geolint MCP¶

Lint your website for AI-search readiness. AI answers are the new front page: ChatGPT, Perplexity, Claude, Copilot and Google AI Overviews send traffic, or do not, based on whether their crawlers can fetch and quote your pages. geolint fetches a page, its robots.txt and llms.txt, evaluates 51 known AI crawler tokens against the rules, runs 51 audit rules across five categories, and prints a scored report with a concrete fix for every finding.

Server type: Local CLI (npx) with built-in stdio MCP server mode
Auth: none (site audit tool)
Install: npx @iliasabk/geolint check yoursite.com
MCP mode: geolint mcp (stdio server for AI assistants)
Version: 0.3.2 (npm, verified)
License: MIT
Category: SEO / AEO / GEO

Why This Matters for Operators¶

Most sites accidentally block or confuse AI crawlers: a stale Disallow: / left over from staging, a noindex that never got removed, a client-rendered page that looks empty to a bot that does not run JavaScript. Existing tools are blocklists or score-only web apps. They tell you to block everything, or give you a number with no path to improve it.

geolint is the linter: concrete findings, concrete fixes, runnable in CI on every pull request. Five categories cover the full AI-visibility surface:

Category	Rules	Examples
AI Crawler Access	10	search-bots-blocked, wildcard-block-all, stale-tokens
llms.txt	10	missing, invalid-structure, broken-links, relative-links
Structured Data	6	no-jsonld, invalid-jsonld, missing-article-fields
Citability	9	thin-content, no-h1, missing-dates, no-question-headings
Technical Foundation	10	client-rendered, https, slow-response, sitemap-missing

What a Report Looks Like¶

Real output from the bundled demo site:

geolint v0.2.1 - AI-search readiness
http://localhost:4173/
200 OK - robots 200 - llms.txt 404

██████████████████████████░░░░  86/100  Grade B

FINDINGS
  AI Crawler Access
    ✗ ai-crawler/search-bots-blocked  PerplexityBot is blocked by robots.txt -
        Perplexity cannot use your pages as AI answer sources
        fix: Remove the Disallow covering PerplexityBot in robots.txt, or add
        an explicit "Allow: /" for it.
  llms.txt
    ⚠ llms-txt/missing  No llms.txt found
        fix: Create /llms.txt at the site root: an H1 title, a short blockquote
        summary, and ## sections linking to your key content.

Every finding carries a rule id, a severity, the matched evidence, and a fix. The crawler report groups all 51 tokens by vendor (OpenAI, Perplexity, Google and the rest) and shows whether each is allowed for training, search, or user-fetch.

Commands¶

Command	What it does
`geolint check <url>`	Audit one URL; `--compare` two pages head-to-head, `--baseline`, `--fail-under`
`geolint crawl <url>`	Crawl same-origin pages and audit the whole site
`geolint init <url>`	Crawl the site and generate an llms.txt
`geolint diff <old.json> <new.json>`	Score delta plus added and resolved findings
`geolint rules`	List the 51 audit rules
`geolint bots`	List the 51 known AI crawlers and the impact of blocking each
`geolint mcp`	Run an MCP server on stdio for AI assistants

Installation¶

No install, no config:

npx @iliasabk/geolint check yoursite.com

Run it in CI with the official GitHub Action:

- uses: iliasabk/geolint@v1
  id: geolint
  with:
    url: https://example.com
    fail-under: 80

And for AI assistants, the MCP mode serves the same audit as tools:

claude mcp add geolint -- npx -y @iliasabk/geolint mcp

Business Relevance¶

Marketing teams find out why AI assistants skip their pages before blaming the model.
SEO and AEO agencies add a scored, fix-per-finding audit to every client engagement, runnable on every deploy.
Publishers and content sites protect the traffic that flows through AI answers by keeping crawler access, llms.txt and structured data healthy.
Product teams get a CI gate that fails the build when a page ships with a stale noindex or a blocked search bot.

Integration with CorpusIQ¶

geolint audits whether AI systems can read your public pages; CorpusIQ answers whether the numbers those pages promise are actually retrievable. An assistant connected to CorpusIQ reads authorized business data (GA4, Search Console, Shopify, Stripe, HubSpot) with source-cited answers, so the same session can pair a geolint fix list with the traffic and conversion data that shows which fix moved. For docs and content teams, the two cover both halves of AI visibility: technically readable, and factually answerable.

Limitations¶

New project (npm 0.3.2) with a single maintainer; the rule set is opinionated and evolving.
Heuristic scoring. A grade is a summary of the rules, not a ranking guarantee from any AI vendor.
Node 22+ required for the CLI.
Crawler token list covers the 51 known bots at release time; new tokens can appear between versions.
The MCP mode is local stdio only - there is no hosted remote endpoint.