Why we stopped calling it "AI-powered," and what we call it now.
I have said, in enough client meetings now that it has become a small tic, that we have stopped using the phrase "AI-powered" in our work. This is true. It is also the kind of observation that gets re-sold in a dozen newsletters as if it were an insight, and by the time you are reading this version of it, it is close to stale. So let me try to say what is actually underneath it, because the replacement advice that tends to get paired with this ("just describe the job") is itself already on the way to being a trope.
What the data says, when we can get clean data
In the categories where we have clean A/B results on landing and store copy, "AI-powered" in the first-frame promise has moved from slightly positive (2024) to neutral-or-slightly-negative in consumer apps, slightly positive in developer tools and automation, and approximately meaningless in enterprise. It is not one trend, it is a fragmentation by context.
The more interesting shift is underneath the label itself. Engagement time on any paragraph containing the word "AI" is down year over year in our sample, across categories. Users are training themselves to scan past it. It reads as a skip-cue. Whether or not the label still tests well, it is becoming a less valuable piece of real estate.
What replacing it with "describe the job" actually looks like
The easy version of the advice is: drop the label, describe what the thing does. "Write emails in your voice." "Summarize the meeting you missed." This is decent advice and it is also, at this point, the default. Everyone is doing it. The baseline has shifted, and "describe the job" copy is competing against a lot of other "describe the job" copy, much of which is written by the same small set of LLMs and starting to sound the same.
The harder version, which is where the interesting work is now, is to write copy that assumes the reader is already past the hype cycle. Your reader in 2026 has tried four or five AI products this quarter. They know what the category can do. They are not evaluating "can this tool do X." They are evaluating whether your specific product does X well enough to earn a permanent slot on their phone or in their workflow.
Copy that respects that stance sounds different. It is more specific. It compares to adjacent tools by name. It admits what the product does not do. Most AI product copy right now does not sound like this. It still sounds like it is introducing the category to someone who has never seen it. That is the real "AI-powered" problem. The two words are a symptom, not the disease.
The shift underneath the marketing
Under the copy question is a real UX shift that matters more: users increasingly ask their apps questions rather than browsing them. The search box is becoming a text box. The menu is becoming a prompt. In the products we have shipped in the last eighteen months, we have watched the dominant interaction move from tap-based to text-based faster than most of the product teams expected it to.
That is the story worth building a product around in 2026. "AI-powered" was never the story. It was the banner someone hung on the front of the building while the more interesting change was happening inside.