How to Make Money in Japanese Real Estate Using AI

The short answer

Japanese property is the rare developed market where a habitable house still sells for the price of a used car, and where the numbers that decide whether it makes money are public, structured and boring enough for a machine to grind through.

Engawa's MCP server puts that catalogue inside Claude or any other AI client. You describe the deal you want in a sentence, and the assistant filters thousands of listings, pulls the full record on the survivors, runs Japanese closing costs on each, and hands you a shortlist. What used to be a weekend of tab-juggling is now one prompt.

This article is the prompts. Copy them.

Why the numbers work in Japan

Three structural facts drive every play in this article.

Buildings depreciate to zero, land does not. Japanese tax convention writes a timber house off over 22 years, and the market broadly agrees with it. A 40-year-old farmhouse on 400 m² is priced as land with a free building attached. That is why a structurally sound kominka lists at ¥3,000,000 while the plot alone would cost more in almost any other country.

There are nine million empty houses. The akiya count keeps climbing as rural Japan ages, and municipalities are actively motivated to move them. Supply is enormous and the buyer pool is thin.

The yen has been cheap for years. A dollar, euro or Singapore-dollar earner buys into a yen-denominated asset with yen-denominated renovation costs and, if the tenant is domestic, yen-denominated rent. If the currency mean-reverts, that is a second engine on the same purchase.

None of that is secret. What has always been hard is the search: tens of thousands of listings, in Japanese, scattered across regional agent sites, with the fields that actually decide profitability buried three clicks deep. That is a data problem, and data problems are what these tools are for.

Play one: the renovation flip

The classic play. Buy an as-is house well under land value, spend on the parts that gate a mortgage or a rental licence, sell or refinance into the repaired figure.

The whole game is finding the house where the gap between asking price and post-renovation value is widest, and where the renovation is cosmetic rather than structural. Two properties can look identical in photographs and differ by ¥8,000,000 in the work they need.

Here is where the assistant earns its money. Ask it:

Search Engawa for as-is houses in Chiba, Ibaraki and Tochigi under
¥4,000,000 with at least 250 m² of land. For each result, tell me
the seismic standard, the land-to-price ratio, and how far it is
from the nearest station. Drop anything built before 1981 unless
the land is over 500 m². Give me the top eight as a table.

Then narrow, in the same conversation:

Pull the full record on the top three. I want structure type,
water and sewage, land rights, and what the source listing says
about condition. Flag anything that suggests septic tank rather
than mains sewage, because that changes my renovation budget by
about ¥1,500,000.

And then the part almost every foreign buyer gets wrong:

Run estimate_closing_costs on all three at their asking prices.
Show me acquisition tax, registration, scrivener and commission
separately, and tell me what percentage of the purchase price the
closing costs represent for each.

That last one is the whole reason to automate this. Japanese closing costs are not a flat percentage. They are largely fixed fees plus taxes on the municipal assessment, so on a ¥2,000,000 house they can run to 15 or 20 percent of the price, while on a ¥40,000,000 house they are closer to 6. Buyers routinely model an akiya flip on a flat 7 percent and lose the margin at the notary. The tool does this arithmetic per listing in seconds.

Play two: the yield hunt

Less glamorous, more repeatable. Buy cheap, renovate to rentable, hold for cash flow.

Rural Japanese rents are low in absolute terms and enormous relative to these purchase prices. A house bought at ¥3,000,000 and renovated for ¥4,000,000 that rents at ¥60,000 a month is running a gross yield north of 10 percent on total capital. The catalogue carries a published area rent benchmark for comparable dwellings where a statistic covers the area, and it tells you how many dwellings sit behind that average, so you can see whether you are looking at a figure drawn from 40,000 homes or from 600.

Find me houses in Nagano and Yamanashi between ¥2,000,000 and
¥8,000,000, already renovated or in good condition, within 2 km
of a station. For each, give me the area rent benchmark and how
many dwellings it is based on, then rank them by benchmark rent
divided by asking price.

Then stress the ranking:

Take the top five and add closing costs to each purchase price.
Recalculate the ratio on the all-in figure rather than the sticker
price. Show me how much the ranking changes.

