Intent detection / query intent detection The automatic identification of what kind of result a user expects from a query: a product, a category, information, or a specific action. Definition Definition Intent detection is the process a search engine runs to figure out what kind of answer a query is asking for, before it decides which specific results to return. It's a classification step that happens first: is this person looking for a single product, a whole category, a piece of information, or trying to perform an action like tracking an order or finding a store location? This is different from search intent, which is the taxonomy itself (navigational, informational, transactional, commercial investigation). Intent detection is the mechanism that assigns a query to one of those categories in the first place, typically using natural language processing to parse the query's structure, keywords, and phrasing patterns. And it's different from user intent, which is the broader, sometimes unstated goal behind a search. Intent detection works with what's detectable from the query text itself, not the full context of why someone is searching. A query like "iphone 15 case" gets detected as product-level intent, someone wants a specific item, not information about it. A query like "how to clean a leather bag" gets detected as informational intent, likely not looking to buy the bag but to find care instructions, even on a store's own search bar. Getting this classification right upstream determines whether the engine tries to match a specific product, surface a blog article, or does both. What it's used for What it's used for Without intent detection, a search engine treats every query the same way: try to find a product that matches the keywords. That works fine for a query like "leather wallet men," but it breaks down for a query like "how to clean a leather wallet," where forcing a product match either returns irrelevant items or a page with no useful results at all. Detecting intent correctly lets a store's search do more than product matching. It can route an informational query to a care guide or FAQ page instead of forcing it through product search, route a navigational query ("track my order") to an account page, and reserve full product-matching logic for queries that are actually shopping for something. That reduces zero-result pages and the frustration of getting product listings for a question that had nothing to do with buying. How Doofinder applies it Doofinder's natural language processing layer analyzes incoming queries to detect their likely intent before running the main search logic. A query phrased as a question, or containing words like "how," "what," or "best," gets flagged differently than a query that's clearly a product name or category term, and the engine can route or weight results accordingly. This detection also feeds into how the AI Shopping Assistant handles a query: a clearly informational or advisory question triggers a conversational response rather than a plain list of product cards, while a query with clear product intent goes straight to filtered results. Example Case study Example An outdoor gear store's search logs show a recurring pattern: shoppers typing "best hiking boots for wide feet" were getting a generic product grid for "hiking boots," with no way to know the results accounted for width at all. The query's real intent was closer to a recommendation request than a straightforward product search. After intent detection is tuned to recognize comparative and advisory phrasing ("best," "for," specific need statements), that same query now triggers a filtered, ranked set of boots with wide-fit options prioritized, plus a short note explaining the filter applied. Search sessions starting with a "best ___ for ___" pattern show a noticeably lower bounce rate after the change. References Sources https://doofinder.com/en/blog/natural-language-search