. CLI tool — npx xanlens audit domain.com (not built yet)

MIT

Fay — building AI products in the era of AGI.

Part of the xanOS fleet.