Autosuggest

Autocomplete · Search Suggestions · Type-Ahead Search · Query Suggestions

Definition

Autosuggest is a site-search feature that, while the shopper is typing, offers query completions, categories and specific products, so they can reach a result without typing the full query.

What autosuggest shows

Suggestions appear after the first few characters and refresh with every keystroke. The shopper does not have to finish the query: they pick a ready option or jump straight to a product page. It is the part of site search a shopper sees before the results page itself.

Suggestion type	Example for the input “sne”	When it helps
Queries	“sneakers men”, “sneakers for running”	Broad query that needs refining
Categories	“Shoes / Sneakers”	Shopper wants to browse with filters
Products	card with image, price and stock	Query for a specific model
Corrections	“snaek” → “sneak…”	Typo or the wrong keyboard layout

Autocomplete and suggestions are not the same thing

Autocomplete finishes the word or phrase being typed — one option, right in the search box. Suggestions offer a list of options of different types and can lead not to a results page but straight to a category or a product. A separate scenario is suggestions on an empty box: the shopper’s recent queries and popular queries before they start typing.

Where suggestions come from and how to order them

Query log — frequent queries that led to clicks and orders.
Catalogue — category, brand and product names, including new arrivals that are not in the log yet.
Dictionaries — synonyms, spelling variants, keyboard-layout fixes and typo correction.

The order of options is set by query popularity, query conversion and stock. Personalization adds the shopper’s history: favourite brands, gender, size. Semantic suggestions work with intent: typing “gift for dad” brings up categories, not just strings that start with those letters.

How to measure it

Suggestion usage rate = searches started with a click on a suggestion / all searches.
Suggestion CTR = clicks on suggestions / impressions of the suggestion block.
Conversion of searches via a suggestion versus typed searches.

Common mistakes

An empty block on frequent input — worse than no block at all.
Suggestions for out-of-stock products and for queries that return zero results.
No layout correction: “zoga mat” typed on a German QWERTZ keyboard, where Y and Z swap places, gets no suggestions.
A suggestion leads to a results page where the wrong product comes first.
On mobile, there are more options than fit above the keyboard.
Seasonal suggestions linger after the season ends because the dictionary does not follow the assortment.

Frequently asked questions

What is the difference between autocomplete and search suggestions?

Autocomplete finishes the query being typed: “sne” → “sneakers”. Suggestions go further: several query options, categories, brands and products, plus typo correction. In everyday use the terms are interchangeable, but in a brief for developers it pays to list exactly which types of suggestion you need.

Should suggestions show products, or are queries enough?

It depends on the query. When a shopper is after a specific model (“iphone 15 pro”), a product card with image and price right in the suggestions saves a step — they go straight to the product. For broad queries (“dress”), refinements and categories are more useful. A common desktop compromise is two columns: queries and categories on the left, products on the right.

Can suggestions be personalized?

Yes. The order of options can follow the shopper's history: viewed categories, favourite brands, gender, size. An empty search box on focus is also a place for suggestions: the shopper's recent queries and what is popular in their categories. Test changes with an A/B test — personalization must not hide the most frequent queries.

How do I know autosuggest is working?

Look at the share of searches started with a click on a suggestion and at the click-through rate of the suggestion block. Compare the conversion of searches via a suggestion with that of typed searches. Separately check how many suggestions lead to an empty or near-empty results page — there should be none.

Gravity Field

Autosuggest in Gravity Field AI Search

Gravity Field AI Search runs on vector semantic search, with suggestions that follow the intent behind the query and tolerate typos.

✦Autosuggest driven by the intent behind the query

✦Typo tolerance — resilient to misspellings and spelling variants

✦Vector semantic search across the product catalogue

✦A lower zero-results rate through meaning-based matches

Request a demo →