Open ChatGPT today and look at the empty box. It says "Ask anything."
Open Google's AI Mode. "Ask anything." Open Perplexity. "Ask anything."
A year ago the industry called that box a prompt, and taught people to write prompts. The word is quietly disappearing from the places ordinary people actually see. It is worth understanding why, because it changes what your business needs to be found for.
The word changed, but only on the surface
The interesting part is what sits underneath. In ChatGPT's own page code, the input box is still named prompt-textarea. In Claude, the box reads "How can I help you today?" while the accessibility label behind it still says "Write your prompt to Claude."
So "prompt" has not been deleted. It has been demoted. The engines still use it internally, and they still use it with developers. They have stopped using it with you.
You can see the split clearly in the documentation. On OpenAI's developer pages, "prompt" appears roughly forty times more often than "ask" or "question." On Microsoft's consumer Copilot page, "ask" and "question" outnumber "prompt" by about ten to one. In the App Store descriptions for ChatGPT, Copilot, Perplexity and Claude, the word "prompt" appears zero times.
The rule that has emerged is simple. A prompt is an artifact: something you save, share, teach, or count. A question is what a person does. The engines decided their users are doing the second thing.
People agree, and the gap is widening
This is not just branding. It shows up in how people search for help.
Over the past year in the United States, searches for "ask ChatGPT" have been rising, while searches for "ChatGPT prompts" have been falling. The two lines are moving apart, not converging. The same pattern holds for "ask AI" against "AI prompts."
People are not learning to write prompts. They are learning that they do not have to.
Why this matters for your business
If customers are asking rather than prompting, the things you need to be found for are questions, phrased the way a person would say them out loud.
That is a real difference from search keywords. Someone typing into Google types "plumber denver." Someone asking ChatGPT asks "who is a reliable plumber near me who can come out this week and does not overcharge for emergency calls?"
The second one carries constraints, context, and intent that the first one leaves out. It is longer. It is specific. And it means the answer that comes back is filtering on things a keyword never captured: whether you are described as reliable, whether you are known for fair pricing, whether anything about your availability is visible at all.
So the practical question for a business is not "what keywords do I rank for." It is "what would a customer actually say, and does the answer include me?" Those are the questions worth tracking, and they are the questions we track.
The second shift: AI is not replacing search, it is inside it
There is a related change that is easy to miss.
Google's AI Mode looks like a separate product. It is not. Visit it and the address bar resolves to google.com/search, and the page title is still "Google Search." Google is treating AI as a mode of Search rather than a replacement for it, and its own guidance for site owners says so directly: asked whether SEO still matters for generative AI search, Google's answer is a plain yes.
Google has also been unusually blunt about the acronyms that grew up around this. Its optimization guide notes that terms like Answer Engine Optimization and Generative Engine Optimization are common online, and warns that many of the "hacks" attached to them are not effective and are not supported by how Search actually works.
The useful reading is not that the acronyms are forbidden. It is that there is no trick here. The work is the same work: be genuinely present, clearly described, and cited by sources the engines trust. What changed is not the method. What changed is where the answer gets delivered, and whether anyone can see you when it does.
What to take from this
Three things.
Track questions, not keywords. The unit of measurement that matters is the sentence a customer would actually say.
Do not chase the vocabulary. GEO, AEO, AIO, LLM SEO: the industry has produced at least six competing acronyms and no consensus, and the largest AI search provider is openly skeptical of all of them. You do not need to pick a side in an argument between marketers.
Watch the engines, not just your score. These systems change underneath you. A model swap or a change in how an engine cites sources can move your numbers without you having done anything at all. Knowing which happened is the difference between a real problem and a false alarm.
Sources: live product interfaces for ChatGPT, Google AI Mode, Perplexity and Claude, checked July 2026; OpenAI, Anthropic and Microsoft documentation; Google Trends (US, trailing twelve months); Google's guide to optimizing for generative AI features on Google Search.
