Are you ready for the AI age?

A search engine ranks pages and sends you to one. An answer engine — ChatGPT search, Perplexity, Google's AI Overviews — does something different: it assembles an answer from sources it can parse and attribute, and most people never click through to any of them.

That difference is brutal for a business with no website. Under the old model you could still be found: your Google listing ranked, someone clicked it, they phoned you. Under the new one, you have no first-party source. An assistant asked "where should I eat in Skiathos Town tonight" can only describe you through third-party aggregates you do not control and cannot correct — or, far more often, skip you and name a place it can quote.

On this island that is not a hypothetical. 87 of 130 venues have no website of their own, and 34 score below 40 out of 100 on the citability measure below. Only 29 score 70 or above.

Four things, in this order:

1. How you appear online. Your website if you have one, your social accounts, your ratings across Google, Tripadvisor and Restaurant Guru, and how far those platforms disagree with each other about you.
2. Your AI visibility score. How citable you are to an answer engine, scored against six weighted signals, with your result on each shown rather than summarised.
3. When your demand arrives. A full week modelled hour by hour, your five busiest windows, and your year month by month.
4. What we would fix, in order — each item traced to an observation, not to a sales script.

So we separate two things on every page and we never blur them:

Measured. Your ratings, review counts, website and social presence. We checked these against published sources and dated them. They are facts.

Modelled. When demand arrives. This is inferred from the structure of the island — the season curve, the charter week, the bus timetable and the 17:00 beach exodus, Greek meal times, and the ferries, flights and cruise calls that land blocks of people at the New Port. Every coefficient is documented and every assumption is labelled as one.

The model is useful because the structure is real and it is the same structure for everyone. It is not a substitute for your own figures. Give us one month of your own covers and it stops being a model and starts being calibrated — and we will tell you how far off we were.

Six inputs, all of them derived from research published elsewhere on this site:

• Season. A monthly index built from published opening and climate data, peaking in August. Beach venues drop to zero in winter because they are shut, not quiet.
• The charter week. Friday is the island's heaviest arrival day — which is also why it is the airport's worst day for delays, since the binding constraint there is aircraft stand availability.
• Hour of day, by venue role. A beach bar and a hilltop taverna are not the same business. Beach trade peaks in the early afternoon and collapses at 17:00 when the buses fill; town dinner service builds from 20:00 and peaks after 21:00, because that is when Greeks eat.
• Arrivals. Ferries, flights and cruise calls each land a block of people, which decays over the following hours and reaches different zones of the island to very different degrees. Cruise passengers on a short call mostly never leave the waterfront.
• The meltemi. The summer wind is northerly, so on a windy day the north coast loses its trade and the south coast gains it.
• 15 August. Dekapentavgoustos triggers a mass domestic migration to the islands, stacking Greek demand on top of the international charter peak.

The code is in assets/demand.js and the assumptions are commented in it, including the ones we are least sure about. You are welcome to argue with any of them.

The model takes ferry, flight and cruise arrivals as an input. There is no single free public API for any of the three on a small Greek island: ferry timetables sit behind commercial aggregator agreements, cruise calls come from port-agent and commercial schedules, and flight data comes from the airport operator or a commercial provider.

The pipeline is built as adapters behind one normalised schema, so switching provider changes one mapping function and nothing else. Right now it is running on an illustrative sample schedule — realistic in shape, fictional in detail, and labelled as such on every page it touches. Nothing in the demo pretends to be a real sailing.

Plainly: some of the 87 businesses on this island with no website of their own will want one, and we would like to build them. That is the business.

But the guide came first and it stands on its own. Your listing on it was written before we contacted you, it is free, it is not for sale, and no business has ever paid for placement or a score. If you never speak to us, your page stays exactly as it is.

If anything we have published about you is wrong, tell us and we will fix it — that is not conditional on anything.

Frequently asked questions

Do you have my restaurant's website traffic?

No, and nor does anyone else. Analytics for businesses at this scale are not published. Any agency showing you a confident visitor number for a single restaurant has bought a modelled estimate and removed the error bars, or made it up. We separate what we measured — your ratings, website and social presence, all dated — from what we modelled, which is when demand arrives.

How can you know when my restaurant is busy?

We do not know, we model it, and the page says so. The model uses the island's structure rather than your data: the season curve, the fact that Friday is the heaviest charter arrival day, the bus timetable and the 17:00 beach exodus, Greek meal times, ferry and cruise arrivals at the New Port, and the meltemi. Give us one month of your own covers and the model can be calibrated against reality.

What is an AI visibility score?

A measure of how citable your business is to an answer engine such as ChatGPT search, Perplexity or Google's AI Overviews. Those systems assemble answers from sources they can parse and attribute rather than ranking pages, so a business with no website of its own has no first-party source — it can only be described through third-party aggregates it does not control and cannot correct.

Did any business pay to be listed or rated?

No. Not one. Every venue on the guide was researched and scored before any contact was made, placement is not for sale, and a business working with us commercially gets exactly the same editorial treatment as one that never replies.