Kie Furusawa

Back to all lectures

Lecture 3

When a Nearby Property Pulls the AI Answer Away

隣の影

Before reading: this lecture builds on lectures 1 and 2. In the first, we examined the digital trace of a ryokan, AI retelling, and the five tracks of first reading; in the second, category shift, where a ryokan begins to sound like a guesthouse or a basic overnight stay.

A composite scene from several observations, with no real names: an email arrives at a family ryokan from a prospective guest. She writes in English, politely and with a little anxiety: “AI says your ryokan is near the large onsen hotel by the lower bridge. Am I right that dinner and the baths are shared?” At first, the owner laughs. The large hotel has a separate entrance, a different dinner, a different schedule, a different turn off the bus route. The two are linked only by the name of the hot spring and by the fact that both listings on one platform landed in the section “accommodation near the old bathhouse.” A small rough detail: AI gave the small ryokan’s phone number correctly, but explained the evening as if it had borrowed part of the rules from the neighbor.

There is no longer just one old translation on the table, as in the previous lecture. This is another kind of trouble. The ryokan itself is named correctly, the category is not entirely broken, and the word ryokan even remains in the answer. But the AI retelling looks at the house through someone else’s lamp. The large hotel shines brighter in texts, appears more often in lists, and has a name similar to the place name. The small house loses independence and starts to sound like “an option nearby.” In this lecture, we move from the question “why did AI flatten the type of the ryokan?” to “why did AI start holding onto someone else’s property, even though your house is present in the materials too?”

A Borrowed Name as a Handle on Your Door

In the first lecture, we said that the model compresses different fragments into an AI retelling. In the second, we saw that it may place a ryokan into a thin category. Now add one more mechanism: sometimes the answer does not so much change the house’s type as take someone else’s name as the handle by which it can pull the whole scene. The reader asks about a small ryokan, and AI starts explaining the area through a more visible property. Then that property quietly seeps into the details.

Neighbor’s shadow — A situation where a more visible nearby hotel, hot spring, or similar name pulls the AI answer away. This definition is important to keep separate from category shift. In a category shift, the ryokan itself starts being called a guesthouse, hostel, or basic overnight stay. In a neighbor’s shadow, the ryokan may still be called a ryokan, but it is explained through the louder neighbor: “near such-and-such hotel,” “part of the same onsen area,” “a similar option by the same station.” From the outside, this looks softer than a blunt error. Inside the guest’s expectation, it may be worse: she arrives with an idea assembled from two different houses.

Why does this happen so often with small ryokan outside major tourist centers? Because several strong borrowed signs often surround them. The name of the spring, the valley, the bus stop, a large hotel with the same word in its name, a district page on a booking platform where all houses sit in one long list. The owner hears the differences instantly. For her, “the upper house by the cedars” and “the hotel by the lower bridge” are separate worlds. AI retelling sees word proximity and convenient links: one village, one onsen, a similar route from the station, a similar set of words about bath and dinner.

Picture a shelf of keys by the entrance. On one hook hangs the small wooden fob of the ryokan; on another, the heavy brass fob of the large hotel. If the light falls only on the brass, the hand reaches for it first. The model does not “want” to make a mistake. It is simply easier for it to explain the area through the property that is more strongly lit in the texts. Then that light falls on the neighboring door.

Where the Neighbor’s Shadow Comes From

The first material that creates a shadow is a similar name. In onsen villages, this is ordinary: several houses use the name of the spring, stream, bridge, slope, or old bathhouse. For a guest looking at a map and reading Japanese names for the first time, the difference is already fragile. For AI it becomes even thinner when English fragments keep abbreviations: “Kawa Onsen Inn,” “Kawa Onsen Hotel,” “Kawa no Yu.” One word holds the area; the second has to hold the house. If the second word is weak, the answer starts to slide.

The second material is shared district pages. Booking platforms and tourist notes often show a selection around a station or spring, rather than the separate life of the house. A large hotel, a family ryokan, a minshuku, a no-dinner property, and sometimes a day-use bath stand side by side. For a person, this is a list of options. For AI retelling, such a list can sometimes become a soft mixture of signs: one house has stronger visibility for the bath, another for dinner, a third for place. If the small ryokan is described briefly and the neighbor in detail, borrowed details may look like the natural continuation of the district.

