People who use an eCommerce site search are often very close to buying. They are not browsing the menu. They are telling the site what they want to buy, in their own words.
The problem is that many shops treat site search as a secondary feature. A bar at the top, a few results, no real intelligence behind it. If the user misspells something, uses a synonym, searches for a brand with a variant spelling, or uses a common word instead of the technical name, the site answers with zero results.
And zero results, in practice, means: "go to a competitor".
Site search is worth more than a lot of visual work because it captures a clear intent. If someone searches for "dog food allergy", "running shoes overpronation", "dark spot face cream", "clear iphone case" or "bluetooth headphones", the site has to understand what they mean even when the query does not perfectly match a title, brand, category or product page.
Site search is not a magnifying glass
A badly built eCommerce search engine only looks for identical words. A well built one interprets intent.
That means connecting:
- product title;
- description;
- brand;
- categories;
- facets;
- internal tags;
- synonyms;
- typos;
- reviews;
- customer questions;
- availability and price;
- margin and commercial priority.
If search only looks at the product name, it misses an enormous number of intents. Users often do not know the exact name. They know the problem, the need, the symptom, the material, the format or the result they want.
Example: users search the way they speak, not the way the catalogue is written
Imagine a pet eCommerce.
In the catalogue the product is classified as:
single protein food for adult dogs
But users search for:
- dog food for allergies;
- dog food intolerances;
- hypoallergenic dog food;
- chicken free dry food;
- dog food sensitive skin;
- single protein dog food;
- dog food sensitive stomach.
If the engine only looks for the exact string, some of these searches return poor or empty results. And yet the intent is commercial: the user wants to buy.
That is the point: site search has to connect human language, catalogue structure and selling pages.
Synonyms: the catalogue has to speak the customer's language
Synonyms are not an SEO nerd detail. They are recovered sales.
Every sector has different words for the same thing:
- shoes, sneakers, trainers;
- food, kibble, feed;
- supplement, vitamin, tablet;
- sofa bed, futon, pull-out bed;
- cover, case, sleeve;
- air conditioner, split system, AC unit;
- cleanser, face wash, facial soap.
An eCommerce that does not handle synonyms forces the user to guess how the catalogue was written. A well built one does the opposite: it understands the customer and takes them to the right products.
The same applies to misspelt brands, abbreviations, trade names, old product lines and mixed-language terms.
Typos: spelling mistakes should not wipe out sales
On mobile, users make mistakes. They type in a hurry, use small keyboards, dictate by voice and shorten words.
A good engine has to handle errors such as:
kibleinstead ofkibble;hypoallergnicinstead ofhypoallergenic;womens perfume dolcegabanainstead ofDolce & Gabbana;shampo dogsinstead ofshampoo dogs;airpods cseinstead ofAirPods case.
If a typo sends the user to an empty page, the problem is not the user's. It is the site's.
Search needs error tolerance, automatic correction, suggestions and ideally some semantic logic. It should not give up at the first character out of place.
Zero results: the most underrated report in eCommerce
No-results searches are a gold mine.
Every query with no results can mean one of these things:
- the product exists but the engine cannot find it;
- the product is missing and the market is asking for it;
- the catalogue uses names that are too technical;
- a synonym is missing;
- a facet is missing;
- a category is missing;
- an SEO page is missing;
- the brand is written in several different ways;
- an Ads campaign is bringing in irrelevant traffic;
- the user is searching for a need, not a product.
The site search report should not be looked at once a year. It should be part of recurring work on catalogue, SEO, UX and campaigns.
When a query has many searches and few results, there is almost always an opportunity.
Can site search improve SEO?
The search bar is not automatically an SEO strategy. In fact, internal search pages generated with URLs like ?q=... usually should not become indexable pages.
But site search data is gold for SEO.
If a lot of people search for "grain free dog food", perhaps that should not stay just an internal search. It can become:
- a canonical facet page;
- a dedicated category;
- a menu section;
- a block inside the main category;
- an FAQ;
- an informational piece linked to the products;
- a more specific Ads campaign.
The difference matters. You do not have to index every internal search. You have to use those searches to work out which real pages to create.
The same principle applies to eCommerce facets: not every facet should become an SEO page, but some facets deserve clean URLs, a title, an H1, copy, FAQs, products and a self-referencing canonical. We explained it in the guide on which eCommerce facets to index and which to block.
Site search and facet pages: when a query becomes a page
A practical example:
- many users search for "dog food for neutered dogs";
- the catalogue already has suitable products;
- the main category is
/dry-dog-food/; - the facet exists but generates a technical URL such as
?f=neutered; - the query has SEO demand and commercial intent.
In this case it can make sense to create a real page:
/dry-dog-food-neutered.htm
With a dedicated title, an H1, SEO copy, filtered products, FAQs and internal links. Not an indexed internal search. A properly built commercial page.
This is one of the strongest points of a custom eCommerce: site search does not stay isolated, it becomes an input for categories, facets, sitemap, content and automations.
Autocomplete and suggestions: guiding without forcing
Autocomplete should not just be "let me finish your word". It should help the user pick the best path.
A good autocomplete can show:
- best-selling products;
- relevant categories;
- brands;
- useful facets;
- popular queries;
- typo corrections;
- synonyms;
- alternatives when a term is ambiguous.
If I search for "sensitive", the site can suggest food for sensitive dogs, sensitive skin shampoo, digestive supplements or different categories depending on the catalogue. If I search for an unavailable brand, it can suggest similar brands or the right category.
The point is to reduce friction. Every character typed should bring the user closer to the right product.
