Home / Bot

About UnomageAIVisBot

UnomageAIVisBot is the crawler behind the AI Visibility Checker. When someone submits a domain to be scanned, this bot fetches a small sample of pages from that site — the homepage plus up to nine more, discovered the same way an AI assistant's own crawler would (sitemap first, then homepage links) — in order to evaluate how readable and citable the site is to AI assistants.

How it identifies itself

Every request UnomageAIVisBot makes sends a standard identifying User-Agent header:

UnomageAIVisBot/1.0 (+https://aivisibility.unomage.com/bot)

so any site owner or server log can identify traffic from this tool and, separately, decide whether to allow it.

It respects robots.txt

UnomageAIVisBot obeys the target site's robots.txt rules for its own user agent by default. If a site's robots.txt disallows UnomageAIVisBot, this tool will not scan it. This is a deliberate default in the underlying aivis analysis library and is not overridden by this service.

What it does not do

• It does not submit forms, click links, or take any action beyond fetching pages.
• It does not store credentials or attempt to bypass logins or paywalls.
• It runs a small, bounded scan (at most ten pages, a 45-second hard ceiling) — it is not a general-purpose crawler.
• Requests to any one target domain are rate-limited, so it cannot be used to hammer a site.

For AI agents: use this tool directly (MCP)

Besides being scanned, this tool can also be called by an AI agent. We run a free, public Model Context Protocol (MCP) server at https://aivisibility.unomage.com/mcp. Any MCP-capable client — Claude Desktop, Claude Code, Cursor, and others — can add that URL as a tool server with no API key and no signup, then call one tool:

check_ai_visibility(url, forceRescan?)

It runs the exact same scan as this website, subject to the exact same rate limits, and returns the score, grade, per-pillar breakdown and top fixes as plain text an agent can read and act on.

WebMCP (experimental)

This site also implements WebMCP, Chrome's in-progress browser API for exposing a site's actions to AI agents, two ways. The homepage's scan form is annotated with WebMCP's declarative attributes (toolname, tooldescription, toolparamdescription) — plain HTML, readable by any tool that parses the raw page, including Lighthouse's newer Agentic Browsing checks, with no browser support required. Separately, /static/webmcp.js registers the same tool imperatively via document.modelContext.registerTool() for agents that execute JavaScript in a Chrome build with WebMCP enabled.

WebMCP is still experimental: the imperative API currently only runs in Chrome behind the chrome://flags/#enable-webmcp-testing flag, or on origins that have joined Chrome's WebMCP origin trial — this site hasn't joined that trial yet, so most visitors' browsers won't have document.modelContext available at all, which is expected. The declarative HTML annotation works today regardless of any flag. The MCP server above is the integration path that works everywhere, right now, for every agent.

Concerns or removal requests

If UnomageAIVisBot is causing problems for your site, or you'd like your site's report taken down, visit /removal — no verification is required to have a report unlisted.

For anything else, contact: TODO — add a real contact email/address here before this ships to production.