I have written one poem in the far-western speech I grew up around rather than in standard Nepali. It is, predictably, the poem that reaches the fewest readers. It is also the only one I did not have to translate before writing down.
That second part is the thing worth examining, because it is not sentimental and it is not about authenticity. Something specific happens in the gap between the language you think in and the language you are expected to publish in, and a poem is exactly the form where that gap does damage.
The standard is a real thing and it was built
Standard Nepali is not the natural state of the language. It is a particular register, historically centred on the Kathmandu valley and the eastern hills, that became the language of administration, textbooks, radio and print. Everything else became a bhasika — a dialect — which is a word that describes a power relationship at least as much as a linguistic one.
Nepal's census records dozens of languages, and some varieties that people casually call Nepali dialects — Doteli in the far west among them — are counted separately. Where the line falls between a dialect and a language is contested, politically loaded, and not something I am qualified to settle. What I can speak to is what it feels like from inside: the speech I grew up with is close enough to standard Nepali that people understand most of it, and different enough that using it in writing is a visible choice.
Three things a dialect can do that the standard cannot
It carries a specific place without describing one. If I write a poem in far-western speech, I have located it before the first image. Standard Nepali is placeless by design — that is what made it useful for administration — so a poem written in it has to do geographical work through detail. Dialect does it in the grammar.
It has words the standard genuinely lacks. Not exotic ones. Ordinary ones: specific terms for slopes, weather, kinship positions, the particular way something is carried. When the standard has no equivalent, the writer either uses a longer phrase and loses the compression, or keeps the word and loses some readers.
It changes what the speaker sounds like. This is the one people underestimate. Register is characterisation. A voice speaking in dialect is not a neutral voice, and in a literary culture where the standard has been the voice of the state and the school, that non-neutrality can be the point of the poem.
What it actually costs
I want to be honest about this rather than romantic, because there is a genre of writing about dialect that treats the cost as noble and skips the arithmetic.
Readers. Fewer people can read it comfortably. Some will not try. That is not a failure of the audience; it is a fact about a smaller speech community.
Critical attention. Reviewing a poem in a dialect you do not speak is difficult, so critics mostly do not. Dialect work tends to be praised warmly and analysed rarely, which sounds like a compliment and is actually a way of being set aside.
Search. A practical modern cost. Standard Nepali is what people type. Dialect spellings are unstandardised — the same word may be written three ways — so the writing is harder to find even by people who would want it. That is a real disadvantage and I have written about the general version of it in Nepali-language search.
The condescension risk. Dialect in Nepali writing has a history of being used for comic characters and rustic servants — the villager whose speech marks him as simple while the educated character speaks standard. Any serious use of dialect is working against that inheritance, and a reader may bring it to your poem whether you invited it or not.
The poem, and what the dialect was doing in it
काँ बगी कर्णाली ? is addressed to the Karnali, and the question in the title — where did the Karnali flow — is a question you ask about something that has already gone.
The reason it had to be in that speech is simple enough: the river is there. Addressing the Karnali in the standard would have introduced a distance between the speaker and the thing spoken to, and the poem is about not being able to hold on to something. Adding a second layer of distance in the grammar would have made it a poem about a river rather than a poem spoken to one.
There is also a plainer reason. That is the speech the question sounds right in. When I tried the same lines in standard Nepali they were correct and inert, like a translation of themselves — which, as I have written about translation into English, is what happens when the words survive and the sound does not.
The choice is not all-or-nothing
Most writers I know who use dialect do not write entire books in it. They use it where the poem's centre needs it and write the rest in the standard, which is closer to how people actually speak — moving between registers depending on who they are talking to.
That mixed approach gets criticised from both directions. Purists call it dilution; readers who want standard Nepali find it inconsistent. I think both are wrong, for the same reason: a poem is not a language policy. It is allowed to use whatever gets the thing said, and a speaker who shifts register mid-poem is doing something true about how people talk.
What I would say to someone deciding
Do not write in dialect to be authentic. Authenticity is not a literary quality and readers can smell it being performed.
Write in it when the poem will not work otherwise — when the standard version comes out correct and dead, when the word you need does not exist elsewhere, or when the voice has to be a particular person from a particular place rather than a general speaker.
And then accept the trade honestly: fewer readers, less criticism, harder to find. In exchange, one poem that is not a translation of itself. On some poems that is a bad deal. On the Karnali one it was not.
More are under Nepali poems, mostly in the standard, and the poets whose language decisions I read against are under Nepali poets — Devkota in particular, who made a related choice about metre rather than dialect and made it for the same kind of reason.