Our written no-AI policy, and the cost of holding it

StrongBrandWords has a written, public, no-AI policy on production copy. It is published on this site, signed by the four people in the studio, and updated when the boundaries of "AI" or "production copy" shift in a way that we want to be clear about. It costs us work. It also defines, in a way that nothing else has so far defined, what we are.

The policy is short. No piece of writing that leaves this studio has been drafted, edited, or assembled by a language model. We use AI for transcripts of recorded calls, for early-stage research summaries, and for occasional internal scaffolding — the studio's own emails, the internal handoff notes, the bookkeeping. The line is the production line: the writing the client pays for is written by a human in this studio. There is no statistical compromise on this and there is no exception for tight deadlines.

The policy costs us work in two visible ways. It costs us pricing competitiveness against studios that use models to produce a first draft and an editor to clean the draft up; their cost structure is fundamentally cheaper than ours and they pass some of that on to the client. It costs us, second, a kind of marketing pressure: we cannot honestly say that we are using "the latest tools" or that we are "AI-augmented" or any of the dozen other phrases the procurement form expects to see. We say, instead, that we are not.

"Our written no-AI policy, and the cost of holding it" Sara Tervo · August 2025

The policy is held because the writing is the product. If the writing has been assembled by a model and edited by a human, what the client has bought is a particular kind of editing — useful, perhaps cheap, but not the thing we believe we sell. What we believe we sell is a sentence that one of the four people in the studio wrote, that another one of the four read aloud, and that nobody else on earth would have written quite that way. That sentence has a kind of legibility to a careful reader that a model-assisted sentence does not have. We can hear the difference. So can the readers who matter to our clients.

There is a softer reason that I will name and that I think matters even more. The four of us became writers because we wanted to spend our working life thinking about sentences. The model-assisted workflow is not a workflow in which a writer thinks about sentences; it is a workflow in which a writer edits the output of a system that did not think. The work, done that way, becomes editorial supervision of a machine's prose. We have all done some of that for other reasons. None of us wants it to be the work of the studio.

I want to acknowledge two reasonable counter-arguments. The first is that models will, in time, write sentences indistinguishable from the ones we write. I do not know whether this is true. If it is true, it is not yet true today, and we will respond when the gap closes by deciding what we are then. The second is that holding the no-AI line will, in time, cost the studio its competitive position. I think this is partly true. I also think a studio that survives the next decade by becoming what it is not has not really survived; it has merely been replaced.

The policy is, in this sense, a kind of compass. It tells us where we are pointed and which work we will and will not take. It costs us the work that would have made us look more modern. It keeps us the work that made the studio worth opening in 2019.