Keyword research is still how you learn what people want and roughly how many of them want it. That has not changed. What changed is the writing. Older SEO rewarded putting the exact phrase in the title, the first paragraph and a few headings. AI retrieval works on meaning, so a page that thoroughly covers heat pump short cycling will surface for dozens of phrasings nobody ever typed literally into a keyword tool.
So write for topic coverage. One page per real service and location, using the terms customers actually use, including brand names, part names, symptoms, slang, and the wrong words people reach for when they do not know the right one. Then add the practical questions: cost ranges, how long it takes, whether insurance covers it, what happens first. Those variations are what match the long questions people now type into an assistant.
Two things to avoid. Do not abandon basic on page practice, because titles and headings still tell both Google and the models what a page is for, and clarity there is cheap. And do not pad pages with synonyms to look thorough, because length without substance reads as filler to a model the same way it reads as filler to a customer. Specific and complete beats long.
Send the site over. We will tell you what is broken, what you could rank for, and whether AI search knows you exist. No charge and no pitch deck.