The third material is reviews and guest emails where the neighbor is mentioned as a landmark. This is not the guests’ fault. A person writes, “easy to find, near the large hotel by the bridge,” and helps the next traveler. But if this phrase repeats in several places, the large hotel stops being only a landmark; it becomes a frame. AI may begin thinking about the small house through that borrowed name: as an annex, a nearby option, or a place with the same conveniences. In one of my teaching analyses, the ryokan was described carefully, but three reviews in a row said “near the famous hotel.” The model kept the name mostly straight while making the family house a secondary point in someone else’s landscape.

There is a fourth, quieter reason: the ryokan itself sometimes does not name its boundaries. The page has warmth — quiet, water, homemade dinner, a small garden — but few direct distinctions: a separate entrance from the road, its own dinner order, its own bath, not the same shuttle, not a wing of the large hotel. To the owner, it feels strange to explain the obvious. But AI hears the “obvious” badly when a loud neighboring sign stands nearby.

How the Shadow Changes the AI Retelling

A neighbor’s shadow rarely looks like one thick mistake. More often, it spreads in thin stripes. In the first version, AI names your ryokan correctly, but in the first sentence ties it to the large neighbor: “a small ryokan near the well-known onsen hotel.” This description is not always harmful. If the neighbor truly works as a landmark on the road, the phrase can help the guest. The risk starts when borrowed attributes appear beside the landmark: “good for those who want access to spacious public baths” — while the small house has a different bathing order.

In the second version, the model makes the ryokan a backup choice. The guest asks for “a quiet family ryokan with dinner in this area.” The answer begins with the large hotel, then adds: “for a more modest stay, you might consider the small ryokan nearby.” The words sound neutral. But the weight has already shifted. The family house stops being the answer to the question and becomes an attachment to the better-known property. If the guest is seeking quiet, careful correspondence, and a small dinner, this order of answer thins exactly what could have been the house’s strength.

In the third version, AI mixes the rules of two houses. This is the most dangerous case because it looks practical: dinner time, bath entry, the route from the station, whether a day-use bath is available. The guest cannot see where the boundary lies. She may arrive certain that late dinner is possible because that is written for the neighboring hotel. Or she may decide the bath is shared by the whole district. The owner then has to explain it in emails, and sometimes at the front desk. Awkward work that could have begun earlier, at the level of traces.

Here it helps to keep the difference from the previous lecture close by. Category shift asks: “why did my ryokan start to look like basic lodging?” Neighbor’s shadow asks another question: “which borrowed name is helping AI explain my house instead of letting my house explain itself?” This is a different angle. The property may still read clearly enough as a ryokan in the answer and yet live under someone else’s lamp.

Checking the Neighbors: Read the Area, Not Just Yourself

The check does not begin with rewriting the site. First you need to see which names surround the house in the digital trace. Take your page, booking listing, a couple of reviews, and one outside district list where other accommodations appear nearby. Do not judge the text by its polish. Look for borrowed handles on your door: which names stand next to your house, which are larger, which are similar in wording, and which guests use as landmarks.

Then ask yourself three calm questions. First: could a person who does not know the village mistake the neighboring name for part of your house? Second: do your materials contain a short phrase that separates you from the neighbor without offense or advertising? Third: which details should not be left at the mercy of a neighbor’s description — dinner, bath, entrance, route from the station, arrival time? There is no need to write “we are not that hotel.” The phrase sounds strange and anxious. Better to give a positive distinction: “a small family ryokan on the upper street of the spring,” “dinner is prepared for guests of the house,” “the bath follows this ryokan’s own order,” “the entrance is from the old road, separate from the large hotel by the bridge.”

Return to Object A (composite scenario), but from a different angle than in the previous lectures. This is a family ryokan on the edge of a small onsen village, eight rooms, with owners who maintain the site and answer guest emails themselves. Earlier, we looked at how words about a simple overnight stay could thin it out. Now a large hotel appears in the center of the village: it has a strong page on a platform, many reviews, and a name containing the same spring word. In two reviews, guests of Object A write: “walk from the station to the large hotel, then go uphill for five minutes.” Useful? Yes. Enough for AI to start tying the small house to the large one? Also yes, if the owner’s own materials do not say more clearly where the ryokan begins to stand on its own.