It usually changes a lot. That is the point.

Play three: land banking near the hubs

The quiet one. Buildings go to zero, land does not, and the land near regional employment centres is the land with a floor under it.

Every listing carries straight-line distance to the two nearest regional centres rather than one, which sounds like a detail and is not. The nearest hub by ruler is frequently not the relevant one, because the rail line running through a town goes where the line goes. A house in Isumi is marginally closer to Yokohama than to Tokyo while the train goes to Tokyo, and a strategy built on the ruler alone buys the wrong side of the peninsula.

List every prefecture in the catalogue with its listing count.
Then search for land and as-is houses under ¥3,000,000 within
60 km of a regional centre, in prefectures with more than 40
active listings. Give me both hub distances for each result,
not just the nearest.

Of those, which sit in municipalities that appear more than once
in the results? I want to see where the catalogue is deep enough
that I could buy two or three adjacent plots over time.

Play four: the long-stay conversion

Renovate to a standard that suits a three-to-six-month guest rather than a permanent tenant, and let the property earn from a different pool. Engawa runs a long-stay side precisely because the akiya that make bad permanent rentals often make excellent seasonal ones: a kominka an hour from a ski field, a coastal house that is empty nine months of the year.

Find kominka and traditional houses in Hokkaido, Nagano and Niigata
under ¥10,000,000 with at least 120 m² of building area. I want
character properties near mountains, and I want to know for each
one what the listing says about heating and insulation.

Heating and insulation is the right filter here. A guest paying for a winter month has a very different tolerance for a cold house than a farmer who grew up in one.

The prompts that separate a good search from a bad one

Four patterns are worth learning, because they change the quality of what comes back more than any filter does.

1. 1

   Ask for the table, not the prose.Say "give me the results as a table with columns for price, land area, seismic standard and station distance". You get something you can scan and paste into a spreadsheet instead of eight paragraphs of description.

2. 2

   Chain the tools in one conversation.Search, then pull full records on the survivors, then run closing costs on the finalists. Each step narrows the set, and the assistant carries the shortlist forward without you re-typing slugs.

3. 3

   Give it your economics.Tell it your renovation cost per square metre, your target yield, your ceiling on total capital. It will apply those to every result rather than handing you generic rankings.

4. 4

   Ask what it cannot tell you.Ask "what do you not know about this listing that I should find out?" and it will name the gaps: no survey, no municipal assessment, no timetable. That list is your due-diligence agenda for the agent.

That fourth one deserves a sentence of explanation, because it is the least obvious and the most useful. The tools are built so that missing information is written out in words rather than left as an empty field. Ask about the commute and you get the map distance plus a flat statement that no timetable data is held. Ask about the assessed value and you are told it lives on the tax certificate the seller holds, and that every purchase tax is levied on it rather than on the price. That turns into a checklist you hand your agent on day one, which is worth more than a guess would have been.

What this replaces

The honest comparison is not against a professional buying agent. It is against what a foreign buyer actually does today, which is: open six regional listing sites in Japanese, machine-translate them one page at a time, keep a spreadsheet, forget to check the seismic year, model closing costs as a flat percentage, and give up around week three.

Against that, an assistant that filters the whole catalogue on your criteria in one message, costs the shortlist correctly, and tells you what to ask the agent is not a small upgrade. It is the difference between running a search and abandoning one.

Getting connected

1. 1

   Take annual Explorer.The AI catalogue is on annual plans only, at $99 a year.

2. 2

   Create an API key.From your account page. Name it after the device it will live on. It is shown once and stored only as a hash.

3. 3

   Paste the snippet.The account page generates ready-made JSON for Claude Desktop, Claude Code or any MCP client on HTTP transport.

4. 4

   Start with a prompt from this article.The Engawa tools appear in your client's tool list after a restart. Paste one of the searches above and change the prefectures.

Sources and review notes

Rent benchmarks are published area averages for comparable dwellings, not projected rent for a specific house. Closing-cost figures are estimates until the municipal assessment certificate is in hand. Engawa aggregates listings and refers buyers to licensed local agents; it is not a broker and none of this is investment, tax or legal advice.

