Spoke
Semantic query understanding
Search systems model meaning rather than matching strings, which is why naturally phrased content now ranks for queries that contain none of its words — and why keyword density stopped being a lever a long time ago.
That modern retrieval models meaning rather than literal strings is well established and reflected in platform guidance. The specific mechanisms any engine uses are not published.
For about a decade, search optimization meant putting the exact query string on the page, repeatedly. It worked, which is why it became an industry, and it produced the unreadable prose everyone remembers: "If you need a personal injury lawyer Durham NC, our personal injury lawyer Durham NC team…"
Systems that model meaning made that unnecessary and then counterproductive.
Do exact-match keywords still matter?
Not as a density target. Search systems model the meaning of a query and the meaning of a page, so content that answers the question ranks whether or not it repeats the exact phrasing. What still matters is using the vocabulary your audience actually uses, because that is evidence of what the page is about and whom it is for.
The practical rule: write the phrase naturally where it belongs, once or twice, and never at the cost of a sentence reading badly.
Before and after
Query: can I still file if the accident was partly my fault
There is no sensible way to put that string on a page. A decade ago the response was to target "comparative negligence" and hope.
Today a page headed "Can I still file a claim if the accident was partly my fault?" — answering in plain language and explaining comparative negligence underneath — matches directly. The old approach optimized for the term; the new one answers the question.
This is why the AEO pillar and this spoke are the same argument. Conversational queries are the input answer engines receive, and conversationally phrased content is what matches them.
What keyword research is for now
Not for extracting strings to sprinkle. For three things:
- Discovering intent. What people actually want to know, in their words rather than yours.
- Sizing demand. Whether enough people ask to justify the page.
- Finding the gap. Questions with real volume and bad existing answers.
The output is a list of questions worth answering, not a list of phrases to hit.
The vocabulary gap
The most valuable thing this analysis surfaces is the mismatch between internal and customer vocabulary. Every organization has one, and it is invisible from inside.
The firm says "matter intake"; the client says "how do I start a case". The agency says "entity disambiguation"; the client says "why does Google think we're two companies". Writing in internal vocabulary means being unreachable by the people you wrote for — which is the same failure keyword stuffing produced, arrived at from the opposite direction.
What this does not mean
It does not mean words no longer matter. A page about estate planning will not rank for car accidents because meaning is modeled. It means you no longer have to damage a sentence to include a phrase, and that writing clearly for a specific person is now the optimization rather than an obstacle to it.
More in Search fundamentals
How search works
Three stages, three distinct failure modes, and one diagnostic habit that saves most of the time people waste guessing. Everything else on this site assumes this model.
Penalties
The tactics search engines act against have barely moved in two decades. AI didn't add a new category. It made the oldest one cheap.
How AI answers are built
It reads the question, fans it out into several searches, retrieves passages, writes an answer, and credits some of it. Knowing the order tells you which stage you failed at, which is the only way to fix anything.