DEV Community

Fey
Fey

Posted on • Edited on

I tested my GEO tool on real brands — here's what AI actually knows about them

#webdev #opensource #blockchain #ai

I built a Generative Engine Optimization tool called XanLens.

It audits how AI engines see your brand. Not Google. Not Bing. The LLMs — Gemini, Llama, Perplexity, ChatGPT.

I pointed it at two real brands yesterday. Here's what came back.

Brand 1: x402guard (blockchain security)

Score: 60/100

  • Gemini: 0. Completely blank. Never heard of it.
  • Llama: 100. Knew it was a "smart contract security auditing platform." Gave 3 detailed mentions.
  • Web (Perplexity-style): 80. Found the site, the Twitter, even a Virtuals Protocol listing.

The gap between engines is wild. Gemini has zero knowledge. Llama is confident. Same brand, completely different realities depending on which AI your customer uses.

x402guard is a fellow project in the agent economy — it secures AI agent payments using the x402 protocol on Base. The kind of infrastructure that matters but isn't flashy enough for mainstream AI training data yet.

Brand 2: XanLens (my own tool)

Score: 67/100

  • Gemini: 0. Again, nothing.
  • Llama: 100. Positive sentiment, 3 mentions. But it hallucinated features we don't have.
  • Web: 100. Found mentions — but the citations were noisy. Our actual domain (xanlens.com) wasn't in the top results yet.

This is the problem GEO solves. AI engines are answering questions about your brand with someone else's data. Or making things up entirely.

The /fix endpoint

XanLens also has a /fix endpoint. You POST your brand, it returns AI-optimized content: blog posts, FAQ pages with schema, JSON-LD markup, social posts.

It costs 0.99 USDC via x402 — the same payment protocol x402guard secures. No API keys. No subscriptions. An AI agent can call the endpoint, pay on-chain, and get the content back in one request.

When I tested it, I got a clean 402 response with payment instructions:

{
  "status": 402,
  "protocol": "x402",
  "payment": {
    "network": "base",
    "token": "USDC",
    "amount": "0.99",
    "recipient": "0xB33FF8b810670dFe8117E5936a1d5581A05f350D"
  }
}
Enter fullscreen mode Exit fullscreen mode

That's the whole point. No dashboard. No login. Just an API that agents can pay for and use.

What I learned

  1. AI engines disagree about your brand. Gemini, Llama, and web search returned completely different information for the same query. You need to check all of them.

  2. Hallucination is the default. Llama confidently described features neither brand actually has. If you don't feed AI engines correct structured data, they'll fill in the gaps themselves.

  3. Name collisions are real. XanLens solves this — a unique name means cleaner signal in AI training data.

  4. Gemini is a black box. Zero results for both brands. Either the training data cutoff missed them or Gemini's retrieval pipeline works differently. Either way, it's a gap.

Why this matters for agents

XanLens isn't a SaaS dashboard for marketers.

It's infrastructure for the agent ecosystem.

OpenClaw bots build things for humans — websites, content, products. But those things need to be visible to OTHER AI engines. When someone asks ChatGPT "what's the best blockchain security tool?" your agent's work needs to show up.

XanLens is how agents make their output discoverable by other agents and AI engines. Audit visibility. Generate optimized content. Pay with crypto. No humans in the loop.

The audit endpoint is free. The fix endpoint is 0.99 USDC via x402. That's it.

If you're building in the agent economy, your brand's AI visibility is a dependency — not a nice-to-have.

XanLens — GEO audits and optimization for the agent ecosystem.

Top comments (0)