This classifier reads a list of keywords and labels each one with the intent its wording suggests: informational, commercial, transactional, navigational or local. It shows the words that triggered the label, a confidence figure and the second-closest intent, then exports the whole table as a CSV.
It is a rule-based first pass. The labels come from a fixed word list, not from Google and not from a language model, so the output is a hypothesis you can check quickly rather than a verdict.
Five labels, not four
Most intent tools use four buckets. This one adds local, because near me, opening hours and directions describe a job that a guide or a product page cannot do. The five labels map to five page formats: a guide or FAQ for informational, a comparison or review for commercial, a product or pricing page for transactional, a home or login page for navigational, and a location page with hours and a map for local.
How the score is built
Each recognised single word adds 2 points to its intent, and each recognised phrase adds 3. A domain name ending in .com, .org, .io and similar adds 4 points to navigational. The phrase how to adds 4 to informational on top of the how it already contains, which is why a how-to query tends to score highly.
The leading intent is the highest score. Ties are broken in a fixed order: transactional, commercial, navigational, local, informational. Confidence is 45 plus 7 for each point of margin over the second intent, plus 2 for each point of the winning score, capped at 95. A keyword with no recognised signal defaults to informational at 30.
That formula explains the odd-looking numbers. One clear signal usually produces 63. A strong phrase such as how to pushes the label into the 90s. A tie produces a low figure even when one label wins, because the margin is zero.
A worked example
Eight keywords produced all five intents. Three were informational (37.5%), two transactional, and one each commercial, navigational and local. One keyword was flagged as ambiguous.
| Keyword | Intent | Confidence | Signal |
|---|---|---|---|
| how to check keyword density | Informational | 95% | how, how to |
| best seo tools for small business | Commercial | 63% | best |
| buy backlinks cheap | Transactional | 81% | buy, cheap |
| ahrefs login | Navigational | 63% | login |
| seo agency near me | Local | 72% | near me |
| keyword density | Informational | 30% | no explicit signal |
| pricing | Transactional | 63% | pricing |
| seo audit checklist | Informational | 63% | checklist |
The last row is the useful caution. keyword density is a plain noun phrase, so nothing matched and the tool fell back to informational at 30%. It might be the right label, but the tool did not earn it. Treat anything at 30% as unclassified.
Confidence is not accuracy
A 95% label means the wording was explicit, not that the page will rank. The classifier cannot see the results page, and the results are what actually define intent in practice. If the top ten results for an informational-looking query are all product pages, Google has decided the query is commercial, and a guide will struggle.
The fixed lexicon also has blind spots. Brand names without a domain suffix, non-English keywords, and local queries that name a place instead of saying near me will often fall through to the 30% default. plumber Leeds is local to a reader and unclassified to this tool.
What to do with the labels
Group the list by intent and look at the shape. A list that is 80% informational is a content plan; a list that is mostly transactional is a set of product or pricing pages. A long tail of mixed intent usually means two clusters should be split before you write anything.
Then check the top queries by hand. Search each one, note the format of the first three results, and compare it with the label. Where they disagree, trust the results. The classifier is fastest at triaging hundreds of keywords; it is weakest at the handful that decide your next page.