Every hour you spend building a relationship with one of these products is an hour invested in a single company's database.
Suppose you have used one of these for six months. The character has a personality you shaped, it knows a hundred things about you, and the conversation has a history. Now suppose the platform raises its price, changes its content policy, replaces the model, or shuts down.
You cannot move any of it. Not to a competitor, not to a file, in most cases not even to a readable archive. There is no interchange format in this category, nobody is building one, and the companion you spent six months with exists only as rows in one company's database.
This is not a scandal β it is the ordinary condition of every consumer product built on a proprietary model. It is worth stating plainly because the emotional weight people put into these makes it feel like something they own, and it is not.
The character definition transfers, because you can rewrite it. Appearance, traits, speaking style, backstory: those are things you know and can type into a new creator. On a platform like Janitor AI, where you write the definition as free text, this is close to a genuine copy.
What does not transfer is everything the companion learned after that. Every fact it picked up about you, every callback, every shared reference β that lives in the platform's memory store, in a proprietary format, and there is no interface that emits it.
So a migration reproduces the character on day one and loses the relationship. On the platforms that lead this site for memory β Nomi AI at 9.3, Kindroid at 8.8 β that accumulated history is most of what you were paying for, which makes them, ironically, the ones with the most to lose in a move.
The commercial reason is obvious: switching costs are the business model, and a platform that let you export your companion would be funding its own churn.
But there is a real technical reason underneath it. A memory store is not a document β it is a set of extracted facts with embeddings, weights and retrieval behaviour tuned to one specific system. Even with total goodwill, exporting it to a competitor would produce something the receiving platform could not use, because the receiving platform's retrieval works differently.
The honest version is that portability here would mean an agreed standard for what a companion is, and the category is nowhere near stable enough for that conversation. Expect this to remain true for years.
The pricing guide argues for paying monthly for the first month on financial grounds β annual discounts here run 50 to 65 percent, and the risk is the platform rather than the money.
Lock-in sharpens that argument considerably. A twelve-month commitment is not just twelve months of payment; it is twelve months of accumulating something you cannot take out. By month eight you are not staying because the plan is paid, you are staying because leaving costs you the relationship.
Replika demonstrated the failure mode in 2023 when it removed erotic roleplay from existing companions with no warning. Annual subscribers were not merely paying for something they could no longer use β they were holding a character that had changed and could not be moved.
One page, in a text file, outside the platform. The full character definition as you entered it. The three or four traits. The speaking-style instruction. The backstory. And a running list, added to every few weeks, of the details the companion knows about you that you would want it to know again.
That last list is the part that feels excessive and is the only part that matters. It takes a minute a fortnight and it converts an unrecoverable loss into an afternoon of setup.
It is also useful for a reason that has nothing to do with migration: writing down what your companion knows about you is a fast way to notice how much you have told it, which is a question worth asking every so often anyway.
Check whether the platform lets you delete your history yourself, without emailing anyone. It is thirty seconds in the account settings, it is the clearest signal of how a company thinks about your data, and a platform that never built the control will be slow to honour a deletion request.
Then assume you will move eventually, because in a category this young you probably will, and set up so that when it happens you lose an afternoon rather than six months.
Financially, almost always β the discounts are steep, often 50β65%. The risk isn't financial, it's that this category is volatile. Platforms swap the model underneath you, remove features, and occasionally close. We've watched a companion's personality change overnight on more than one platform. Committing twelve months to something with no track record is a different bet from committing to one that's been running for years, and we'd weigh that before the discount.
Cancel inside the platform's own account settings first, and keep the confirmation. Refund policies vary and annual plans are frequently non-refundable, so read the terms before committing rather than after. If a platform will not respond and you believe the charge is wrong, a card dispute is the fallback β but treat it as a last resort, because chargebacks generally end the account and any characters on it.
It depends entirely on the platform, and it is one of the sharpest differentiators between a serious operator and a careless one. If you are in the EU, EEA or UK, the GDPR gives you a right to erasure that applies regardless of what the interface offers, and a platform that ignores a valid deletion request is breaking the law. Before subscribing, check whether deletion is self-service or requires emailing support β the difference predicts a lot.
We signed up to all fourteen platforms without paying. Three free tiers are genuinely usable, most are demos that stop exactly where it gets interesting, and two platforms have nothing at all.
Read postEvery platform in this category calls itself uncensored and every one of them refuses things. Here is where refusals actually come from, which limits are permanent, and what the word is worth when you see it on a pricing page.
Read postThe same generator is almost never good at both styles, and picking by style narrows fourteen platforms to two or three faster than any score does. What each end of the category is actually good at.
Read post
