Your strongest proof points, reorganized around the specific buyers and scenarios where AI is helping them — so AI models can consistently find the right evidence in the right place. Hosted at a subdomain you control. The structure adapts as our platform analysis learns what's landing, and the same signals feed back into your own content cycle.
A set of pages we host at a subdomain you control — typically llms.[your-domain] or info.[your-domain]. Each page is built around a specific buyer persona and a specific scenario where AI is helping them, and pulls together your strongest proof points for that combination.
The pages are written for AI agents to consume, not for humans to browse. Your existing site keeps doing what it does for human visitors; the documentation does a different job alongside it.
AI models are already shaping how your buyers learn about you. The question is whether they're consistently surfacing your strongest proof points — or hitting whichever page happens to rank, even when it's not the most persuasive evidence you have.
Your existing site was built for humans browsing. The proof points that matter most live scattered across pages written at different times for different audiences. Company Documentation pulls those proof points together and reorganizes them around the persona × scenario combinations where they actually matter, so the right evidence is in the right place when AI needs it.
A second surface, dedicated to your proof points and how AI uses them — sitting alongside your main site, feeding signals back into it.
Setup is one-time, takes about 15 minutes of DNS work, and your IT team owns it. After that, four things keep happening:
utm_source=unusual, so any AI-influenced human traffic that does come through to your main domain shows up cleanly in GA4, your CRM, and your existing channel reports — alongside everything else.Because every page is a discrete unit of proof tied to a specific persona × scenario, we can show you which ones are actually doing the work — and that becomes a content signal you can act on:
Static documentation does heavy lifting, but visiting AI agents often probe with follow-up questions, edge cases, or scenarios a single page can't cover. AgentDesk is an agent we host alongside your Company Documentation that handles those follow-ups.
It's reachable through the same subdomain, but only visiting AI agents discover and use it — humans browsing the documentation don't see it. When a visiting agent (acting on behalf of a human buyer) needs more than what's on the page, it can interact with AgentDesk for additional answers grounded in your documentation and the category knowledge we maintain.
We capture the full back-and-forth conversation, not just individual Q&A — the whole multi-turn exchange between the visiting agent and AgentDesk. Then we review the transcripts and surface insights you'd never see from web analytics — what visiting agents are asking, how the conversations evolve, where the documentation is straining, where to invest next. We're also slowly rolling out the ability for AgentDesk to share additional proof points beyond the documentation pages — for example, anonymized case studies your sales team would normally walk a prospect through.
Google treats llms.example.com as an independent site from example.com. Authority, ranking, and indexing operate independently — nothing about the subdomain can degrade your main domain's standing.
Each page is a fresh restructuring of your proof points around a persona × scenario — different shape, different language, different page templates. Google's duplicate-content detection requires substantial textual overlap, which restructured proof-point pages don't produce.
The documentation is published for AI agents to consume. We track human clicks and human visits to each page — they're extremely rare, because the URLs are long and topic-specific. The hypothetical case where one would get meaningful human traffic: a person searches for the exact narrow topic of that page. We haven't really seen this happen, and if it does, that itself becomes a useful signal we surface back to you.
No. Google indexes and ranks subdomains independently of their root. The subdomain has its own ranking signals; nothing it does flows up to your main site. If the subdomain ranked or didn't rank for a query, that wouldn't change your main domain's position on the same query.
No. The subdomain hosts rewrites, not copies. The structure is restructured for AI parsing (more explicit claims, denser context, fewer marketing flourishes), the language is reworked, and the page templates are different. Google's duplicate-content systems look for substantial textual overlap, which rewrites don't produce.
It can, and that's fine. Subdomains are public web. Anything Googlebot indexes there ranks on its own merits, separate from your main domain. The documentation is built for AI agents to consume — that's where its traffic comes from in practice.
Every outbound link from the subdomain to your main domain carries utm_source=unusual. Filter or segment by that source in GA4, your CRM, or your attribution stack to see AI-influenced traffic alongside your other channels.
Three steps for your IT or web team — DNS records, an indexing link, and the AgentDesk endpoint. About 15 minutes of work.
Open the setup guide