Keyword lists were built around how people type into a search box: short, clipped, stripped of context. Best accountant Perth. Commercial cleaning Sydney cost. Two to four words, no sentence.
Nobody talks to an assistant that way. They write sentences, they include their situation, and they ask for judgement. Optimising for the old phrasing means optimising for a question that is being asked less every month.
What a real prompt looks like
A search query might be "conveyancer Perth". The equivalent prompt is closer to: "We're buying our first place in the northern suburbs and settling in about six weeks. What should we look for in a conveyancer and roughly what should it cost?"
Notice what is in there that the keyword lacks. A situation, a constraint, a timeframe, two distinct questions and an implicit request for advice rather than a list. Content written for the keyword answers none of it. Content written for the prompt answers all of it and gets quoted.
Where to find the real ones
You cannot get these from a keyword tool, because the tools report search volume for search phrasing. You get them from the places people actually explain their situations.
- Your sales calls. The first three minutes of a discovery call is a prompt, spoken aloud. Start writing them down verbatim.
- Your inbox. First contact emails are close to perfect prompts, complete with context and constraints.
- Your support tickets. Post purchase questions reveal what buyers did not understand beforehand, which is exactly what the next buyer will ask an assistant.
- Community threads. Reddit, industry forums and local groups are full of people describing their situation and asking for recommendations. This is the closest public analogue to prompt behaviour, and engines read these threads too.
- The assistants themselves. Ask one what people typically want to know before buying in your category. The answer is a serviceable starting list.
Turning prompts into content
Once you have thirty real prompts, group them by underlying need rather than by wording. Several differently phrased prompts usually share one answer, and one substantial page answering that need properly will match all of them. This is the opposite of the old approach of one page per keyword variant.
Then write the answer the way you would say it: lead with the direct response, follow with the reasoning, include the numbers, name the trade offs honestly.
Your customers stopped typing fragments. Content written for fragments has stopped matching them.
The awkward questions matter most
The prompts businesses least want to answer publicly, about price, about who they are not right for, about how they compare to a cheaper option, are the ones assistants most need an answer to. Whoever answers them becomes the source. Usually nobody does, which makes it the cheapest position on the board.