Are you "experimenting" with AI on human subjects? I think it's fairly well established that it's not considered ethical to experiment on someone without their knowledge or consent.
Perhaps you have some disability or impairment and you're augmenting what would be your own response with the help of AI. You're probably proof-reading the response and you're happy that the response represents or closely represents your views. In that case, perhaps a disclosure is not warranted. You might instead put a note in your profile, but I would say it's not strictly needed if you're willing to accept that the comment you posted represents what you truly wanted to say.
Is your service essentially a bot? I'd say that should be disclosed.
Do you think people would respond or feel differently if they knew your response was generated by AI? eg. I might upvote a well-intentioned reply even if I don't agree with all points made, as I think the person has made a considered response to my comments. If someone simply fed my comments into an AI and posted the reply, then I don't think they're really adding value and I'd rather they not reply. If I wanted to talk to an AI, I'd skip the middle man.
Also, why not go further and include a disclosure label for human-generated content?
But further yet, why only disclose that content is AI generated, but not also disclose the manner in which the content was generated and what human-in-the-loop verification and validation took place before posting?
Unless we are talking about an AI-agent posting to social media, in which case, the disclosure could be implicit in all posts made by the AI-agent.
But to seriously answer your question, yes. It should be marked, just as "screen sequences have been sped up for presentation purposes" and "re-emactment" are clearly marked on TV.