Let ChatGPT, Gemini and Google understand your brand and be able to cite it.
A GEO and technical SEO service for brands and ecommerce: we look at what ChatGPT, Gemini, Perplexity and Google’s AI summaries answer when someone asks about your category, prepare your site so their bots read it well, write the content that answers buying questions and track how the answers change. Without promising what nobody controls: what the AI says is decided by the AI.
What our clients say
For brands that do not show up when the buyer asks the AI
You ask ChatGPT and you are not there
You ask ChatGPT or Perplexity for the best in your category and your competitors come up, or your brand comes up described with old or plainly false data.
Google answers before the click
Google’s AI summaries answer on the results page itself and you see fewer visits on searches where you used to be found.
A site the AI reads badly
Content that only appears with JavaScript, AI bots blocked without you knowing in robots.txt or the CDN, product pages without specifications and no structured data.
Six jobs on what does depend on you
We do not control what the AI answers. We can make your site read well, make your data match everywhere and give you answers worth citing. It can be one of these jobs or several, with the scope written before we start.
AI answers audit
We prepare the questions your buyer asks (comparisons, “the best for…”, alternatives to your competitors) and look at what ChatGPT, Gemini, Perplexity and Google answer: whether they cite you, whom they cite, which sources they use and what they say about you.
A site AI bots can read
Robots.txt and CDN reviewed so AI search crawlers can get in, content that shows without relying on JavaScript, speed, sitemap and indexing in order.
llms.txt, Markdown and structured data
An llms.txt with the map of your site, a Markdown version of your pages for agents and well-built Organization, Product and FAQ schema. The same as we have on epinium.com.
Content that answers buying questions
Honest comparisons, buying guides, FAQs and complete specifications: pages an assistant can cite because they answer something concrete, reviewed by you before they are published.
Consistent brand signals
Who you are, what you sell, where you are and who you sell to, told the same way on your site, your structured data, your marketplace listings and your profiles. We fix what contradicts itself, which is what makes a model doubt.
Citation tracking
We repeat the same set of questions as often as we agree and show you what changed. It is a hand-made sample, not a ranking: answers vary from one query to the next.
Our own site is already ready for AI agents.
Before offering it, we did it on epinium.com. We publish an llms.txt with the site map for language models and an llms-full.txt with every article. Every page has a Markdown version, served to whoever asks for it with the “Accept: text/markdown” header. The 404 tells an agent where to go instead of leaving it at a dead end. Company data goes in Organization schema, taken only from what our legal pages publish. The robots.txt handles AI bots one by one, and new or changed pages are announced through IndexNow to the search engines that support it. It is the same we will do on your site, and you can check it yourself.
See our llms.txt →Four steps, and you can stop at the first
Diagnosis
A call about your category and your competitors. We run a few of your buyer’s questions in the assistants, show you what they answer today and tell you where we would start.
Audit and scope in writing
An audit of the answers and of the technical side, and a closed scope: what we touch, in what order and what we need from you. Before any invoice.
Technical work and content
We fix the technical side (bots, llms.txt, Markdown, structured data, speed) and write or rewrite the pages that answer buying questions. Nothing is published without your approval.
Tracking
We repeat the same questions as often as agreed and tell you what changed and what we would do next. If you want to stop, you stop.
We work on your site and your content. We do not sell citations.
We do not promise you will appear
The answers of ChatGPT, Gemini or Perplexity change from one query to the next and nobody outside those companies controls them. Whoever guarantees you will show up is selling something they do not have.
Not mass content
We do not publish hundreds of generated articles to “feed” the AI. A few pages that truly answer what your buyer asks and that you have reviewed.
Not an account agency
We work on your site and your content. We do not run your ad campaigns with a human team: that is Roicos, the group’s sister agency, and we will tell you on the first call if that is what you need.
Related
WordPress to Astro
Fast sites that search engines and AI agents can crawl and read from day one, like this one.
PlatformGEO in the Epinium platform
If you prefer to do it yourself: your product listings prepared for shopping assistants, with every change approved by you.
GuideWhat is GEO
Generative engine optimization explained: how it resembles SEO and how it does not.
What people ask before the first call
What is GEO and how is it different from SEO?
GEO (generative engine optimization) is working on your site and content so assistants such as ChatGPT, Gemini, Perplexity or Google’s AI summaries understand you and can cite you in their answers. SEO looked for a spot in a list of links; GEO looks for the mention inside the answer. The base is the same (a crawlable, fast site with useful content), so we do it together with technical SEO, not separately. More in our GEO guide.
Do you guarantee ChatGPT will cite me?
No, and be wary of anyone who does. What an assistant answers is decided by its model, changes from one query to the next and nobody outside OpenAI, Google or Perplexity controls it. What we do is make your site easy for them to read, make your data say the same thing everywhere and give you pages that truly answer what your buyer asks. That gives you more chances; it does not guarantee anything.
What is llms.txt and do I need it?
It is a file at the root of your site that tells language models what is on the site and where to find it. It is a recent proposal and not every assistant reads it, so it does not work miracles, but it costs little, does no harm and forces you to organise the site with agents in mind. Ours is at epinium.com/llms.txt.
How do you measure whether I appear in AI answers?
With a fixed set of questions from your category that we run in ChatGPT, Gemini, Perplexity and Google as often as we agree, noting whether you appear, which sources are used and what is said about you. It is a hand-made sample, not a share of voice: answers vary between queries and between users, so we read it as a trend and never as an exact number.
Can I do it myself with the Epinium platform?
Yes, if what you want is to prepare your product listings (Amazon, Shopify, WooCommerce) for shopping assistants: it is in GEO in the Epinium platform, with every change approved by you. This service is for when you prefer that we do it across your whole site: technical work, content and brand signals.
How much does it cost and when will I see results?
It depends on the scope, so we do not publish standard rates or timelines. The diagnosis is free and the scope and price come in writing before any invoice. As for when it shows, we will not give you a date: it depends on when the assistants read your site again and on what your competitors do. The rest of what we do is in Services.
See what the AI says about your brand today
A free diagnosis: one call, a few questions from your category in the assistants, and we tell you what we would fix first. Sometimes the first thing is not the site, but making your data say the same thing everywhere.