AI search

What Is Query Augmentation?

Query augmentation is the expansion of a user's question into rewrites, drill-downs and next steps before retrieval, and the pages that answer the expanded set are the ones that qualify for the answer. Two Google documents describe the mechanism: a lapsed patent and a pending filing.

Query augmentation, illustrated
A user query expanded by stored and synthetic queries into one set, which a content brief maps to page components

What does query augmentation add to a search query before retrieval?

Query augmentation adds synonyms, the entity type, attribute terms, and context terms to the literal query, then retrieves against the expanded set.

A query for semantic SEO audit cost expands toward audit pricing basis, audit deliverables, and audit versus technical audit, so a page that answers only the literal string covers one row of that set.

Google's documents describe two mechanisms. A lapsed patent matches the user query to stored queries that already perform well, and a pending filing has a generative model write new queries from the query and its session.

Augmentation is not a longer keyword list: a keyword list collects strings by volume, and an augmented set collects the phrasings that share one meaning and one answer. Strings with a different answer belong to a different page, and the Query Deserves a Page test makes that call.

How did Google's query augmentation patent turn titles and anchor text into queries?

Google's query augmentation patent stored page titles and widely repeated anchor text as queries a page answers, then matched user queries to them.

The patent is US 9,916,366, granted in 2018 and since expired for non-payment of fees. Its stored queries come from two sources: user queries from logs that pass performance signals such as long clicks and click-through rate, and synthetic queries generated from titles, anchor text, tagged sections, and listing fields.

A document title counts as a well-performing query even when no user typed it, because submitting the title returns that document. Anchor text that many pages use for one target, 500 pages in the patent's example, is stored as a query for that target.

Matching favors nouns. Candidate stored queries share at least two terms with the user query, allowing synonyms and stems. Verbs and adjectives drop before nouns. The patent has lapsed, so this is design logic, not a guarantee.

How does Google's stateful chat filing generate synthetic queries?

Google's stateful chat filing has a generative model write synthetic queries: rewrites, simplifications, drill-downs, supplemental queries, and next steps.

The filing is US 2024/0289407, published in August 2024 and still pending.

Session state is a set of tokens that accumulates attributes. A buyer researching agencies starts with {semantic SEO agency}, and two queries later the set reads {semantic SEO agency, audit, flat fee}. Every later rewrite carries those tokens, so retrieved documents are matched against the attributes as well as the head entity.

Documents are selected for the query and for the rewrites, so a page that answers the head query and none of its rewrites competes for one slot in a set built from several. Next-step suggestions follow the categories the results fall into, such as price range or a best-for variant.

Why do drill-down queries change after a user reads a tutorial?

Drill-down queries change because the system in the filing tracks what the user has already read and then seeks more detailed, lower-level information.

The filing describes a familiarity signal drawn from the user's profile and session. Content the user already read is omitted from the summary, or the summary goes deeper, and later synthetic queries seek lower-level detail.

The page-side response is depth tiers inside a cluster. A pillar page holds the high-level layer, and each cluster page holds one level down: a specific type, application, or comparison. No two pages in a cluster sit at the same depth on the same entity.

Each cluster page states its depth in the standfirst. For example, a page on query templates would say it assumes the reader knows what a topical map is.

How does a content brief map augmented queries to page components?

A content brief maps each augmented query to one component, the triples that component states, and the named entities it contains.

The method has five steps: list the augmented variations of the query, assign one component to each, write its triples, list the named entities it carries as text, and order the components. The four-column result is the brief.

Order follows function: identity and trust augmentations sit in macro context above the fold, attribute augmentations sit in micro context below it, and question augmentations close the page.

Take a dental clinic's page for dental implants. The augmentation dental implant cost maps to a cost component with the pricing basis as triples, and implant procedure steps map to an ordered step list. Candidacy questions map to the closing FAQ, and the clinic's dentists appear as named entities in the trust block.

Which page elements carry query augmentation phrasings?

