AIO Copilot

Search Intent Classifier

Paste keywords and get each one labelled informational, commercial, navigational or transactional, with the words that decided it. The labels come from the words alone, so check the live results before you build a page.

How the search intent classifier works

It reads the words in each keyword. The results page has the final say.

  1. Paste your keywords

    One keyword per line. There is no cap on the list, and nothing you paste leaves your browser.

  2. Classify

    Each keyword is checked against 29 word patterns. Every pattern it matches adds a point to one of the four intents.

  3. Read the label and signals

    The intent with the most points wins, shown with the words that matched. Confidence is high when the signals agree and low when they conflict or none are found.

  4. Check the results page

    For keywords you plan to build a page around, search them and look at what ranks. Where the results disagree with the label, the results win.

What is search intent?

Search intent is the goal behind a query: to learn something, compare options, reach a particular site, or get something done, such as buying or booking. Google says in its explanation of how it ranks results that to return relevant results it first has to establish "the intent behind your query." A page can contain every word of a search and still lose to pages that answer the need behind it, which is why intent is the first thing to settle about a keyword. Our glossary entry on search intent has the full definition; this page covers sorting a keyword list by intent, and where sorting by words goes wrong.

What are the four types of search intent?

The scheme most tools use started with three types. In A taxonomy of web search (2002), Andrei Broder, writing up work done at the search engine AltaVista, split queries into navigational ("to reach a particular site"), informational ("to acquire some information assumed to be present on one or more web pages") and transactional ("to perform some web-mediated activity"). He found that informational queries made up less than half of web searches. Marketers later carved commercial investigation out of informational, for research before a purchase, such as "best crm for agencies" or "ahrefs vs semrush". This tool uses that four-way split.

Google's own raters use different labels. The September 2025 edition of the Search Quality Rater Guidelines describes Know queries (with a narrower Know Simple group for questions that have a short answer most people would agree on), Do queries for accomplishing a goal, Website queries for reaching a specific site or page, and Visit-in-person queries for going to a business. The guidelines add that many queries carry more than one intent. Their example, [walmart], can mean the nearest store, the shopping site or news about the company.

How does this classifier decide?

It counts signal words. The tool holds 29 word patterns: eight for informational (how, what, guide, tutorial, statistics), seven for commercial (best, review, vs, alternatives, software, pricing), seven for navigational (login, official, dashboard, docs and six large brand names) and seven for transactional (buy, price, coupon, book, free trial, near me). Each pattern a keyword matches adds one point to that intent, and the intent with the most points wins. A tie goes to whichever comes first in the order informational, commercial, navigational, transactional. A keyword that matches nothing is labelled informational at 40% confidence.

Confidence measures agreement. One matching pattern with nothing pulling the other way scores 80%, two score 90% and three or more score 95%. Conflicting signals drag it down. From the sample list, "moz pro pricing" scores one commercial point and one transactional point for the same word, ties, goes to commercial and lands at 45%. Anything under 50% shows a red bar: check those keywords by hand. The content type column is a rough suggestion from a fixed list for each intent, and it changes with the number of signals rather than their meaning, so trust the intent column over it.

Why check the results page as well?

Because words are weak evidence, as the taxonomy's author said himself. Broder wrote that there is "no assumption here that this intent can be inferred with any certitude from the query" and called inferring intent from a query "at best an inexact science, but usually a wild guess." The rater guidelines show why with examples. Know Simple queries frequently have no question words, so [barack obama height] asks the same thing as [how tall is barack obama]. Meanings also drift: [iphone] meant the first model in 2007 and means the newest one now.

This tool misses in the same way. "dentist toronto" contains no signal word, so it comes out informational at 40%, though most people typing it want a dentist to visit, the Visit-in-person intent in Google's guidelines. The results page is the better evidence, because it shows what Google has decided searchers want. Seer Interactive, for one, classifies intent from results-page features: a shopping box or map results mark a query transactional, a popular products block marks it commercial, sitelinks mark it navigational, and a featured snippet or Knowledge Graph panel marks it informational. Search any keyword you plan to build a page around and read what ranks. Mixed results mean Google sees several intents, and pages of different types can each rank.

