Part 2: users, queries and intents

What the document says

Part 2 changes the lens: the page exits, the user enters. The vocabulary it defines is the foundation of everything the industry calls “search intent” — worth knowing at the source:

A query is what the user types, speaks or photographs. An interpretation is what the query might mean. An intent is what the user wants to accomplish. One query has several possible interpretations; each interpretation, several intents; and one user can hold several intents at once.
Interpretations have a hierarchy: dominant (what most people mean — [apple] = the company), common (what many mean — the fruit), reasonable minor (few, but legitimate), unlikely minor and no chance — the interpretation that technically matches the words but almost nobody has in mind: nobody searching [hot dog] wants an overheated pet. A no-chance result “feels off-topic even if it matches the query words.”
Meaning shifts over time. [iphone] in 2007 meant the iPhone 1; today it means the latest model; tomorrow, another. The default assumption is that users want current information, the newest model, the most recent edition of a recurring event.
The four classic intents: Know (learn — with the subcategory Know Simple: a specific, correct, complete answer in 1–2 sentences, with consensus), Do (accomplish: calculate, watch, pay, download), Website (reach a specific site) and Visit-in-Person (find a physical place). The document says not to agonize over the Know/Do boundary — many queries are both.
Two precision examples worth the whole chapter: in [lewis county non emergency number], ignoring the word “non” ruins the result — “users can become frustrated when results ignore important words.” And for [is egg salad healthy], an egg salad recipe is not a reasonable intent — the user wants nutrition information; the recipe is the next step, not the answer.
Location matters conditionally: [verbena] in Austin might be a local restaurant; in Flint, only the plant. And an explicit location in the query ([New York hotels] searched from Dallas) outranks the user’s own location.
A rare methodological warning: when researching a query on Google, don’t rely on the top results — “a query may have other meanings not represented on Google’s search results pages,” and no page deserves a high rating just for ranking.

In practice

The interpretation hierarchy is the correct mental model for keyword research. Before contesting a query, classify yourself: do you serve the dominant interpretation, a common one, or a minor one? You can rank serving a reasonable minor interpretation — but the traffic ceiling and the possible score are proportional. And if your content matches the words while serving a “no chance” interpretation, an algorithm trained on this scale will treat it as off-topic no matter how optimized it is.

The egg salad example is a lesson in adjacency not being relevance. Content related to what the user wants is not content that answers what they asked. Half the “top of funnel” content that underperforms fails exactly here: it answers the question next door, not the question asked.

Know Simple maps what AI and featured snippets swallow — and what’s left. A two-sentence consensus answer is direct-answer territory (and now AI Overviews): fighting for that click is a lost war. The document lists what is not Know Simple: broad queries, controversial ones, those with no single answer, exploration, inspiration, and searches for the opinions and perspectives of real people. That’s the map of defensible content.

Freshness is a built-in assumption, not a bonus. If the dominant interpretation slides toward “the latest” over time, evergreen content on an evolving topic is always decaying in relevance. Real revision dates, updated examples and updated models cited aren’t cosmetic — they’re staying inside the dominant interpretation.