Assessing a Companion App You Have Never Heard Of
New apps arrive in this category constantly, and with a new one every ordinary signal is missing. There is no history of the operator keeping its own terms, no deletion flow anyone has tested, no evidence the thing will exist in a year, and no reputation it would lose by behaving carelessly. That absence, rather than any specific suspicion, is what you are working around — so the assessment is about the operator and the shape of the business, not about the feature list.
That is fortunate, because the operator is the part you can actually check in ten minutes, and the features are the part that will have changed by the time you have finished evaluating them.
Ten minutes on the operator
Find the legal entity. The terms or the policy should name a company and a place. A document that names nobody is telling you that no one has accepted responsibility for it, and that is the single most useful negative signal available.
Read the policy for signs of a template. Filled-in boilerplate is normal and fine; a placeholder left unreplaced, a clause referring to a different kind of product, or a policy describing a website when the product is an app all indicate that nobody has thought about the specific handling of what you are about to type into it.
Check whether the six clauses that matter are present at all. Retention, training use, human review, processors, change of control, and a deletion route — the set covered in how to read a companion app’s privacy policy. A new operator that has bothered with all six has told you something about its intentions.
Find a support channel and use it. Send one question before you sign up. Whether a reply comes, how long it takes, and whether it reads as a person are worth more than any store rating.
Establish who takes the money. Store billing gives you a cancellation path and a refund process that do not depend on the operator’s goodwill; a direct payment page does not. This matters more with an unknown operator than with an established one, and the distinction is set out in who you actually bought the subscription from.
Confirm the app is the app. New and unknown is exactly the profile that clones imitate, and the imitation is usually of a well-known app rather than an obscure one. Spotting a lookalike companion app covers what to compare.
The thin-wrapper question
A great many new entrants are a front end over somebody else’s conversational service. This is a legitimate way to build a product and it has two consequences worth knowing about.
The first is that the material you type passes through a party the app may not have named, which is precisely why the processor clause in the policy is worth finding. The second is durability: a wrapper’s continued existence depends on an upstream arrangement it does not control, and upstream terms change more often than companies fail.
The tells are ordinary rather than technical. A very small team with a very broad feature set on day one. Behaviour indistinguishable from other apps in the category, including the same characteristic phrasings. Terms that mention a third-party provider. Outages that do not correlate with anything the operator is doing. None of these is damning; together they suggest what you are actually buying is an interface, and the interface is the part that will be maintained.
What not to base the decision on
The launch feature list. Feature lists are the cheapest thing to write and the first thing to change.
Terms that are unusually generous at launch. That is a customer-acquisition decision, not a commitment, and it is normally the first thing revised once acquisition is no longer the priority.
Early reviews. A new listing’s reviews are the least informative reviews it will ever have, for reasons that have nothing to do with honesty: there is no long-term use to report on yet.
Any current capability claim. What a conversational product can do is a moving target across the whole category, so a claim about capability is a claim about this month. Judge how the app is structured and how the operator behaves, both of which are considerably more stable.
Commit in a way you can reverse
The practical answer to an unknown operator is not to avoid it but to keep the first month cheap to walk away from.
Attach as little identity as the app allows — a dedicated email address rather than a platform sign-in, no optional permissions, store billing if there is a choice. The reasoning is in what a companion app operator actually holds.
Do not build continuity you would mourn. Retained material is not portable and generally not exportable, so treat the first weeks as a trial rather than a foundation. When a companion app stops remembering covers how easily it goes.
Test the exit early. Find the account-deletion flow and read what it says it does before you need it. An app with no discoverable deletion route has answered a question you were going to ask later anyway.
Re-look after the first substantial update. New apps change fastest, and the first release after launch tells you what the operator considers negotiable — see what an app update can change without asking.
What only time can tell you
Whether the operator honours its retention promise. Whether the deletion request did anything. Whether support is still answering in six months. Whether the terms you signed up under survive the second funding decision. None of these is knowable at install, and no checklist substitutes for them.
Which is the honest conclusion: a new app cannot be assessed the way an established one can, and the correct response is a smaller commitment rather than a more elaborate evaluation. Check that a real entity stands behind it, that its documents were written for it, that you can get your money back and your account closed, and then let the first few months supply the evidence a listing page never could.