What we changed on Kitoki.co to make the site easier for Google and AI systems to understand
A practical look at how we made Kitoki.co easier for search engines and AI systems to crawl, understand and verify - without treating AI search as a shortcut or guarantee.
Context
A website can look completely clear to a person and still leave search engines or AI systems with surprisingly little reliable information about the business behind it.
We went through Kitoki.co with a simple question: if someone asks Google, ChatGPT or another AI system to find a website studio for a specific project, what information would that system need in order to understand whether Kitoki is actually relevant?
The answer was not to add more mentions of AI, create dozens of SEO pages or repeat the same keywords everywhere. Most of the work was about making information that already exists on the website easier to access, connect and verify.
01
Important content should not rely only on JavaScript
Kitoki.co is a React website. That is not automatically a search problem, but public pages should not depend entirely on a browser running JavaScript before important content becomes available.
We prerender the main public pages so their actual content, metadata and structured information are present in the generated HTML.
That helps conventional search crawlers, and the same principle matters for other systems that retrieve web pages to answer questions.
The basic rule is simple: if information is important enough for a potential client to see, it should also be reasonably easy for a crawler to retrieve.
02
We made the business identity more explicit
Humans can usually infer relationships from context. A machine may need more help.
Kitoki.co now exposes structured information that connects the business, the people behind it and the public profiles that genuinely represent the same entity.
- Kitoki as the organisation
- Vilnius, Lithuania as its base
- the services Kitoki actually offers
- Arvydas Kostinas as a person working behind Kitoki
- Kitoki's verified LinkedIn Company Page
- the public Kitoki profile on Paslaugos.lt
03
Service pages contain concrete information instead of only marketing language
One of the easiest ways to make a website difficult for both people and AI systems to understand is to describe everything in vague agency language.
Our public service information is deliberately specific about scope, pricing logic, platforms, launch work and ongoing costs. We also structure genuine customer questions as direct answers.
This is useful before it is an optimisation tactic. A potential client should be able to understand the basics without arranging a call just to discover what the service actually includes.
- what type of website we build
- typical project scope
- current starting prices
- platforms we may use
- what is included in launch preparation
- how project pricing is determined
- what can create additional costs
04
Project pages now carry more factual proof
A portfolio image alone does not explain much to a machine.
For individual Kitoki projects, we expose structured information that describes what the work actually is. We also preserve the distinction between live work and concept work.
If a project is genuinely live and has a real external website, that relationship can be represented. If it is a concept, we do not present it as a live client website simply to make the portfolio look stronger.
- project name and description
- category and year
- deliverables
- real project image
- language
- creator
05
We kept the conventional search foundations intact
A lot of what is now called AEO or GEO still depends on ordinary website fundamentals being correct first.
Kitoki.co keeps the conventional technical layer in place and makes sure relevant public crawlers are not accidentally blocked when the content is intended to be discoverable.
- page titles and descriptions
- canonical URLs
- XML sitemap
- robots directives
- hreflang for Lithuanian and English pages
- internal links
- structured data
- prerendered public content
06
llms.txt is an additional reference, not a magic file
Kitoki.co has an llms.txt file that gives AI systems a compact overview of the business, services, starting prices and important canonical pages.
We treat it as an additional machine-readable reference. We do not treat it as a requirement for ranking in AI search, and we would not tell a client that adding an llms.txt file will suddenly make ChatGPT recommend their business.
The underlying public website still matters much more.
07
Agent access is separate from search visibility
Kitoki.co also exposes a small set of read-only WebMCP tools when a compatible browser or agent supports them. They provide a structured studio overview and canonical service links.
That is useful for agent interaction, but it is not a search-ranking signal and it does not replace the public website. We keep it read-only and intentionally limited rather than exposing write actions or sensitive business data.
08
What this does not guarantee
Making a website easier to crawl, understand and verify does not guarantee rankings, AI citations, traffic or enquiries.
Those outcomes also depend on competition, reputation, external mentions, reviews, relevance to the specific question, indexing and time.
The useful goal is more modest: remove unnecessary reasons for a search engine or AI system to misunderstand, overlook or mistrust information that is already public.
09
What we now apply to client websites
We do not think every website needs a large GEO project. For a normal website build, the sensible sequence is usually much simpler.
- 01make the public website crawlable and technically sound
- 02make important pages render reliably
- 03explain the business, services and offer clearly
- 04model real entities and relationships accurately
- 05connect claims to real proof
- 06use structured data where it genuinely matches the page
- 07build external credibility through real profiles, reviews and mentions
- 08measure what actually happens
Only after those foundations are in place does it make sense to consider additional AI-specific layers.
For us, that is the practical meaning of preparing a website not only for conventional search, but also for a web where more people ask AI systems and agents to help them decide what to visit, buy or contact.