AI & SEO automation
Search intent classification in SEO
I start with a kick-off: I establish the implementation goals, the sections of the site to be analysed, the markets and languages, and the SERP result types that matter from a business perspective. I then collect data on queries and landing pages, organise the keywords, cluster them and map them to existing URLs, pointing out gaps and keyword cannibalisation.
About the service
Intent classification and query-to-URL mapping based on data and the SERP
About the intent classification service
A few service details- Kick-off and scope of the analysis
- Defining markets and languages
- Collecting query data
- Clustering and intent signals
- Mapping clusters to URLs
- Recommendations: structure and content
Work process
How we work together on intent classification and URL mapping
I start with a kick-off, in which we specify the scope of the intent analysis, the markets and languages, and how changes will be approved. I then collect data on queries and pages in order to organise the keywords and move on to clustering and mapping to URLs.
- 01/ 03
Kick-off and scope
I establish the implementation goals, the list of site sections to be analysed, the languages and markets, and the relevant SERP result types.
- 02/ 03
Access and data
I collect queries, clicks, impressions, landing pages, the site architecture and product priorities as the basis for further analysis.
- 03/ 03
Clusters and matching
I cluster the queries, analyse the intent signals and match the clusters to existing URLs, detecting keyword cannibalisation and missing pages and agreeing changes to structure and content.
Kick-off and defining the scope of work
The kick-off and defining the scope of work consist of specifying exactly what is to be optimised as part of the intent classification. At this stage I establish the implementation goals and the list of site sections to be included in the analysis. I also clarify which languages and markets we cover, so that the later collection and organisation of queries corresponds to the real scope of the site. In parallel, we establish which SERP result types matter from a business perspective, because this affects how intent is interpreted.
At the kick-off we also agree how changes will be approved, which brings order to later decisions about matching pages to intent. These arrangements are crucial, because the scope determines how broad the set of queries will be and how many clusters need to be worked through. The scope may grow when the site has many sections and markets, poorly organised existing URLs, heavy keyword cannibalisation, or when new templates and content formats are needed. In practice, the number of stakeholders approving the changes matters too, because it affects how the cooperation runs.
Access and input data needed for the analysis
The access and input data needed for the analysis include information about queries and pages, the current site architecture and a list of priority products or services. I need data such as queries, clicks, impressions and landing pages to be able to collect and organise the keywords and move on to query clustering. On that basis it is then possible to analyse the search results for the clusters and extract intent signals, including query modifiers and the dominant content types in the SERP. The input data also makes it possible to assess the expected level of detail and the stage of the user’s decision journey.
Mapping intent to existing URLs requires matching the clusters to the current pages and identifying situations in which several URLs answer the same intent, or in which an intent has no target page. Access to the site architecture supports decisions on whether changes are needed to page structure, content merging, redirects or modifications to headings and sections. The collected data is also the starting point for later recommendations on aligning content and page structure, as well as architecture and internal linking.
Collecting and organising queries
Collecting and organising queries consists of compiling a list of keywords from various sources and preparing it for further work on intent. At this stage I remove duplicates, so as not to multiply the same topics in the next steps. I also normalise the forms of the queries, taking into account inflected forms and common typos, in order to standardise the notation and make grouping easier. In addition, I label the queries as brand/non-brand and organise them into topical segments.
The organised set of queries is the basis for working on clusters rather than on individual keyword variants. Thanks to the normalisation and labelling, it is possible to assess more precisely which topics concern the same user need and which require a separate approach. Topical segmentation also makes it easier to map queries to site sections later and to identify areas that need clarifying. The result is a list of queries ready for clustering and search results analysis.
Query clustering and SERP analysis
Query clustering and SERP analysis consist of grouping keywords into coherent clusters and checking which result types dominate for each cluster in the search engine. In clustering, I group topically related and synonymous queries so that they can be assigned to one intent and one type of target page. I work on clusters because they reflect users’ real needs better than individual query variants. This lays the groundwork for unambiguously assigning a content format to the expectations that emerge from the SERP.
In the search results analysis, I check which page types and content formats appear most often for a given cluster, e.g. guides, categories, lists or product pages, and which elements of the results are featured. On that basis I extract intent signals from two sources: the query itself and the dominant results in the SERP. I assess, among other things, the modifiers in the query (e.g. “price”, “reviews”, “how”), the expected level of detail and the stage of the user’s decision journey. These observations are later used to assign intent consistently within the defined clusters.
Intent model and granularity
I establish the intent model and granularity by choosing how precisely user needs are to be labelled within the query clusters. The decision concerns whether we use a simple classification (informational/navigational/transactional/comparative) or a more elaborate one, e.g. with subtypes or local intents. In parallel, I define the rules for resolving disputed cases, so that the assignments are consistent across clusters. This stage follows directly from the earlier assessment of intent signals and is meant to bring order to later implementation decisions.
The adopted model governs how I assign intent to a cluster and how I interpret the expected purpose of the page (e.g. education, comparison, purchase, contact). Granularity also affects whether one page format is enough within a single topic, or whether separate destinations are needed for different sub-intents. Thanks to the resolution rules, the classification can be carried out consistently even where the queries are similar but the SERP shows different content formats. The result is a set of unambiguous rules that bring order to the subsequent mapping and optimisation recommendations.
Recommendations for intent-based optimisation
I prepare the recommendations for intent-based optimisation as a set of guidelines for the content, structure and elements of the page, so that they meet the user expectations arising from the assigned intent. In practice this covers matching the layout of sections, the scope of information, the CTAs and the format (e.g. list/guide/category) to what is consistent with the cluster’s intent. I also indicate how to adjust titles and headings and which subtopics to include, so that the page fulfils the primary goal assigned to the intent. An important element of the recommendations is avoiding mixing conflicting intents on one page, because that makes it harder to match the content unambiguously to the user’s needs.
In the recommendations I also take into account the architecture and internal linking, indicating where new pages are needed and how to lay out the paths between intents (e.g. info → comparison → transaction). I specify which contextual links support the topic clusters and help the user move on to the next stage of the decision. If conflicting assignments are identified during the work (e.g. several URLs serving the same intent), the recommendations are adjusted to the decisions on merging content, redirects or changes to the structure of headings and sections. The scope of the recommendations may grow with the number of sections/markets, the level of disorder in the URLs, the scale of keyword cannibalisation, the need for new templates or the number of stakeholders approving the changes.
Testimonials
What clients and the industry say
Feedback from clients and industry people I have worked with on SEO projects.
I had the chance to work with Kuba at Kulturalnie o SEO, an event I organise. Kuba did a great job as a speaker and received high marks from the audience. He showed professionalism and broad knowledge. In other projects at Vestigio, Kuba shows enormous commitment, a willingness to explore and implement new ideas, and excellent organisation of his work.
01 / 08
You talk to me, not a salesperson
Tell me what you want to achieve. I’ll reply personally
I work as a freelancer: the same person reads your message, prepares the quote and then runs the project. No sales team in between.