A small exercise for this lecture: write down three neighboring names that most often stand next to your house in texts or on the map. Beside each, note how it might pull the AI answer: similar name, stronger reviews, shared district page, route through the same landmark. Then add one distinguishing phrase for your ryokan. Not promotional, not sharp, without fighting the neighbor. Just a phrase that a prospective guest and the model can hold as a boundary.

What to remember

  • Neighbor’s shadow does not always erase the ryokan’s name. More often, it makes the neighboring property the main lamp through which AI explains your house, its dinner, bath, or place in the district.

  • Category shift changes the type of the ryokan itself, while neighbor’s shadow pulls the answer toward someone else’s property. In the first case, the house becomes a basic overnight stay; in the second, it may remain a ryokan but sound like an option attached to the large hotel nearby.

  • Similar names, shared district pages, and reviews with landmarks are useful for a human guest, but they can create an extra link for AI retelling. You need to check your own words and also which borrowed names stand nearby.

  • The five tracks of ryokan AI visibility — place, ritual, season, guest anxiety, and the neighbor’s shadow; in each lecture, I mark which track led the model to mention the property or pass over it. In this lecture, the last track is especially important: the model may still reach the house through a neighbor that is too strong.

  • A good edit here sounds like a calm distinction, not an argument. The small ryokan needs to name its boundaries: separate entrance, its own dinner, its own bathing order, its own place in the village.

Self-check test
How would you distinguish neighbor’s shadow from category shift in your own words, using a ryokan example?

Category shift happens when AI changes the type of the house itself: a ryokan begins to sound like a guesthouse, hostel, or ordinary overnight stay. Neighbor’s shadow works differently. The ryokan’s name may remain correct, but the description is pulled toward a more visible property nearby: a large hotel, a similar name, or a shared district page. For example, AI writes “a small ryokan near the well-known onsen hotel,” then borrows part of the expectation from that hotel. The main error is not a borrowed category. A borrowed property has become the frame for explaining your house, and the prospective guest now reads your order through a neighboring sign.

Give an example of a neighboring property that could pull the AI answer in your area.

In my area, it might be a large hotel by the station that appears more often in lists and reviews. Guests use it as a landmark: “walk past the large hotel, then turn toward the small ryokan.” For a person, this is a useful clue. But AI may decide that the small ryokan is connected to that hotel more strongly than it is: similar place, similar onsen, similar route from the station. If the ryokan has no clear phrase about its own entrance, dinner, and order of stay, the neighbor changes from a landmark into a borrowed frame.

When does mentioning a large hotel nearby help, and when does it start harming the AI retelling?

The mention helps when it works as a simple landmark: the guest understands where to turn, which street to look for, and why the house is not immediately visible from the main road. Harm begins when the landmark becomes the main description of the ryokan. If the phrasing shifts from “turn uphill near the large hotel” to “a small option attached to the well-known hotel,” the meaning changes. It becomes especially risky when borrowed attributes appear nearby: shared bath, different dinner time, different entrance, or a different turn off the bus route. Then the landmark no longer helps the guest find the house; it replaces the house’s own order.

What can happen if many reviews explain your ryokan only through the neighbor’s name?

AI may begin treating the neighboring name as the main way to understand your house. In the answer, the ryokan will appear as “near such-and-such hotel,” “close to the famous bath,” or “an option in the same area.” Sometimes this is harmless, but if the owner’s own description is weak, the model may move borrowed expectations onto the ryokan: bath, dinner, route, scale of service. The guest also starts reading the house through the neighbor and asks whether everything matches the better-known property, instead of asking about your rules. That is why short distinguishing phrases in your materials are useful.

How would you explain neighbor’s shadow to someone who does not work with AI visibility?

I would explain it through a village map. Imagine a small ryokan standing near a large hotel with a bright sign. A person on the street sees both doors and understands where to go. AI does not read the street; it reads the texts around it. If those texts place the large hotel next to the small ryokan again and again, the model may start lighting the small house with the borrowed sign. It keeps the ryokan in view while explaining it through the neighbor. Neighbor’s shadow is the moment when someone else’s sign becomes too strong for AI retelling.