Services · Grow
People now ask ChatGPT, Claude, Gemini and Perplexity before they search. Those assistants answer from what they can read and trust. We make your site readable, your facts consistent, and your business something an assistant can act on, then we measure whether it worked.
- A business that ranks for its own name but never appears when an assistant is asked what it does.
- A company whose site is built in a way crawlers and assistants read badly: scripts first, facts buried, every page saying the same thing.
- A team that wants assistants to do something with it, like book a call or send a request, not only mention it.
What we do
The work, in plain words.
- 01
Let the right crawlers in
A crawler policy that names the AI search and assistant bots and lets them read what should be found, and a sitemap with real dates instead of every page changing on every build.
- 02
Say who you are, the same way everywhere
One category line, one name, one address line, on the site, in the structured data and on the profiles assistants cross-check. Organization, person, service, breadcrumb, article and FAQ markup that matches the visible page.
- 03
Pages written for the question
A page per thing a buyer asks for, in the buyer's words, with proof and visible answers. Generic pages that all say the same thing do not get cited.
- 04
Text built for machines
llms.txt and a full-text version generated from the same content as the pages, and a markdown copy of every page, so an assistant reads the facts without the layout.
- 05
Something an assistant can do
An MCP server and WebMCP tools on the page, so an assistant can look up your services or send a request to your team, with limits and a log.
- 06
Measure it
AI crawler visits by bot and page, assistant calls by tool, Search Console and Bing reports, and the queries that matter, checked monthly against a baseline.
How it starts
Three steps, each in writing.
- Step 1
A read of your site
What crawlers and assistants see today, what they get wrong about you, and the queries you should own.
- Step 2
A written plan
Fixes in order of payoff, the new pages, and what we will measure.
- Step 3
Build, ship, measure
Phase by phase, each one live and checked before the next.
Proof
Running today, not a promise.
This site
Everything on this page runs here: the crawler policy, the structured data, the generated llms files, a markdown copy of each page, a checked sitemap, and a gate that fails the build when a page loses its title, canonical or share image.
Our llms.txt →Our own MCP server
Assistants can ask this firm questions directly and get answers with the page each came from: the MCP server, its server card and its catalog entry.
Blue Coast Water for AI assistants →Samplyn for AI assistants
Our own product's agent kit: a public MCP server in the official MCP Registry, WebMCP on its demo page, per-page markdown and a monthly AI traffic report.
samplyn.com/agents ↗MCP servers
The tool side of AI search, built and run as its own service.
MCP server development →
What we do not do
Where we draw the line.
- We do not promise a position or a citation. No one can; we show you what changed and what it moved.
- We do not write pages for keywords we cannot back up. Every claim on a page traces to something real.
- We do not hide text for bots. What an assistant reads is what a person sees.
Questions
What people ask first.
Is AI search different from SEO?
It builds on it. Assistants still crawl and rank pages, but they also read structured facts, plain-text files and tools, and they cross-check you against other sources. We do both together.
What is llms.txt?
A plain-text file at the root of a site that lists its key facts and pages for language models. Ours is generated from the same content as the pages, so it never drifts from them.
Do we need an MCP server?
Not to be found. It matters when you want an assistant to do something with you: look up a product, check availability, send a request. We start with being found and add tools when they pay.
Will this hurt our Google rankings?
No. The same work (clear pages, honest structured data, real dates, fast pages) is what Google asks for too.
How do we know it worked?
We set a baseline before we start: crawler visits, assistant calls, Search Console and Bing numbers, and how assistants answer your key questions. Then we report against it monthly.
What does it cost?
The founder quotes after reading your site, in writing. There is no price list on the site.
No prices on this site: the founder quotes in writing after the first conversation. How engagements work.
Tell me what you need.
Philip Radvan, founder. Every message lands with me, and it's me on the call.