Do AI Overviews appear on commercial searches?

Often, and how often depends on how the queries are sorted. Seer Interactive's April 2026 study of 53 accounts, covering January 2025 to February 2026, sorted about 49,000 tracked queries two ways. Sorted by wording, queries containing price, cost or buy words showed an AI Overview 83.4% of the time, "best of" queries 81.3% and comparisons 95.4%. Sorted by the features on their results pages, commercial queries showed one 8% of the time and transactional queries 5%, against 36% for informational ones. Seer reports the wording-based rates within the queries its results-page method classed as informational.

Read together, a query with buying words in it is often treated as a research question and answered with an AI Overview, while a query whose results page is built around shopping listings or a map rarely gets one. Google says in its guide to AI features that AI Overviews are shown only when its systems judge them "additive to classic Search", and that there are "no additional requirements" to appear in them. Plan for AI Overviews on research-style commercial queries as well as on how-to questions, and treat a commercial or transactional label from this tool as a reason to go and look.

How do you turn intent labels into pages?

Match the page to the intent, then group keywords that share both intent and results. Informational keywords become guides and explainers. Commercial ones become comparisons and roundups with prices and limits stated. Transactional ones become product, pricing or booking pages, and navigational searches for your own brand need your own pages to be the obvious result. When several keywords return the same pages, one page should cover all of them, which is the method in our guide to topic clusters.

Our guide to keyword research with AI shows where an intent pass fits between pulling queries and choosing pages. To start from demand you already have, the Search Console regex guide shows how to pull the question-style queries your site already gets impressions for, ready to paste in above.

Search intent questions.

What the classifier can and cannot tell you.

How accurate is this search intent classifier?

It is as accurate as the words in the keyword. Keywords containing "how to", "best" or "buy" get the label those words suggest, but a keyword with no signal word is labelled informational at 40% confidence whatever it means. "dentist toronto" gets that label, though most people searching it want a dentist to visit. Search the keywords you plan to build pages for and let the results page overrule the label.

What does the confidence score mean?

It measures how far the signals agree. One matching pattern with no conflict scores 80%, two score 90% and three or more score 95%. Conflicting signals lower it: one commercial and one transactional signal scores 45%, and one signal for each of the four intents scores 28%. A keyword with no signals at all gets 40%.

Can a keyword have more than one search intent?

Yes. Google's Search Quality Rater Guidelines say many queries have more than one likely intent, and their example, "walmart", can mean the nearest store, the shopping site or news about the company. This tool gives one label per keyword, so a low confidence score is the hint that a keyword is mixed. The results page shows how Google splits it.

What is the difference between commercial and transactional intent?

Commercial searchers are still comparing ("best crm for agencies", "ahrefs vs semrush"); transactional searchers are ready to act ("buy", "book", "near me", "free trial"). The line is blurry, and "pricing" scores a point for both in this tool. Check the results page: comparison articles point to commercial intent, product and pricing pages to transactional.

Is search intent the same as keyword difficulty?

No. Intent says what kind of page searchers want; keyword difficulty is a score each SEO tool calculates its own way to estimate how hard a keyword is to rank for. Google says third-party tools "don't have access to our internal ranking data", so treat difficulty as the vendor's estimate. Intent you can check yourself, for free, by searching the keyword.

Is my keyword list sent anywhere?

No. The classifier runs as JavaScript in your browser, so the keywords you paste are not uploaded or stored. There is no signup and no cap on how many keywords you classify at once.

Classified your keywords? Group them into pages next.

Intent tells you the format of a page, not which keywords belong on the same one. Our topic clusters guide shows how to group keywords by SERP overlap, with an example and a template.

Read the guide

What to do next.