What we asked, and who answered

Twenty questions, each asked in Italian, English and French, at four grounded AI engines: ChatGPT, Perplexity, Gemini and Claude. That is 240 calls, and 209 of them returned a usable answer. The run is dated 30 September 2026 and every figure below comes from those stored answers. Google AI Overviews and DeepSeek were not part of this run, so this case speaks for four of the six engines named on this site, not all six.

Ten of the twenty questions were regional: where would someone stay anywhere in the Langhe, the wine country around Alba, Barolo and Barbaresco, which is the demand this property sits inside at 12km from Barbaresco, 9km from Neive and 20km from Alba. The other ten were narrow: they named something the property verifiably has, a hot tub, a dog friendly room, wheelchair access, a family room, a garden and a sun terrace, parking, or a location between two named towns.

The regional answers never named it

Across the ten regional questions in three languages, 96 answers came back, and Nonna Giulina appeared in none of them. Castagnole delle Lanze did not appear once either: the engines did not pick a different guest house in the village, the village itself never came up when the question was regional.

That absence sits inside a market with no leader to blame it on. The 96 regional answers between them named 118 distinct properties, and the one that came up most often still only appeared in 19 of the 96, under a fifth. Nobody owns this question yet, which means the opening is real.

The narrow answers found it by one route and no other

The ten narrow questions returned 113 answers, and the property was named in seven of them. Six of those seven are the same question, asked once in Italian, once in English and once in French: where to stay in Castagnole delle Lanze, a village of a few thousand people. The seventh is a French question about a stay between two named towns.

Every other narrow question returned nothing. The hot tub question, the dog friendly question, the wheelchair access question, the family room question, the garden and terrace question and the parking question all came back without the property, and each of those six questions describes something the property’s own public booking listing already states. The engines are not guessing wrong. They are not reading that page at all.

Castagnole delle Lanze is a useful control here. The village was named in 20 of the 113 narrow answers, nearly three times as often as the property that sits inside it. The place is well known to these engines. The property is attached to almost nothing.

Where the four engines are actually reading this market

After TripAdvisor (114 citations) and Booking.com (100), the sources behind these 209 answers are Italian wine travel writing and specialist directories: winetravelguides.com (24), patatofriendly.com (24), dogwelcome.it (22), bed-and-breakfast.it (19), viaggichemangi.com (19) and intothevineyard.com (18).

One of those is worth sitting with. A pet directory is cited 22 times across this market, and Nonna Giulina allows dogs, stated plainly on its own booking listing. The dog friendly question still named it zero times. The signal exists. It has not reached the one directory the engines are actually citing for that exact question.

It tells them their next hour of work is a form, and the new website can wait. The engines already trust a site whose entire purpose is to say which places welcome dogs, and they cited it 22 times in their region, yet a house that welcomes dogs isn’t on it. The facts already exist on their Booking listing, and AI doesn’t connect them to the hotel because no source it relies on says them out loud. Effort should go where AI already reads, so being listed in the right three places will do more this year than anything they write on their own site.

Sebastián Ocampo, founder of Y8Y

Two failures, not one

A sister audit, run with the same instrument in the same month on Hotel Savoy Bern, a 64 room four star hotel, named that hotel in 46.6% of its answers and missed it in a single category. This property fails the opposite way: it is findable by one route nobody types, where to stay in a village of a few thousand people, and absent from every quality it actually offers.

Same four engines, same month, two different diagnoses. A reader checking their own property against either story should ask which failure is theirs: unknown everywhere, or known for the wrong single reason while every real attribute sits unread.

What is still open

There is no rate for this property on the public record, no counted guest mix and no site of its own that we have confirmed, so none of that appears here and no figure above is estimated to fill the gap. What the run establishes is that nobody finds this property today. It does not establish who should, and that is a separate piece of work from the one this audit did.

For an independent property without a chain’s marketing budget, the fix is not a new website. It is getting the attributes that are already true, stated on the public listings these engines already cite, in the specific language each question uses. A property audit is the next hop in that work, one listing, one directory and one named attribute at a time.

See also: Airbnb and STR guide | Does Airbnb listing quality affect AI visibility | What makes an Airbnb listing appear in AI search | Entity salience

Which one are you

Ask ChatGPT for a place to stay in your village or street, then ask for the thing you’re best at with no place name. If only the first returns you, you’re Nonna Giulina, because AI knows you exist and doesn’t know what you’re for. If you show up for both but not for the question that pays your bills, you’re the Savoy, because AI knows you well and describes you to the wrong guest.

Sebastián Ocampo, founder of Y8Y