Query Intent Classification (Complete 2026 Guide)

Query intent classifies what a searcher wants from a query string. Google segments intent into four canonical buckets: navigational, informational, commercial investigation, transactional. Per Google's ranking systems guide, the interpretation step determines which surface renders for each input (https://developers.google.com/search/docs/appearance/ranking-systems-guide).

Navigational queries carry brand or destination signals. 'Gmail login' wants the inbox, not a tutorial. Informational queries carry learning signals. 'What is structured data' wants an explanation, not a product page. The classifier routes each query to a different content type.

Commercial investigation sits between research and purchase. 'Compare crm platforms' reads as a buyer's-guide request. Transactional queries carry purchase signals. 'Buy ahrefs subscription' wants a checkout flow. The four buckets drive distinct SERP layouts across Google.

Google classifies intent using neural matching, RankBrain, and BERT-based language models. Each model scores queries against indexed entities and click patterns. The combined score routes the query to a layout template selecting ad slots, AI Overviews, and organic positions.

Per multiple 2026 intent-classification studies, BERT handles the natural-language parsing layer. BERT resolves pronoun references, prepositional attachments, and word-sense ambiguity before intent scoring runs. The classifier then maps the parsed query to the closest canonical bucket.

Neural matching handles the entity-similarity layer. RankBrain handles long-tail disambiguation. Each layer votes on the intent bucket. The highest-vote bucket wins. Queries below a confidence threshold fall back to broad informational rendering.

The classifier runs at query-time, not index-time. Each search triggers a fresh classification pass against the live model. The same query can land in different intent buckets depending on Google's rolling model updates.

AI Overviews rewrote intent signals in 2025. Informational queries now return a synthesized answer block above the ten blue links on most trigger-eligible queries. Per Ahrefs' February 2026 update, AI Overviews trigger on 13.4% of all queries and over 35% of informational queries (https://ahrefs.com/blog/ai-overviews-reduce-clicks-update/).

Informational intent now competes with the AI Overview for position zero. The traditional ten blue links lose CTR when an AIO appears above them. Commercial investigation queries trigger AIOs less often, and transactional queries rarely trigger them.

Per Search Everywhere's 2026 AI Overviews analysis, 'what is' queries trigger AIOs at 92%, 'how to' queries at 78%, and commercial comparison queries at 34% (https://www.searcheseverywhere.com/blog/google-ai-overviews-in-2026-search-data). The classifier weights informational intent higher for question-format query strings.

Intent classification now drives feature assembly. The classifier picks the rendered set: ads, AIO, image pack, video pack, knowledge panel, local pack, or organic. One query string produces six different SERP layouts depending on the inferred intent.

Map each content piece to one intent bucket. A piece covering 'compare crm platforms' targets commercial investigation. A piece covering 'crm pricing comparison' targets the same bucket at a later funnel stage. Different intents require different content shapes.

Informational intent rewards long-form explainers, structured data, and FAQ schema. Commercial investigation intent rewards comparison tables, product spec sheets, and review-style prose. Transactional intent rewards pricing pages, checkout flows, and product schema. Each format matches a different ranking template inside Google.

Per Search Engine Land's 2026 intent-to-format mapping study, commercial pages with comparison tables outperform text-only commercial pages by 34% in CTR on the same query class (https://searchengineland.com/seo-2026-higher-standards-ai-influence-web-catching-up-473540). Format alignment produces measurable ranking gains beyond raw content quality.

Audit your top 20 pages against the four intent buckets. Reclassify each page. Pages mismatched to the dominant intent bucket lose rankings within one to two core updates. Mismatched pages also fail to earn AI Overview citations for their queries.

Weekly content brief disagreement with the SEO team lands. You pull your top 50 commercial queries from Search Console. You record the intent bucket Google inferred against the bucket your brief targeted. The mismatch list is your rewrite queue.

You sample 20 queries in an incognito window. You screenshot the SERP for each. You classify each result against your intent hypothesis. Pages on the wrong bucket fail to convert on the expected funnel stage.

Per Semrush's 2026 intent-classification dataset, 41% of top-ten commercial pages target the wrong intent bucket for their target query. The misclassification costs those pages an average 22% of expected organic traffic (https://www.semrush.com/blog/intent-misclassification/).

You document the mismatch patterns in a shared sheet. Marketing and SEO settle on a single intent taxonomy before the next brief cycle. The taxonomy pins every piece to one of the four canonical buckets at draft time.

Note the gap. This post draws on 2025-2026 data from five sources: Google's ranking systems guide, Ahrefs' AI Overviews update, Search Everywhere's 2026 analysis, Search Engine Land's 2026 SEO audit, and Semrush's intent-classification dataset. Three neural-matching scoring weights remain undisclosed. Replication required.