April 28, 2025
Attribution in the Age of AI
Thank you to everyone who replied with feedback about my AI attribution message from a couple days ago.
The core question was this:
Who should I say “wrote” an article that I generated by having an AI summarize the transcript of a talk I gave?
The source material was all mine, and I reviewed and approved the AI’s output, but it felt weird saying that I “wrote” it, especially because the result didn’t “sound” like me.
One of my more pro-AI buddies in Slack had this to say:
I don’t think you’d have this question if you hired a ghostwriter or used interns.
Fair point!
Here’s my reply:
It is hard for me to imagine letting other people literally write my stuff for me, but there are probably plenty of content marketing pipeline tasks that I would be comfortable with them doing - e.g., chopping a long video into snippets, or creating the LinkedIn carousel version of a slide deck.
That said, I’m fairly certain people like GaryVee or Hormozi have assistants (including AI) doing all sorts of content creation for them, and I have nothing against that, either as a consumer of their content, or a fellow “business thought leader-y” type. So me being uncomfortable doing the same thing seems inconsistent to me.
Maybe I’m holding myself to a double standard for historical reasons - i.e., people know I don’t have assistants, so if I started using assistants (or AI), I would feel the need to disclose that fact at least for a while until the general expectation changed to “JS doesn’t write all his own stuff.”
It’s worth noting that GaryVee, Hormozi, et al seem to use a video-first content generation workflow, which I think lends itself better to processing by assistants. My content generation workflow is written-first, which has less built-in provenance.
Interestingly, this conversation was the first time the term provenance entered my mind in the context of content creation.
Maybe in a couple of years, everyone will assume that all content was at least partially created with genAI, and bothering to give attribution about exactly which tool did what will seem silly.
But I don’t think we’re there yet.
At least I’m not.
Yours,
—J