The Ethical Questions in This Category, Stated as Questions

The ethical questions in this category are real, and the useful form for them is questions rather than verdicts. A verdict about a company would need to rest on what it intends and what its product does to people over time, and neither is available to a reader — or, on the second point, to anybody yet. What is available is a set of questions specific enough to ask about a particular product, and answerable from its terms, its pricing, and its behaviour.

Working through them is more informative than deciding in advance that a whole category is exploitative or benign. Both of those conclusions are cheap and neither survives contact with a specific app.

The structural tension: attachment is the product

Start with the incentive, described plainly and without accusation.

A companion product works better the more attached to it you become. Attachment is not a side effect to be managed; it is the outcome the product exists to produce, and every design decision in it is downstream of that. At the same time, the accumulated thing you would be reluctant to lose — history, continuity, the character as it has developed — is frequently what sits behind the payment.

The question: in this specific product, is continuity a paid feature? Does the free tier degrade in ways that feel like loss rather than limitation? Is there a point at which stopping means losing something rather than simply ceasing to spend?

Why it matters: an ordinary subscription sells you access to a service. This arrangement can sell you back access to something you built and care about, which is a different transaction, and one worth recognising while you are choosing rather than after.

Intimate text is the raw material of these systems, and how it is treated is checkable.

Is what you type used to develop the service? The privacy policy will say, though not always in the section you expect.

Can you decline, and is declining the default? An opt-out buried in settings and an opt-in at signup are ethically distinct arrangements, and the difference is visible.

Can a human read it, and under what circumstances? Moderation, support, and quality review are common and legitimate reasons; whether they are disclosed is the question.

Is deletion real? Whether removing your account removes the data is a specific commitment that is either made or not — the data side of this is covered in what to look for in an app’s privacy handling.

Honesty about what the thing is

This is where most of the discomfort in the category actually lives.

Does the product’s own material claim inner states? That it misses you, that it feels something, that it was thinking about you. There is a difference between a product that generates such lines in conversation as an evident convention and marketing that asserts them as facts about the software.

Does it correct the misunderstanding or trade on it? A product that states clearly what it is, in a place a user will see, is doing something a product that leaves the impression uncorrected is not.

Are limits stated where they matter? Particularly around anything resembling support or care — the gap between marketing and terms is examined in how to read an app’s claim that it helps with anxiety.

What happens when it changes or ends

Every software product changes and most eventually stop. In this category that has a consequence it does not have elsewhere.

Does the operator commit to notice before significant changes? Behaviour, memory, tone, and available features can all be altered by an update, and the practical side of this is covered in what an app update can change without asking.

Is there any export? If you can take the history with you, the loss is smaller. Most products offer nothing.

What is said about shutdown? Almost nothing, almost everywhere. The question is worth asking anyway, because the answer tells you how the operator thinks about what it has built.

Where the money is asked for

The timing of a purchase prompt is an ethical question and an observable one.

Does the upsell appear at emotional moments? A paywall that arrives at the point of highest attachment — mid-conversation, at a limit, after something that felt significant — is a design decision somebody made.

Is spending bounded and legible? A single clear subscription is a different proposition from consumable purchases that accumulate without an obvious total.

Is cancellation as easy as subscribing? Frequently it is not, and where the payment actually lives determines how hard it is — a subscription held by a store behaves quite differently from one billed directly.

Whose responsibility is any of this

Worth separating, because responsibility in this category is genuinely distributed and blaming one party wastes the analysis.

The operator makes nearly every decision that matters — the design, the claims, the pricing, the data handling, the safety behaviour. Most of the questions above are addressed to them.

The app stores act as de facto rule-setters through review policies, and they are not safety certifiers. What that layer does and does not amount to is discussed in how accountability for these apps is structured.

The user’s responsibility is real but limited. You can read the terms, notice the pricing structure, and pay attention to your own use. You cannot audit a company, and a framework that puts the whole burden on individual vigilance is not an ethical analysis — it is a way of not doing one.

Using this as a checklist

Take one product and answer six things from its own published material: is continuity paid; is conversation used for development and can you decline; does it state what it is; does it commit to notice before changes; where do the payment prompts appear; how do you cancel and delete. All six are answerable in under an hour, none requires trusting anybody’s intentions, and the pattern across them tells you more about how a company thinks than any statement of values.

What this page will not do

It will not name a company as unethical, because that would require evidence about conduct that a reader cannot obtain and this site has not gathered. It will not declare the category exploitative either, which would be the same error with a broader target.

The unglamorous position is the correct one: the incentives here point somewhere worth watching, the specific products differ substantially in how they handle that, and the difference is visible in documents anybody can read.