AI search engines like ChatGPT, Gemini, Perplexity, and Claude are
changing how people discover brands. Instead of clicking through
Google results, users now ask AI — and the AI picks who to mention.
The question is: is your brand being mentioned?
Most tools to answer are built for marketers, not developers. So I built webscore-sdk — an npm package that lets you measure this programmatically.
What is GEO?
GEO (Generative Engine Optimization) is the practice of improving
your brand's visibility in AI-generated answers. Think of it like
SEO, but instead of ranking on Google, you're trying to get mentioned
when someone asks ChatGPT "what's the best tool for X?"
Two key metrics matter:
- Mention rate — % of relevant prompts where your brand appears
- Citation rate — % of prompts where your website is linked
Install
npm install webscore-sdk
Quickstart
import { WebScore } from 'webscore-sdk'
const client = new WebScore('your_api_key')
const result = await client.geo.scan('https://yoursite.com')
console.log(result.mentionRate) // e.g. 67
console.log(result.citationRate) // e.g. 42
console.log(result.sentiment) // "positive"
console.log(result.score) // 0–100 GEO score
Get a free API key at webscore.dev.
Free Tier
The free plan includes 10 GEO scans to try it out — no credit
card required. Paid plans with higher limits are coming soon.
npm: webscore-sdk
Live tool: webscore.dev
API docs: webscore.dev/api-docs
Would love feedback on the prompt methodology or the
mention/citation split. Drop a comment below!
Top comments (2)
finally someone actually showing how to track this stuff, wonder if the results vary much between claude and gemini though
Great question! Yes, the results vary quite a bit between engines.
In our testing, Gemini tends to cite sources more frequently (higher citation rate), while Claude gives more nuanced sentiment around brands.
Perplexity is often the most "honest" — it will mention smaller brands if they have solid content, even without massive domain authority.
That's actually why we query all four in parallel and show per-engine breakdown — the aggregate score can hide some interesting differences underneath.
Try scanning your own brand and check the engine breakdown — would love to hear what you find!