Trendsloppy AI

I’d rather be lied to by AI than by a human. Perhaps that’s just me, but the nice thing about AI being untruthful is that you can immediately call it out, and it can’t walk away. When you call out a human for being dishonest, you have to deal with their wounded ego, they might get aggressive, and try to get retribution. All of this moves you further from the truth, and is a distraction from the problem you were addressing in the first place.

It’s no surprise that AI can be misleading. According to a recent Harvard Business Review study, AI generally regurgitates the same trendy business advice, even when fed materially different sets of circumstances. AI’s buzzword-heavy corporate advice is referred to in this article as trendslop, which is a fantastic term that immediately defines itself without needing explanation.

Since I write about real estate, this immediately made me think of complex business decisions like the negotiation of a real estate transaction. Increasingly, clients reference decisions ChatGPT made for them regarding the asking price for their house, whether a house is a good investment, etc. Sometimes Chat can give good answers, but the more complex and personalized the queries get, the more trendsloppy the answers become. It’s nice to have the Harvard Business Review confirm that the goal of AI in complex business situations is not to deeply analyze the personalized data it’s fed, but to say something that sounds superficially plausible.

Most real estate agents seem to have more work because of AI, not less. An article in The Real Deal describes agents who have to respond to lengthy AI emails from clients that cite non-existent laws and give highly misleading advice. One agent said she started receiving long, detailed emails from a client who says they hate reading, emails that referenced laws from other states that weren’t actually applicable to the transaction. Another agent says of client interactions, “[Now] I can mentally prepare myself for the 15 questions you’re going to bring up on ChatGPT on the phone, and then you’re still going to send me a ChatGPT email, and then I gotta respond to the ChatGPT, which I’m probably just putting into ChatGPT because I don’t want to have to write the email.”

Like most tools, AI is limited by how the user interacts with it. If you don’t understand what it is, and what it isn’t, your AI interactions are going to be misleading. The problem I have with AI is not that’s it’s trendsloppy, it’s how AI is perceived to have superhuman intelligence, when it’s really a trend aggregator. If you view AI as a super interactive encyclopedia, it’s extremely helpful. However, asking an interactive encyclopedia for complex, personalized business advice is kind of insane. Asking an interactive Wikipedia to summarize the general thinking on a popular topic makes sense, but it would make less sense to ask Wikipedia “how much should we increase production in Q2 while taking into account the debt service on our new factory, as well as the launch of our IPO in Q4?”

I think we often subconsciously assume that AI is a quantum supercomputer, performing magical calculations that account for every circumstance and running mockups of every possible scenario before spitting out an answer. However, AI is not a quantum computer; it does not run through every possible circumstance while keeping your personal goals in mind. Quantum computers barely exist at all, and are still much more theoretical than applicable. What AI does is quickly dive into the most widely discussed aspects of a topic and create an interactive executive summary. This is really, really helpful, and I use it all the time.

It’s not so much AI itself that frustrates me, but all of the unreasonable expectations we place on it. This is partly because of how it’s been marketed to sell it to the public. If AI were referred to as an Interactive Trend Aggregator, the general perception of its usefulness might be more reasonable. However, “Interactive Trend Aggregator” makes me sleepy compared to “Artificial Intelligence,” which makes me want to throw billions at it immediately.

That said, an encyclopedia you can argue with is a game-changer. One basic thing that sets humans apart from other animals is the degree to which we share knowledge with one another through language, creating an ever-expanding human encyclopedia. AI is a way to directly debate with this encyclopedia; it’s a rapid-fire personal researcher that queries the most-referenced sources on any topic. I can skip over the laborious initial research and zoom in to relevant sources after my AI assistant has helped me narrow my focus. However, AI is only a tool, it is not objective reality.

The marketing of AI almost as a quantum supercomputer makes it appear as though it can see into parallel universes, or at least model them. If you haven’t bought into this marketing, it shouldn’t be a surprise that AI is trendsloppy rather than quantum.

https://fortune.com/2026/04/10/what-is-trendslop-llm-hidden-bias-bad-advice-workplace-consultants/

https://lyletagawa.com/posts/llm-context-biases/

https://medium.com/@jerrygrzegorzek/trendslop-03a5928bf506