Family Storage Is Now Open for Beta Testing — Official Launch on September 14

On August 28, 2026, the fully renewed family service Family Storage opened for beta testing. The official launch is scheduled for Monday, September 14.

It took about three weeks from the day we were commissioned on August 6. In our previous report we promised to write again once the work was complete, so here is the rest of the story.

The new site is at familystorage.jp.

Nothing that was recorded has been lost

This was our first concern. Everything our members had built over the years — family names, the connections in each family tree, photographs — was carried over to the new system as it was. The domain remains familystorage.jp. The same entrance, the same records.

What is new

The family tree was rebuilt from scratch

Previously, adding one relative could rearrange the whole diagram, moving cards you had already placed. The new tree arranges the family automatically and then keeps calculating until no two cards overlap. If you would rather place them yourself, drag the cards with your finger and save the arrangement.

You can change who the tree is centred on. Move the centre to any card: centre it on your mother and the tree extends into her family; centre it on your child and it extends towards your grandchildren. If you record maiden names, chips such as "the ○○ family" appear above the tree, and one tap moves you to that household.

Printing and PDF export are supported. Fit it onto one A4 sheet, split it across several A4 sheets to be taped together, or export one large sheet from A3 up to A0. Even a large family can be printed at a size where the names stay readable. You can also print only the direct ancestral line.

Ancestors whose names are not yet known appear as dotted cards, so you can see at a glance where the record stops — and therefore what to look for next.

Each person's record now holds far more

Beyond names and dates, you can record what the person was like. The screen is divided into About, Education & Career, Marriage, Health, Grave, Notes and Photos.

"About" holds the origin of the name, personality, a motto, favourite words, a person they admire, foods they love and dislike, hobbies. It is a place for the things you can only ask while they are still with you.

Health records accumulate by measurement date, so height, weight, waist and blood pressure can be traced over time. BMI is calculated automatically and judged by the Japan Society for the Study of Obesity criteria. Illnesses, allergies and check-up results are kept separately. The grave record holds the site, address, plot and posthumous name, and entering the address shows a map.

Dates of birth are shown alongside the Japanese era year, and the zodiac sign is calculated automatically. For those who have passed away, the age shown is the age at death, not the age they would be now. Addresses can be looked up by Japanese postal code or Korean road-name address.

Anniversaries are found and announced automatically

Birthdays, memorial days and wedding anniversaries are picked up from what you have already entered. There is no need to register them again. Correct a date of birth and the reminder moves with it.

Reminders arrive ahead of the day, more than once, through notices on the site, phone notifications and email.

Google Calendar is supported. Each member gets a private address; add it to your calendar and the anniversaries appear year after year. How many days ahead to be reminded is decided in your own calendar.

An AI storyteller speaks from the family record

It reads the family record and answers questions such as "who is the oldest generation here?" or "what was my grandfather like?" It reads aloud as well as writes, so you need not keep watching the screen. Japanese, Korean and English are supported.

The storyteller runs entirely on our own server. Family names and relationships are never sent to an external AI service.

A guide answers on the spot

A guide (chatbot) sits at the bottom right of the screen. It has been given the contents of the site in advance, so it can answer questions about pricing, how to apply, or how to use the family tree. Anything it cannot answer, or anything urgent, is passed on to a member of staff.

Built to work comfortably on a phone

Larger type, wider buttons. The family tree can be pinched to zoom and the cards dragged into place. Add it to the home screen and it opens in one tap, like an app.

The interface is available in Japanese and Korean.

So that search — and AI — can find it

However carefully a service is built, nothing begins until people can find it. Alongside conventional search engine optimisation, we treated answer engines (AEO) and generative AI (GEO) as separate problems and addressed each.

For search engines

The site is offered in Japanese and Korean, so every page carries a pointer to its counterpart in the other language, allowing search engines to show the Japanese page to someone searching in Japanese and the Korean page to someone searching in Korean. The sitemap lists something over a hundred and twenty pages, and each page declares a canonical address so that the same content does not appear under several URLs.

The two-line description that appears in search results is written separately for each page. One detail deserves mention: search results are truncated by width, not by character count. A Japanese or Korean character occupies the width of two Latin characters, so counting characters as one would in English produces a line twice as long as intended, cut off mid-sentence. We therefore measure width rather than characters, and end the text at a sentence boundary.

Page content is also supplied in machine-readable form: the operating company, the site itself, the position of the current page, articles, and the research service each follow their prescribed schema.

For answer engines

Search increasingly returns the answer itself rather than a list of pages. To be chosen as an answer, text has to be written as a question.

We therefore rewrote the FAQ — over twenty entries — in the words people actually use. "Is it really free?" "How much does it cost?" "I am worried about my family's personal information." The headings were left as questions rather than rephrased as topics, and the most-read questions can be opened directly from the front page.

For generative AI

The way people look things up is moving towards conversation with AI. So that an AI can describe the service correctly, we placed an llms.txt file on the site.

It is a single plain-text guide written for machines: what this site is, and which page holds what. The robots.txt file points to its location.

We deliberately do not block AI crawlers. Blanket blocking has become common, but for a service whose name is not yet known, not being read at all means never appearing in an AI's answer either. Members' records sit behind a login and are not visible to crawlers. The decision followed from separating what may be read from what must not be.

We kept our promise

Last time we wrote that data would be stored encrypted, and that the design would not send family information to outside generative AI services. That is how it was built.

Names, addresses and phone numbers are stored encrypted. The AI storyteller runs entirely on our own server; family information is never sent to an external AI service. Family records are visible only to the family members who were invited.

What comes next

Until the official launch we will keep refining the details while the service is in beta. If anything catches your eye, we would be glad to hear about it.

We will keep tending to it carefully after launch, so that it can stay with families for a long time.

▶ Visit Family Storage (familystorage.jp)

— THEVOS ECO Institute Co., Ltd.