• Statistics Bureau: Housing and Land Survey
• MLIT: building standards and seismic retrofit guidance
• National Tax Agency: depreciation and property taxation
• Model Context Protocol: specification and client list

Editorial ownerEngawa Editorial Team

MethodEngawa catalogue tooling, Japanese tax and building-code references, national housing statistics

Checked on6 August 2026

This article is general information, not personalized legal, tax, structural, financing, visa, or investment advice. Property and service availability can change.

Frequently asked questions

Can you actually make money buying cheap Japanese houses?+

Yes, through four recognised routes: renovation flips, where the purchase sits well below land value and the repair work is cosmetic rather than structural; rural rental yield, where a total capital outlay of six or seven million yen can meet a rent in the tens of thousands per month; land banking near regional employment centres, where the building depreciates but the plot does not; and long-stay or seasonal conversions, where a property earns from guests rather than permanent tenants. The constraint is never finding cheap houses, of which Japan has millions. It is finding the ones where the numbers survive closing costs and renovation.

How can an AI assistant help me find Japanese investment property?+

Through Engawa's MCP server, an AI client such as Claude can query the live property catalogue directly. You describe the deal in a sentence and it filters by prefecture, price, land area, building age and renovation state, pulls full records on the survivors, compares them side by side, and runs Japanese closing costs on each. It returns results as a table you can rank on any figure you care about, including ratios it computes for you such as rent benchmark against all-in cost. It is included with annual Explorer membership.

What is the best prompt for finding akiya investment deals?+

Chain three prompts rather than writing one long one. First, a broad filter: as-is houses in named prefectures under a price ceiling with a minimum land area, returned as a table with seismic standard and station distance as columns. Second, a detail pull on the top few: structure, water, sewage, land rights and what the source listing says about condition. Third, closing costs on the finalists, shown as a percentage of purchase price. Each step narrows the set and the assistant carries the shortlist forward without you re-entering listing slugs.

Why do Japanese closing costs matter so much on cheap property?+

Because they are largely fixed fees plus taxes levied on the municipal assessment rather than a flat percentage of the price. On a forty million yen apartment they land near six percent. On a two million yen akiya they can reach fifteen or twenty percent, since the judicial scrivener, registration and acquisition tax barely shrink as the price falls. Investors who model an akiya flip on a flat seven percent lose the margin at the notary. Running the estimate per listing before making an offer is the single most valuable calculation in the whole search.

Do Japanese houses really depreciate to zero?+

For tax and, broadly, for market purposes, yes. Japanese convention depreciates a timber house over twenty-two years, and buyers price accordingly, which is why a forty-year-old farmhouse on four hundred square metres is effectively priced as land with a free building attached. This is the mechanism that creates the opportunity: the structure can be entirely sound while carrying no value in the asking price, so renovation spending converts directly into a value that the original purchase never paid for.

What is the 1981 seismic standard and why does it affect returns?+

Japan revised its building code for permits issued from 1 June 1981. Buildings meeting the newer specification are shin-taishin; earlier ones are kyu-taishin. It matters commercially rather than just structurally: it determines whether a Japanese lender will finance the buyer you eventually sell to, how insurers price the building, and whether earthquake-resistance tax reductions apply. A pre-1981 property is not unbuyable, but it is a narrower resale market, and every Engawa search result carries the label so a shortlist can be filtered on the cutoff immediately.

What rental yield can you get on a rural Japanese house?+

Rural rents are low in absolute terms and high relative to these purchase prices. A house bought at three million yen and renovated for four million that lets at sixty thousand yen a month runs a gross yield above ten percent on total capital. Engawa carries a published area rent benchmark for comparable dwellings wherever a statistic covers the area, along with the number of dwellings behind that average, so you can judge how much weight it deserves. It is an area average and not projected rent for a specific house.

Do I need an annual membership to use the AI tools?+

Yes. The AI catalogue is included with annual Explorer at ninety-nine dollars a year, and monthly Explorer does not carry it. A monthly member can create an API key and connect a client, but calling a catalogue tool returns an upgrade notice instead of listings. Switching from monthly to annual takes effect immediately.