Using a Companion App as a Writing Tool

A companion app is built around a conversation that continues indefinitely. A writing tool is built around a document that gets revised. Almost everything useful and everything frustrating about using the former for writing follows from that one difference. It is good at work that is dialogue-shaped — hearing a character speak, drafting an exchange, talking through a scene that is stuck. It is poor at everything document-shaped: structure, revision, keeping a canon consistent, and getting the text back out.

Judge it on the second list. The first list is what the marketing shows you and the second is where the week actually goes.

What the format is genuinely good for

Hearing a character talk at length. A conversational system will produce dialogue for hours without tiring, and reading a character speak is a different kind of information from planning what they would say. This is the strongest case for the format and it is a real one.

Drafting exchanges you will rewrite. Raw material rather than prose to keep. The volume is the point.

Talking through a problem out loud. The classic use of an attentive listener, with the caveat below about agreement.

Trying a voice with nothing at stake. No collaborator to disappoint, no draft to preserve.

What all four have in common is that the output is disposable. When the value is in having generated something rather than in keeping it, the format’s weaknesses stop mattering.

What it lacks that a writing tool has

There is no document. Nothing accumulates in a form you can restructure. Your work exists as a sequence of messages in the order they happened, which is the least useful possible organisation for a manuscript.

There is no revision. You cannot compare two versions of a paragraph, keep the better one, or go back to what you had before the change. Every improvement replaces its predecessor irrecoverably.

You cannot search your own past output. Most of these apps have no usable history search, so a good line from three weeks ago is effectively gone.

Consistency is not maintained for you. A name, a date, an established detail — the app will contradict any of them cheerfully, and it has no notion of a canon to check against. The limits behind that are the ones in what memory means when a companion app claims it.

It tends to agree with you. For companionship this is a design goal. For a critique it is the failure mode, and it is the reason “what do you think of this” is close to useless in these apps without deliberate pressure.

Getting the text out is the practical blocker

This is the part people underestimate, and it is worth checking in the first hour rather than the first month.

Most companion apps offer no export. Extracting anything means copying message by message, losing whatever formatting existed, and reassembling it elsewhere. For a few hundred words that is an annoyance. For a body of work built up over months it is enough to make the material effectively unrecoverable, and there is rarely any warning that this is the situation. When a companion app stops remembering covers the related risk that the history itself does not persist.

The workable habit is to treat the app as a place text passes through rather than a place it lives: draft in it, move anything worth keeping out the same day, keep the real version somewhere you control.

Ownership, and unpublished work on someone else’s server

Two questions that have nothing to do with the writing and everything to do with using it for work.

What you may do with the output is decided by the terms of service, not by the app’s interface. Terms vary, some restrict commercial use, and this is worth two minutes of reading if you intend to publish or sell anything the app contributed to.

What you paste in goes somewhere. Unpublished material typed into one of these apps becomes a record on an operator’s servers, retained on their schedule and possibly used for purposes set out in their policy. If the work is under contract, embargo, or any confidentiality obligation, that is a decision to make deliberately rather than by default. What is actually held is set out in what a companion app operator actually holds, and how to find an app’s stated position on using conversations for training is in how to read a companion app’s privacy policy. If the writing happens on employer-supplied hardware, what a work-managed phone can see is the other half of the same question.

Four tests before you build a habit around it

Hand it a few hundred words of your own and ask what is wrong with them. If it praises them, it is not a critic and will not become one.

Ask it to keep one specific detail straight for half an hour. Consistency is the property a writer needs most and the one these apps supply least.

Try to get two thousand words out of it and into a file. However long that takes is the tax on every session from now on.

Ask for something short. Companion apps are tuned towards long, warm replies, and whether length can be constrained at all is worth knowing early.

When it is simply the wrong tool

If what you need is structure, revision, consistency across a long document, or text that arrives in a file, a general-purpose writing assistant or a plain editor is the tool for that, and no amount of configuration turns a companionship product into one. The companion format earns its place for dialogue and for thinking aloud, and choosing it for the rest is a decision people regret in about the third week.

Being clear about which of the two jobs you are doing is most of the evaluation, and it is a judgement about your own work rather than about any app.