45% of searchers take less than 5 seconds to act on a Google search results page.
You can assume that most of your users aren't carefully analyzing the search engine results page before deciding which result to click.
What does this have to do with Google product taxonomy?
Your customers are making purchase choices based on the results that come back from searches relative to their initial query and what products they expect to match that. Having your products correctly categorized helps ensure that your catalog is as easy to search as possible. The hierarchy structure of a taxonomy allows your products to naturally refine into more specific categories as the user improves their search granularity for what they’re looking for. If you have 10 different types of shoes in your catalog, you can create categories for them so users only see the specific type of shoe they’re looking for if they search for it. They don’t want to dig through a generic “shoes” category if they know they’re looking for “oxfords”. The taxonomy structure of Google Product Taxonomy allows for this granularity.
In this article, we'll cover what Google product category taxonomy is, and how it can be used to benefit your business.
Before explaining Google's current taxonomy, let’s understand Google product categories.
The product categories are the groupings that you can assign to products that are similar and a user would assume should be together. The options you have for categorizing your products and their relationships to other products all fit into one large category tree. This system is known as a "taxonomy."
The right google product category for a given product is completely dependent on your catalog and how users interact with it to find what they are looking for. The categories you use and the depth that you go depends on how much variance is in your catalog.
The Google taxonomy (also known as “Google shopping taxonomy”) is Google's list of product categories used to organize goods in a particular data feed.
There are about 5,595 categories in total, although this number is subject to change as Google updates their taxonomy to account for new products and categories, and has slight variations based on country. Each category has a numeric id tied to it that can be tied to google shopping.
Each category in Google's taxonomy is carefully organized and labeled to help shoppers find products quickly and easily. Since Google's taxonomy can be standardized across sites, shoppers also benefit from consistent navigation and expect to find similar categories in your online store as the ones they are used to from other sites. For you as the merchant, the taxonomy is used to help merchants optimize their product listings to attract more customers and maximize their sales.
We mentioned there are currently around 6,000 categories. Each begins with a general main category such as "Animals & Pet Supplies" or "Apparel & Accessories."
From there, the categories branch off (subdivide) into sub-categories such as "Pet Supplies" or "Clothing Accessories."
Here's what the hierarchy in Google's product taxonomy looks like:
Apparel & Accessories > Clothing > Activewear > Motorcycle Protective Clothing > Motorcycle Jackets
Google's product categories are meant to organize and present your products in a standardized way.
Correct and thorough categorization helps optimize your website for multichannel marketing.
Additionally, Google can quickly determine if a user's search query matches your website's products, based on product categorization.
A couple of other channels that use the Google Product Category include Bing Shopping and Facebook Dynamic Product Ads.
Categorizing your products makes it easier for potential customers to locate your online store.
With SEO, relevancy of your results to user search intent is key. By adopting Google’s standard product categories you can ensure that your products are indexed correctly and show up in the results for the right searches. This way, you can get more organic traffic and potentially increase your sales.
Simply put, if their search terms match your products, they're more likely to visit and buy from your website.
While GTIN can assist Google in matching products to search queries, Google Shopping relies on Google Product Categories for both comparison and navigation.
Google’s standard product categories can also help you save money on ad campaigns.
Google decides what to show searchers depending on several characteristics and qualities such as Title and GTIN.
For each category, Google has various editorial criteria and advertising practices.
No matter what type of goods you sell, you must first define the relevant Google product category in order for them to be accepted.
When you categorize products in a way that Google understands, Google can show your ads to more relevant buyers. If more people who see your ads click on them, you will probably see higher click-through rate (CTR) and lower customer acquisition costs (CAC).
You can use Google product categories on Facebook to create additional product sets.
However, Google's Taxonomy contains multiple levels (hierarchies) that can help you simply categorize your products advertised on Facebook without breaking them into product sets.
Just as with Google’s native ads, standardizing your product categories on Facebook can help you see a higher ROI with Facebook ads. By using correct product categories for your listings, you can make sure that your ads reach the right audience. This way, you can create more effective ad campaigns, improve visibility, and increase your reach.
