Transparency is Not Trust
When does experimentation affect trust? And who decides?
The most interesting thing about AI experimentation is how indefensible it is. I’ve been very clear about testing AI to figure out how it works or doesn’t. My first podcast about AI was published in January 2021, 5 years ago. I’ve followed my curiosity but I haven’t really used it much to actually build until March of this year when I realized I could use Claude to build a new website.
I’ve been testing out writing long form with it. I’ve customized Claude skills to eliminate language choices, structures and patterns I don’t love about AI and trained it on how I write. I often start with a podcast first and then repurpose the transcript. It’s my language, my ideas but Claude formatting them. I edit intensely. Often I never publish what I’ve created at all because it’s just not good enough. I never use AI to write comments or short posts anywhere. Mostly I use it as a thinking partner to identify blind spots or to suggest resources I might not have considered during the research phase of my work. I’ve been transparent online that I’m exploring AI to test capabilities so nothing is hidden. The funny part about this story is the title of the piece in question. Transparency is not trust.
This week, I did publish this piece on LinkedIn and Substack. Don’t look for it. I deleted it. I spent about 5 hours working on the piece, two versions customized for each platform. I even tagged a creator I felt had inspired some of my analysis when I got significant compliments from two different marketing experts who pride themselves on effective communication, one of whom has a background as a psychologist. The next day, that creator wrote a whole piece about her refusal to restack a post and whether or not she had an obligation to tell the person who wrote the post that the voice was “too AI”. Ironically, she offered to generously share her process for using AI well, a process that almost mirrors my own process to a T. She did not tag me, she did not send it to me or contact me in any way but based on the timing of the post and her interaction with my writing, I suspect she’s talking about me. She wrote that she admired the ideas but didn’t want to endorse the writing because she suspected AI, even though she herself uses AI.
I was instantly angry but I know enough to reflect on what’s really triggering the response. It felt patronizing. It felt judgmental. It felt hypocritical. Use AI but only in the way I use it or I’ll tune out because it’s not worth my time. It felt as if she was claiming to be the only person with the discernment to assess what makes writing worthy. Once I let myself feel the emotions, what I was left with is how indefensible it is to use AI to write, even if you have a detailed process to avoid generating AI slop. Slop is in the eye of the beholder. As I said, two people I really respect loved the writing. One of them actually told me that he hoped he could write like me when he grows up. The other said she devoured the writing and shared it with her audience, something I rarely see her do.
Despite this, I can’t defend the writing if someone says it sounds like AI. It was used as a contributor. How much editing do you have to do before you eliminate the digital footprint entirely? It’s not defined. There is no guideline. The only way to defend it is to avoid AI use entirely. I’m typing this word by word. Do you believe me? You can’t prove it and neither can I really. There’s no reliable test. You have my word. It’s really not enough anymore. Or maybe it’s not about being enough. Maybe it depends on the context and your understanding of my intent and capability. I’d say the people who complimented me trust me and the context. The person who wrote the unaddressed feedback does not.