Result ranking: finding is not enough, you have to sell
A search can find 200 products. But if it puts the wrong ones at the top, it is not working.
Ranking should take into account:
- relevance to the query;
- availability;
- the most consistent category;
- the brand searched for;
- popularity or sales;
- margin;
- price;
- reviews;
- newness or seasonality;
- products you should not push because of stock or margin.
Showing out of stock, off-target or barely relevant products first is an elegant way to lose orders. Site search has to work together with the catalogue and the commercial strategy.
Can AI genuinely help?
Yes, if it is not used as decoration.
An AI agent connected to the catalogue can:
- read titles, descriptions, brands and categories;
- recognise real synonyms;
- map frequent typos;
- work out which products answer a given need;
- suggest new facets;
- merge duplicate facets;
- uncover missing categories;
- turn no-results searches into actions;
- create autocomplete suggestions;
- flag badly described products;
- link reviews and FAQs to the results.
In our work on AI applied to eCommerce and automation, this is one of the most interesting cases: AI does not invent content at random, it helps you read the catalogue and surface buying journeys that would otherwise stay invisible.
Site search and AI chatbots
Site search and the AI chatbot should talk to each other.
If a user searches for "product for an itchy dog" and does not know what to choose, the engine can show products and categories. The chatbot can go a step further: ask about age, size, known allergies, preferences and budget, and guide them to a choice.
This logic works when the chatbot has access to catalogue, facets, brands, availability, reviews and frequently asked questions. Not when it is just a generic window answering with vague sentences.
We looked into this in the article on the AI eCommerce chatbot connected to the product catalogue: context is what makes the difference.
Site searches help Google Ads too
Internal queries are not only useful for SEO. They are useful for campaigns as well.
If a lot of people search for a product or a need after arriving from a campaign, you have to ask yourself:
- is the landing page specific enough?
- is the campaign bringing the right traffic?
- is the product they searched for visible?
- does the category contain useful facets?
- does the Shopping feed have correct titles and attributes?
- are there no-results searches generated by ads that are too broad?
Well tracked site search can surface both wasted budget and opportunities for more precise campaigns. So it also connects to the digital strategy and performance side.
How to measure site search
Search has to be tracked. Otherwise you are working on gut feeling.
The minimum metrics are:
- number of internal searches;
- most searched queries;
- queries with no results;
- queries with few results;
- clicks on results;
- add to cart after search;
- orders after search;
- abandonment rate after search;
- queries corrected from typos;
- synonyms that produced results;
- products searched for but missing from the catalogue.
GA4 can collect events related to site search, but flipping a switch is not enough. You have to check that the query parameter is read correctly, that the reports are useful and that the data turns into actual work on catalogue, SEO and campaigns.
The zero results page should not be a closed door
When nothing is found, you should not just show "no results".
A good zero results page should offer:
- a corrected version of the query;
- similar products;
- nearby categories;
- alternative brands;
- recommended facets;
- best-selling products;
- a quick contact option or the chatbot;
- a way to report what they were looking for.
The zero results page should not be a defeat. It should be an intelligent fork in the road.
Checklist to work out whether your site search is losing sales
Check these signals:
- a lot of searches return zero results;
- search does not correct simple mistakes;
- misspelt brands are not recognised;
- common synonyms do not work;
- categories and products are badly mixed together;
- out of stock products appear before available ones;
- autocomplete only shows words, not useful paths;
- you do not know which queries generate orders;
- there is no report of no-results queries;
- nobody uses that data to create facets, categories or content.
If you find more than three problems on this list, site search is not a detail to fix "one day". It is a daily loss.
How we handle it on a custom eCommerce
On our custom eCommerce, site search is not designed as an isolated bar. It is part of the catalogue system.
That makes it possible to:
- add synonyms in a structured way;
- correct frequent typos;
- read no-results queries;
- suggest products, categories, brands and facets;
- connect search to AI, reviews and FAQs;
- use the data to create new SEO facet pages;
- reduce empty searches;
- push available and relevant products;
- understand which products are genuinely missing from the catalogue;
- improve feed, SEO and Ads with the same data.
The important thing is that search should not just "work". It should learn. Every failed search is a signal. Every synonym added is an obstacle removed. Every query turned into a facet or a category is one more commercial page.
This also connects to the article on eCommerce facets, AI and automation: when catalogue, facets and search talk to each other, the shop becomes far easier to navigate and much stronger on the SEO side.
Useful Google sources
For this article we cross-checked the official Google guidance on eCommerce site structure, URL structure for eCommerce sites, handling faceted navigation and enhanced measurement in GA4, including site search.
FAQ: eCommerce site search, AI and SEO
Does eCommerce site search really help sales?
Yes. People who search on the site often already have a clear intent. If they quickly find relevant products, categories or alternatives, the probability of an add to cart and an order goes up.
Should site search pages be indexed on Google?
Usually not. Internal queries with parameters should not automatically become SEO pages. Site search data is instead there to help you work out which real categories, facets or content to create.
What are synonyms in eCommerce search?
They are different words that point to the same product, need or category. For example cover, case and sleeve. Handling them avoids zero results and brings the catalogue closer to how customers speak.
Why do typos matter so much?
Because many users search from mobile and get words, brands or spellings wrong. An engine that does not tolerate simple typos loses high-intent searches and therefore potential orders.
Can AI improve site search?
Yes, if it is connected to the catalogue. It can suggest synonyms, identify no-results queries, propose facets, understand needs and connect products even when the query does not match the exact title.
Which metrics should I check?
Most frequent searches, no-results queries, clicks on results, add to cart after search, orders after search, corrected typos, synonyms used and products searched for but missing from the catalogue.