AI search visibility: not a surcharge, but standard equipment
llms.txt, a quotable paragraph structure and a permissive robots.txt. This is not a package you buy afterwards but part of the build — and it does not cost extra. If you need standalone GEO/AEO work on an existing site, our sister brand aiseo42 handles that.Why is it not a separate line item?
Because bolted on afterwards, half of it does not work. Structured data is only reliable if it is generated from the same source as the visible content — otherwise the product price changes on the page while the declared price does not, and that is exactly what Google penalises. The same goes for machine mirrors and update dates: if someone maintains them by hand, sooner or later they will lie.
In a template system this is a question of plugins, and the quality depends on the plugin vendor. With hand-written code, findability is simply part of the system. That is why there is no "SEO package" surcharge here: the base price contains it.
What gets built into every page?
Complete structured data. Per page: WebPage, breadcrumbs, Service with price on service pages, Article with author and date on articles, FAQPage from the frequently asked questions. Not from a template: the schema is built from the actual content of the page, so it cannot drift away from it.
A machine-readable mirror of every page. The same content as clean, unmarked text, on the same path with a .md extension, and declared in the HTML head. A language model gets the substance in one request, with nothing to misread.
llms.txt and llms-full.txt. The first is a compact list of facts and pages with dates, the second is the complete content in a single file. These few kilobytes are the cheapest way for a model to process the whole site cleanly.
A quotable structure. Every page opens with a 40–70 word summary that stands on its own; the headings are the questions people actually search for; every number comes with its source and date. Models lift content at paragraph level — this is the shape they can lift.
A permissive robots.txt. GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot, ClaudeBot, Google-Extended, Applebot, CCBot, Amazonbot and the rest — allowed by name. What is not allowed in cannot quote you.
Server-side output. The content is real text in the HTML, not something that appears after JavaScript runs. Most models never see what only assembles after execution.
Speed and clean headers. Hashed, indefinitely cacheable assets, fonts served from our own server, zero external trackers on the critical path.
The proof: this page
You do not have to take my word for it, and you do not have to dig through my code either. It can be checked with Google's own tool, available to anyone:
- Rich Results Test — paste in any of my page addresses and look at the structured data, error list included.
- Ask ChatGPT, Perplexity or Gemini what RootCR does — the answer depends on what is written here.
The rest of the machine layer — the content mirrors, the fact and page lists written for models, the access rules — I show you working on your own site when you request a quote. I explain how it is put together; I do not put the finished solution in the shop window.
The same equipment goes into every client project. The bucipek.hu case study shows what that means in numbers: 192 indexed URLs, zero structured data errors.
How do I measure whether it lands?
Instruments instead of slogans. What I collect weekly from the server log:
| What | Where it shows up |
|---|---|
| Whether models reach the content | GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot, bingbot hits |
| Whether visitors arrive from AI surfaces | referrer: chatgpt.com, perplexity.ai, gemini.google.com |
| Whether they actually cite you | the same five questions every month to ChatGPT, Perplexity and Gemini — recorded |
| Speed | daily automated measurement, with the same command |
The third row is the important one, and the least automatable: the questions have to be asked, and it has to be recorded who shows up as a source.
When is this not enough, and when does it need separate work?
What I described here is the technical foundation: the fact that a model can read your site does not mean it will cite you. Citation needs content — pages that answer one specific question with numbers and a decision rule. That is writing, not code.
If your site is being built here, the technical foundation is a given, and we put the content structure together jointly. If, however, you already have a website and specifically need work on getting language models to cite it, that calls for a standalone service — handled by our sister brand:
Frequently asked questions
Is this the same as traditional search engine optimization?
Largely yes, but the emphasis is elsewhere: a search engine ranks pages, a model quotes paragraphs. I spell out the difference here: how does a small business get into ChatGPT and Google AI answers.
Do you guarantee that I will appear in ChatGPT's answers?
No, and anyone who guarantees it either does not understand it or is not telling you what they are selling. The model decides, and the decision changes month by month. What can be guaranteed: the content is reachable, readable, factual, dated and interpretable in machine form — without those you certainly will not appear.
How much does this add to the project?
Nothing. The amount on the pricing page includes it; the price is moved by the number of subpages, the integrations and the custom logic, not by findability.
Will you do it on an existing site built by someone else?
Yes, but first I look at the system and tell you how much of it is patching and how much is rebuilding: taking over an abandoned website. If the site is fundamentally sound and only visibility needs work, aiseo42 is the better route.
What exactly is structured data?
A machine-readable layer that states what the page depicts as data rather than as text. Spelled out in a separate article: what is structured data.