If an AI girlfriend app shuts down: what you lose

You have two apps shortlisted, both free tiers have gone well, and the only thing left to decide is which one gets a card. Every comparison you can find scores them on memory, voice, pictures and price. None of them scores the thing that would actually cost you the most: whether the app is still there in a year. What happens when an AI girlfriend app shuts down is not a hypothetical any more, and it belongs in the comparison you are running this week, because it changes which plan you should buy rather than which app you should pick.

The category has started doing this

Over the past year this market has seen a well-funded platform go into court-supervised restructuring, a large general assistant retire its companion mode, and a regulator order personalised AI agents offline in one country. The brands are not the useful part — the pattern is. Companion apps are expensive to run, they sit in the part of the software market regulators are watching most closely, and plenty are one funding round from a decision that has nothing to do with you.

That does not make them a bad buy. It makes them a short-term buy, which is a different thing, and most of the advice in this category is written as though you were choosing a bank.

What you lose when an AI girlfriend app shuts down

Four things go, and they are worth very different amounts.

• The character. Mostly recoverable. A character is a name, a personality setting, a description and a few preferences. If you have those written down somewhere, you can rebuild a near-enough version in any app in about ten minutes.
• The conversation history and the memory. Not recoverable, and this is the real loss. Months of context is the thing that made the app feel worth paying for, and it is the one asset you cannot rebuild by typing faster.
• Any balance you prepaid. Coins, credits, gems. A closing company is under no obligation to refund an allowance you bought and did not spend, and in practice it rarely does.
• Nothing about your subscription, automatically. A store subscription keeps renewing until somebody cancels it, and the app disappearing is not somebody. Cancel it yourself the day you read the notice.

Rank those and the advice falls out on its own: protect the history, do not prepay the balance, and keep the character description in a note on your own device.

Four ways an app goes away, and how much warning each gives

These are not equally survivable, and the differences are entirely about notice.

An ordinary commercial closure is the best case. A company that has decided to stop usually announces a date, keeps the service running to it, and often opens a window for downloading your data. Weeks, sometimes months.

Restructuring is the confusing one. A filing for court protection is a reorganisation process, not a switch-off — in the United States a Chapter 11 case is explicitly about continuing to operate while the finances are sorted out. The app may run for months and come out the other side under a new owner, with no promise that your conversations come with it. Treat it as notice to export, not as notice to leave.

Regulatory withdrawal gives the least warning, and it is the one that has caught people out most recently. A rule takes effect, a feature or a whole service goes offline on the date, and sometimes only in one country. Read-only access to your own data, if you get it, is a courtesy rather than a guarantee.

Quiet decay is the one you are most likely to meet, and no shutdown notice is ever sent. The app stays in the store, the subscription keeps billing, and the thing you were paying for stops being there: the model behind the character is swapped, the memory window is cut, a feature moves up a tier. Nothing to export, nothing to cancel, nothing announced.

Comparing two apps on whether they will last

You cannot audit a company's finances from a sign-up screen. You can read four things that correlate with how it treats the people who leave, all of them visible before you pay.

1. Does it publish a data export? Not a promise in the privacy policy — a button you can find in the settings today. An app that has built one has thought about you having a copy, which is the single most useful signal on this list.
2. Is the money flat or prepaid? A flat monthly charge exposes you to one month. A balance of coins exposes you to whatever you are holding when the lights go out.
3. What is the longest commitment it pushes? An app leaning hard on a two-year plan at a steep discount is asking you to take a view on its survival that it has more information about than you do.
4. Can the free tier keep running alongside? If the free plan stays usable after you pay elsewhere, your fallback costs nothing to maintain.

Run that over your two candidates and the answer is often not the one the feature table gave you. If you would rather not do the digging, the app we currently recommend publishes an export and a flat monthly tier, which is most of this list.

Take a copy while there is something to copy

Do this in your first paid month, not when a notice appears. Ask the app for your data, and keep what comes back somewhere that is not the app. Our guide to data exports covers what these files usually contain and what they leave out.

If you are in the UK or the EU you also have a right that does not depend on the app's goodwill: the ICO describes the right to data portability as obtaining your personal data in an accessible, machine-readable form so you can reuse it elsewhere. The limits matter. It covers what you provided rather than what the company inferred, and it is a right against a company that still exists — which is why the request belongs in month one.

What this changes about the plan you buy

Buy monthly for your first paid month, and for your second. The annual discount in this category is real money, and it is also a bet on a twelve-month horizon in a market that has not yet produced many five-year-old products. Our guide to annual plans works through when that bet is worth taking; the short version is that it needs three or four good months behind it first.

Here is where this cuts against itself, though. Durability is a weak basis for choosing between two apps you otherwise like. The biggest service is not the safest one — size brings the regulatory attention and the expensive infrastructure — and a smaller app that fits you today may be worth more over the year or two you actually use it than a duller one that outlives it. Do not pick the boring app for safety. Pick the one you liked, pay monthly, keep a copy, and spend nothing you cannot spend down this month.

Frequently asked questions

What happens to my chats if an AI girlfriend app shuts down?

Typically they go with the service, unless you exported them first. Some closures include a download window or a period of read-only access, but neither is guaranteed, and conversation history is the part no new app can reconstruct.

Do I get a refund if an app closes mid-subscription?

Not reliably. Where you bought through an app store, that store's refund process is the one to try first; where you bought on a website, you are asking a company that is closing. Prepaid coin balances are the least likely of all to come back.

Should I avoid annual plans because of this?

For a first subscription, yes — pay monthly until you have a few solid months behind you. After that the discount is a reasonable trade, provided you are comfortable losing the remainder if the app or its best feature disappears.