Titles, headings, and internal link anchors carry the phrasings augmentation matches against, so each one is written as a query.

The title is a phrase a user types: entity type, entity, and one constraint. A working title returns its own page when pasted into search, and a slogan matches nothing.

Headings cover the augmented terms, not only the literal keyword, and each leads with the head noun, because verbs and adjectives drop first in matching. Each page also has one canonical anchor phrase that every internal link uses, since varying the anchor splits the phrase a repeated anchor builds.

Sources

Patents and filings describe systems Google has invented. None is a confirmed ranking factor.

Illustration for the knowledge article What is AI citation readiness?AI citation readinessAI citation readiness is the state of a page whose sentences can each be verified against one passage and matched to what other trusted sources say.Illustration for the knowledge article What is a query template?Query templatesA query template is a repeatable search pattern with a variable entity, such as how to [task] or [service] in [city].Illustration for the knowledge article What is a content brief?What is a content briefA content brief is the document that turns one row of the topical map into one page.Illustration for the knowledge article What does query deserves a page mean?Query Deserves a PageQuery Deserves a Page is the test that decides if a search query earns its own URL or becomes a heading on an existing page.Illustration for the AI Search Visibility serviceAI Search VisibilityAnswer-first passages, entity signals, mention building, and AI Overview and prompt tracking.Illustration for the Semantic SEO Strategy serviceSemantic SEO StrategySeven documents, from the topical map to the 12-month publishing roadmap.Talk to us speech bubble: message the Digital Vikingz team on WhatsAppTalk to usMessage the team on WhatsApp

Questions about query augmentation

Is query augmentation the same as query fan-out?
Close to it. Query fan-out is the name Google uses for the AI Mode version of the same idea, and the page-side work is identical under either name.
Does query augmentation mean a page targets every synonym?
No. One page answers one intent, and its synonyms share that page as headings or labeled blocks. A variant without the head term is a different entity or needs the head term added.
Where do augmented query lists come from without access to Google's logs?
Three places: Google's related searches, the titles of the top ten results, and the anchor text pointing at those results. Titles and anchors count because the lapsed patent stored both as queries.
Do misspellings need their own pages or sections?
No. The lapsed patent matched candidates through spelling correction and edit distance as well as synonyms and stems.
Are brand and place names expanded like other terms?
No. The lapsed patent does not transform location names, brand names, product names, famous people, organizations, or tourist destinations. The canonical spelling goes everywhere, and abbreviations go in one alias line.
Does the page title still matter now that the patent has lapsed?
Yes. Titles, anchor text, and structured listings reappear as inputs in later Google systems, so a title written as a query fits both the old logic and the new.
Does query augmentation affect local service pages?
Yes. The lapsed patent generated synthetic queries from listing fields: business name, address, city, state, category, and website. Local pages keep the service and the city as one contiguous phrase in the title, H1, and standfirst.
Which contextual signals does the stateful chat filing use?
The filing lists prior queries, prior results pages, documents the user opened, schedule, location, time, and activity. A page has no control over that context, so it covers the attributes a session adds most often.
Does a page ever rank for an augmented query it never mentions?
Sometimes. Neural matching, a deployed Google system, relates query concepts to page concepts, so a page can rank for phrasings it never states. Stated coverage removes the guesswork.
What is the difference between query augmentation and query templates?
Query augmentation expands one query into phrasings that share its meaning, and a query template repeats one pattern with a variable entity, such as [service] cost in [country], across many entities. Augmentation shapes one page, and templates shape a network.
Does next-step coverage mean adding more pages?
Not always. A next step with its own demand and answer earns a page. A next step that shares the page's answer earns a section, and Query Deserves a Page decides which.
Does a long click still matter for query augmentation?
Google has not said, but the lapsed patent used long clicks to qualify user queries as stored queries. The patent counts a click on another link on the landing page as a long click, so a page with a next-click path turns a satisfied visit into one.
Book a callWhatsApp