Outside of the reasons above that focus on interacting with external products that use google product categories there are a number of reasons your catalog should be categorized and use a hierarchical structure to better group products.
Google’s product taxonomy is a useful tool for:
Consumers are impatient and, in today's digital world, they expect instant gratification. This is why it is important to reduce the amount of time and clicks they have to take to find what they are looking for.
By helping your customers find products faster, you can not only reduce the bounce rate but also increase the number of items added to shopping carts.
Furthermore, shoppers may not even be aware of the products that you have available. If they can’t find a product in the category that they expected, they might assume that you don’t sell a specific item, even when you do.
Through the use of Google’s product taxonomy, you can provide consumers with the most streamlined and convenient shopping experience possible. This will give them the ability to quickly and easily find the exact product they are looking for, allowing them to make their purchase with minimal friction.
If your visitors can't simply discover what they want to buy, 79% will go to a competing site for identical goods.
Additionally, a lot of Google searches end without a click, leading to lost opportunities for businesses. To combat this issue, Google product categories are designed to help businesses like yours and your products to appear within relevant search results for potential buyers. When your products appear for more relevant searches, you will see an overall improvement in your click-through rates (CTR) and search rankings.
As a result, your business will be able to efficiently guide consumers to the products that they are searching for, thus improving your website traffic and ultimately your sales.
Businesses that make use of these product categories can benefit from the improved visibility and presence that comes from being listed as a relevant search result.
By improving the structure of your catalog you can make better product inventory decisions and further understand how your products break down into groups. With the depth focused approach of a taxonomy you can better understand the types of products you have. For example if you just have a “shoes” category and have 20 products in that category you might think you’re good to go, as potential customers won’t have much to search through. At this high level it also looks like all your “shoe” products are pretty similar. But if you’re using a taxonomy with depth to it and multiple sub categories you can split those up into more granular product groups that better fit searches. You might have 7 different types of shoes in that group of 20, and a customer just looking for “oxfords” does not want to look through the other types of shoes.
There are two main ways that catalog managers handle categorizing products.
As you can probably guess, manual categorization is extremely time consuming. It can seriously limit your ability to scale your catalog and onboard new vendors. Companies have to hire teams of virtual assistants to review these catalogs and categorize them correctly.
Some companies take it a step further by creating some high level keyword rules to try to whittle down the possible categories for a given product so manual reviewers don’t have to go through all categories in the taxonomy.
Regardless - manual categorization is a painful process that doesn’t scale easily without adding more manual resources, and money.
In 2025 automated categorization mostly falls into two main buckets for how companies are building systems to automate understanding which category is best fit for product data.
Companies will build systems using keyword matching and string matching to try to understand where the product belongs. These systems can be a bit difficult to scale as you usually have to keep making adjustments to the rules and parameters as your product data changes or you want to add new categories. The set of rules and comparisons that make sense for one vendor might not make sense for the other, and you then have to repeat the process of mapping these keywords and strings.
Recently companies have started using ChatGPT to try to categorize products. LLMs have a better understanding of language variations in product data so these systems perform better than the old school matching systems. They’re also a bit easier to get off the ground as they just require you to add your taxonomy to the prompt and your product data. ChatGPT also offers an API that you can plug into your product catalog to further improve your automation.
But there’s a few key issues we’ve seen with this workflow that lead to customers not sticking with it long term. For one GPT is definitely faster than manual categorization, but is much slower than matching algorithms. These large language models are incredibly slow when using them via an external API service, and there isn’t much you can do to speed it up.
