Search intent is the purpose a person is likely trying to fulfill with a query. The model separates learning, destination-finding, evaluation and action-oriented purposes rather than assuming that every query has only one use.

Informational queries seek explanations; navigational queries seek a known destination; commercial queries compare alternatives; transactional queries signal an intention to complete an action such as purchasing, registering or booking.

Yes. A query can support multiple plausible user goals. The secondary intent distribution helps reveal whether different page types may be relevant to the same search.

Dominant intent is the highest-probability category returned by the classifier. It is a model result, not proof that every searcher has that goal.

It expresses the classifier’s relative assessment for an intent category. It is not a measured percentage of users and should not be interpreted as a Google ranking signal.

Not necessarily. The backend may provide only the primary and selected secondary scores. Missing category scores must not be treated as measured zero probabilities.

The gap is the difference between the two strongest returned intent probabilities. A narrow gap suggests that the classifier does not strongly separate the leading purposes.

A mixed-intent query has a meaningful secondary purpose alongside its primary purpose. Such queries may warrant multiple content formats or carefully structured sections.

Collision is a TAO interpretation used when the leading intent scores are close. It describes classification ambiguity, not a penalty or a metric published by Google.

The page derives a diagnostic from the returned probabilities, including the leading score and its gap to the runner-up. This interpretation remains separate from the underlying classifier output.

No. Intent alignment may inform content decisions, but the classification displayed here is not a direct ranking factor or a prediction of where a page will rank.

Observed result types provide an independent view of the formats surfaced for a query. They can support or complicate the classifier’s interpretation.

People Also Ask can reveal question-oriented subtopics visible in a captured search environment. Its presence alone does not prove that the entire query is informational.

Shopping and product-oriented features are useful commerce signals, but they do not exclude research and comparison behavior. The overall SERP composition matters.

A local pack indicates a location-oriented result surface in the captured context. Its relevance depends on the query, geography, device and other search conditions.

AI-generated answer surfaces change the composition of some search results. Their presence should be recorded as observed evidence rather than converted automatically into an intent score.

It is the set of adjacent queries returned for a seed term and their intent classifications. The distribution can expose nearby informational, comparison and conversion demand.

The seed is a single wording. Related terms reveal neighboring needs and help determine whether an editorial cluster should contain several page types.

No. Volume estimates demand, not purpose. A high-volume query may still be informational, navigational, commercial, transactional or mixed.

No. Keyword difficulty is a separate competitive indicator. It does not directly describe the probability distribution across search intents.

The same phrase can refer to different entities, products or tasks across markets and languages. Intent interpretation should remain attached to its search context.

Yes. Events, product launches, seasonality and changing result formats can alter the observed query environment. Repeated captures should be compared with timestamps.

A missing feature or incomplete capture should not be mistaken for evidence that the feature never appears. Inspect the capture coverage and backend response before drawing conclusions.

Use the primary and secondary signals to choose suitable page purposes, such as guides, comparisons, category pages or conversion pages. Validate the choice against actual result formats.

Outputs depend on the connected classification and search-data services, their coverage and the specified context. The lab reports modeled probabilities and captured surfaces, not direct access to Google’s internal ranking systems.