I would say no.

What if you use it to give you ideas and you come up with your own greeting.

And then you generate one more, and it's the greeting you came up with.

Humans have been doing this long before AI. It's called Hallmark.

(and no, it's not genuine)

Hallmark cards tend to leave a lot of blank space to write. It’s absolutely genuine when you grab a pen and start writing. Bonus points if the Hallmark message is relevant to the person receiving it and maybe something you can riff on in the handwritten part. Funny cards are also great!

So I would say the golden rule continues to apply: the more effort is visible, the more genuine it is.

You could even say that Hallmark is giving you a prompt.

If you paid someone to do it for you, is it genuine? Ghostwriting birthday cards is…quite grim, whether that’s an AI or not.

To me, yes. Key point being, "if I review and approve of the message" - if you actually do this, then yes.

Being genuine is all about how much actual care and heart you pour into thing. That's really something only you can truly know. Using generative AI doesn't automatically make it not genuine, much like using a grammar checker or a thesaurus doesn't.

Conversely, not using AI tools doesn't suddenly make YouTube and OnlyFan creators' comments genuine. They never were. There is no care and heart in there, only salesmanship, and it's most likely outsourced to a brand management company anyway.

I recently went to an interest group meeting. It begins with a presentation, which is generally confirmed in advance by the meeting organizer based on a text abstract of the planned presentation submitted by a future presenter. This time, the presentation turned out to be staggeringly bad, mostly consisting of filler words and interjections, and presenter was struggling to express even the basic idea. I could not fathom how that would happen, considering the idea was presumably expressed in the abstract.

At some point, the organizer (to help situation) quoted from the abstract that was submitted, and it was indeed apparently decently written. After that, I could not help thinking that the likelihood the abstract was written by an LLM is very high. In that case, the presenter certainly reviewed it, but crucially did not write it—the thought process that makes the idea part of your active vocabulary, and you capable of expressing it would not have taken place.

In fact, even if no LLM was involved in this particular case (perhaps it was just a particularly violent case of stage fright, despite the event being very small and the environment very casual), it would be beyond silly to assume that this would not be happening going forward.

I used to think that it is beneficial if an LLM can help in handling certain boring signaling communication for people who are very bad at it, acting as a sort of an equalizer. A model writes stuff, you approve it, and you gain access somewhere without having, say, the written language flair of someone who went to a prestigious school, or having to spend time on something that seems unnecessary.

I am changing my mind now. Sure, the case I have described is one of the more extreme ones, but it made me think how signals are actually signals for a reason, and when these signals go away the communication field does not become equalized—instead, other signals and barriers are used: money, IRL meetings, invitation, some sort of privacy-violating invasive check of humanity, etc., or the communication that relied on some signals before would simply not happen now. I personally do not see that as a positive thing.

Only if you also tell me you did. Then I know I'm not worth your time.