The more common issue we see is that long term the accuracy of this system is capped. Its pretty easy to get to 80% or even 87% with smaller taxonomies with ChatGPT, but becomes incredibly hard after that. Everyone's product data and relationship between data and taxonomy is different, and ChatGPT isn’t trained on your specific product data relationships. Outside of what you give it in the prompt, it doesn’t know anything about your use case, your business, or complex domain specific knowledge. Any business case specific stuff you want to add to your categorization has to be correctly portrayed to that model in the prompt, while managing everything that goes into a high accuracy prompt (data variance coverage, hallucinations etc). These language models are not trained specifically on this task and use case, and that outside knowledge can bog down the ability for very specific use cases. It's why many AI practitioners end up choosing smaller task specific models in other domains over ChatGPT in the long run if their task is extremely granular and that type of model exists.
So what about us?
We’ve built a state of the art categorization specific ai architecture that focuses on product data categorization 24/7. It blows ChatGPT out of the water in terms of accuracy and speed. This architecture is trained on your specific data and learns how to differentiate product records from each other to better understand the exact categories that match even as they become super granular. It also fine-tunes over time as you use it so it gets better and better and understanding your specific use case and domain specific logic.
We’ve achieved results like:
97% top level and 93% 6 level deep accuracy on Google Product Taxonomy for a marketplace.
92% on a multilingual product catalog.
97% on a 5 level deep taxonomy for an ecommerce solutions company.
All of these companies were using old school keyword and fuzzy matching systems and had reached a limit on accuracy. Another downfall of matching systems or ChatGPT is they don’t give you a model derived confidence score based on the models understanding of how likely it is the result is correct. You can use these confidence scores for human in the loop intervention to understand how much of your categorized results are ready to be uploaded and which ones you should review.
Let’s take a look at how you can get started with Pumice.ai categorization to the Google Product Taxonomy.
Pumice.ai has a dashboard as well as an API endpoint that can both be used to categorize product records. CSV of product records are used in the dashboard and JSON requests are used with the API. Pumice has two required fields - title and description, but has optional fields such as image, brand, SKU, price. We strongly recommend using as many fields as possible allowing our ai to fully evaluate the product data based on the taxonomy.
Pumice.ai has two different types of categorization models - custom and fine-tuned.
Custom Models: These take your existing product info and the taxonomy you want to use, then blend them with our basic categorization model. They’re not trained on your specific data or how it connects, but they do use your data as a reference. This is handy if you don’t have enough training data yet to upgrade to the fancier model below.
Fine-tuned Models: These are fully trained on your product data and how it links to your taxonomy. They only understand your stuff, and they’re built using SOTA architecture. We guarantee to reach at least 90% accuracy during development or you don’t pay.. Every time we tweak the training, we’ll share a progress report so you know where things stand and what’s next. This is the best model when you have your own product category list or taxonomy.
If you’re using the API you’ll have to provide a model_id to interact with the specific model or taxonomy you need. If you have a custom Google Product Taxonomy model with the model_id we’ve provided you, you include it, if not the base model is “google”.
If you’re using the single product record API the format will be JSON with the same parameters for the model.
If you’re using the dashboard you simply have to choose the correct model from the dropdown under “dynamic categorization”. We load these models into your account so if you don’t see the one you’re looking for please let us know.
Once you’ve got everything loaded up just hit run or call the API! Our ai models will run in the background and return your data categorized in either a CSV or JSON request. The results will be in the exact same format as you provided with additional columns for the correct google product category and confidence score. If you’re using the batch API you will get a task_id back that you can use to check when your results are finished.
If you need larger volume via the API we have separate high volume infrastructure that is deployed as a standalone. We have customers running 50 million records per month on that infrastructure.
The models also return only one category (the best fit) by default, but can be adjusted to return multiple categories ranked in order from most likely to least.
Once your results are back from Pumice you can integrate them into your system. Most customers connect the API directly to their PIM or third party software for catalog management for a seamless onboarding experience. We’ve helped a number of companies set up workflows to automatically kick off Pumice.ai jobs when new products without a category hit their PIM and update the products automatically with the new category. Other customers hook their Shopify store directly to the Pumice.ai API to do the same thing and add these categories as “product type”.
If you're ready to leverage AI to automate your product information and catalog management for you, reach out today.
.png)