Keyword Intent Classifier — Info / Comm / Trans / Nav, Batch

Intent decides what kind of page can rank

Intent is the search engine's inference of what the searcher actually wants, and it maps directly onto page type:

• Informational — the searcher wants to understand. A guide, explainer or tutorial can rank. A product page cannot, however well optimised.
• Commercial investigation — comparing before buying. Comparisons, reviews and roundups rank; both a pure explainer and a pure product page struggle.
• Transactional — ready to act. A product or signup page ranks; a blog post about the topic will not.
• Navigational — looking for a specific site or brand. Rarely worth targeting unless the brand is yours.

Why optimising the wrong format cannot work

Search engines infer intent largely from what searchers click, so the winning result type is set by behaviour rather than by your preference. You cannot rank a product page for an informational query by improving the product page — the format is wrong, and optimisation does not fix a format mismatch. That is the argument for classifying intent before planning content rather than after writing it.

What this tool actually checks

Word patterns, in a fixed order of precedence, and the order matters more than people expect. A transactional signal is checked first (buy, order, book, reserve, hire, sign up, subscribe, free trial, coupon, discount, cheap). Then a commercial one (best, top, review, vs, alternative, compare). Then a navigational one (login, sign in, account, dashboard, and a few platform names). Informational is last, and its question words only count at the start of the phrase — apart from "how to", which counts anywhere.

Two consequences worth knowing. First, precedence means a keyword matching two patterns takes the earlier one: best cheap coffee maker comes back Transactional, because cheap is tested before best. Second, a phrase where the question word is buried — coffee maker what is the best kind — misses the informational rule entirely.

Verify with real results, because that is the actual answer

Search the keyword and look at what ranks. If the top ten are all guides, the intent is informational whatever any classifier says. Real results beat every heuristic here, because they reflect behaviour the engine has already measured across far more searchers than a word list can represent. The tool exists to sort a few hundred keywords quickly enough to be worth doing, not to settle the question.

Anything the patterns cannot place is returned as Informational and flagged low confidence. Treat that flag as an instruction to check manually rather than as a weak answer.

Mixed intent, where the engine is uncertain too

Many keywords serve more than one intent, and the results page shows a blend — a couple of guides, a couple of product pages, a comparison. That mixture is itself information: it tells you which format still has room. A page can serve two intents; trying to serve all four serves none.

Once the intents are assigned, the keyword clusterer groups what is left into the pages that should actually be built — since keywords sharing an intent and a topic usually belong on one page rather than several.