Even the Eliza chatbot from 1964 helped some people therapeutically. The "personhood" of the AI doesn't matter in this case as long as it's more beneficial than harmful. I think of it as 'interactive journaling', in OPs case.