Your website, your brand, your winery —
visible to AI.

Not necessarily — and that is normal. AI assistants rarely give the same answer twice: results vary with the exact phrasing, your chat history, your location and ongoing model updates. That is why we never judge by a single answer. We put the same fixed set of buyer questions to several AI systems every month and count how often you are named across all of those answers. A single answer fluctuates; the naming rate across dozens of answers is a stable, comparable measurement. And every example in your report is a real answer we received — stored word for word, with date and system.

Because the measurement never changes: the same questions, the same AI systems, every month. Your buyer questions are defined at the baseline measurement and then frozen, so each monthly report shows the share of answers naming you compared with the months before — measured identically. Every underlying answer is archived word for word, so you can verify the trend yourself at any time. This stored evidence is also the basis of our money-back promise.

They measure two different things. The AIRI score rates your website itself: can AI systems reach it, read it, and verify your facts (eight dimensions, scored by three independent AI models, published as an open standard). AI visibility measures the outcome: how often you are actually named in AI answers. Our industry benchmark measures visibility with generic standard questions ("which wineries in Piedmont should I know"), while your individual report uses the questions your real buyers ask — derived from your customer profile: grape varieties, region, price range, target markets, languages. Different questions produce different answers, so the two numbers can diverge widely. A niche estate can be prominent for specific buyer questions and invisible for generic ones; a famous house often the reverse. That gap is not an error — it is the point: being visible in general is not the same as being visible where people actually buy. Between the two sits the third quantity, the one usually forgotten: presence — whether there is anything of yours at all that answers the question being asked. Readability is the precondition, presence is the work, visibility is the result. An estate can score 91 out of 100 on readability and appear in zero answers if nothing there answers the question.

Because AI systems draw on two sources. First, everything written about you elsewhere — Wikipedia, press, guides, retailer pages, links. Second, what your own website makes readable. For a handful of world-famous names the first source dominates: they get recommended no matter how their site is built. Our benchmark shows exactly this — one Austrian estate scores 1 out of 100 on website readiness and still appears in 39% of answers. For everyone else, and that is the overwhelming majority, the first source is thin or missing, and the website is the one lever you actually control. That is who this service is for: good estates whose reputation is bigger than their machine-readable footprint.

Usually not. A very low score almost always means one of three things, and only the last is about quality: the site blocks AI crawlers (robots.txt or bot protection turns them away), the site could not be read at all during the measurement (server errors, or content that only appears after JavaScript runs), or the site is genuinely thin on verifiable facts. The first two are technical settings that can be changed in an afternoon — which is why we report the eight dimensions separately instead of just one number: you can see immediately which case you are in.

Open yourdomain.com/robots.txt in a browser and look for entries like GPTBot, ClaudeBot, PerplexityBot, Google-Extended or CCBot. A line saying "Disallow: /" under one of those names means that AI system is shut out. No robots.txt at all is usually fine — the stricter problem is bot protection (Cloudflare and similar) that silently blocks anything that is not a normal browser. Our free report checks this for you and names the crawlers concerned.

What we do is LLM optimization — sometimes called GEO, generative engine optimization: making sure AI assistants like ChatGPT, Claude and Gemini can read, verify and cite the facts about your business. Classic SEO — competing for ranked links in Google search — is not our product, and we want to be exact about that: the twin is deliberately built NOT to rank in Google. It tells Google to index your own site instead, because two results for one estate would confuse your buyers and split your own site’s ranking signals. So the twin adds no Google rankings — by design. What does help your classic search is the report: it shows you where your own website blocks machines (crawlers turned away, facts a machine cannot verify, missing structured data), and fixing those on your site helps AI and Google alike. Your existing website and any SEO work you have done stay completely untouched.

Like an audiobook next to the book: the same content, a different form, for a different audience. Your website is made for people — images, atmosphere, design. AI systems don’t read images, they read facts. The twin is the same information about your estate in a form machines can read and verify reliably. You don’t get a second website to look after: you change nothing, maintain nothing, and your existing site stays exactly as it is.

No. The twin explicitly tells Google “the original site is the right address, index that one” (technically a canonical link to your domain). Your own website is what should appear in Google results — not the twin. We don’t block the twin from the web entirely, because that is the whole point: AI systems have to be able to read it. In short: readable for AI, while your own site is what shows up in Google search.

Almost — but let’s be precise, because it isn’t strictly true. The twin is an ordinary, publicly reachable page; anyone with the address can open it in a browser. What we steer is who *finds* it: AI systems are explicitly invited, and Google is pointed at your original site. In practice that means it is seen almost only by AI — and by people following an AI recommendation. What is a twin? →

They are not redirected automatically, and that is deliberate. Redirecting people while machines see something else counts as cloaking at Google and can penalise your own domain — we won’t take that risk on your behalf. Instead the visitor sees your content in your design, with your logo and your photographs, and a clear button takes them on to your website, your shop and your contact details. And note: what the measurement counts is that your name is *mentioned* in the answer — not which address sits behind it.

Yes. Minutes after your payment arrives our agents start by themselves: they read your website, take over your colours, typefaces, logo and photographs, and build your twin from them. Before it goes live a person at Operal looks at it — we deliberately do not automate that last step. You normally receive the link within one working day; the binding commitment is 20 working days.

No. Your website and your shop stay exactly as they are. We build a separate, machine-readable twin that lives on its own address under operal.tech and mirrors your content. If you ever stop the service, your website is exactly as it was before.

Then you will see exactly that in your report — every answer is stored word for word, nothing is dressed up. And we are honest here: we cannot guarantee mentions, because the AI systems belong to third parties and decide for themselves whom they name. Our money-back promise therefore fully covers what is in our hands: twin live within 20 working days, AI-readability of at least 90 of 100 points, reports delivered — and if your twin’s measured readability later falls below 90/100 and we do not restore it within 30 days, you get the annual fee back. The full conditions are published in our terms.

Yes. Every published twin is an indexable page on our domain operal.tech that links openly to your existing website in three places and explicitly marks it as the original via a canonical annotation. That means one more public, machine-readable place on the web exists that names you and points at you — on top of the twin’s actual job of making your facts AI-readable. What we deliberately do not derive from this: a ranking promise. Nobody can honestly guarantee whether or how strongly search engines or AI systems weight a single reference — which is why our money-back promise covers measurable readability, not link effects. Read the terms →