How to leverage the Google Natural Language API to boost your ASO Efforts

Google Natural Language API

The world is changing fast – and Google is changing with it. Over the last year, Google’s ongoing push on their machine learning and AI capabilities has continued to accelerate. Everyone knows about ChatGPT, but did you know that Google recently relaunched their AI model chatbot Bard as Gemini? The likelihood is that we’re going to see this more broadly utilised as they catch up to the current market leader. But there’s other tools they’re launching too which are more relevant to us app marketers.

Some eagle-eyed ASO specialists may have noticed the beta launch of a feature that automates the writing of Google Play long descriptions for Custom Store Listings (custom landing pages on the GP store). While this is limited in use case right now, our expectation is that before too long we’ll see this launched as a feature for your main store listings. We’ve also seen the launch of a translation feature which can automatically translate your long description into 10 different languages!

Great, right?

Or perhaps not. As Google grows in confidence with their machine learning tools, we’re likely to see them push these more as the solution to app developers. But we need to be aware of how this might affect our strategies – we don’t know how these tools would help us with the core goals of ASO, discovery and conversion. Plus, if everyone uses the same tools to create their listing, we’re all likely to lose any competitive edge that we may have had with more control over the content.

Having said that, there is a machine learning tool from Google which is completely underrated when it comes to ASO – and that’s the Google Natural Language API (GNL).

Before we dive into what GNL can do for us, it’s important to be aware of some key background changes we’re seeing to ASO. There have been an unusually high number of algorithm changes in the past 6 months, and with that we’re seeing diminishing impact from traditional, established ASO techniques such as keyword density as a tool for visibility. It seems that, as with all other areas of Google, they’re trying to improve their machine learning within the algorithms to understand text better, allowing them to understand true relevance of particular keywords based on broader categorisations of apps and their functionality. This in turn means less importance is placed on keyword density, because Google can determine itself what is relevant.

So where does GNL come in?

The aim of the Google Natural Language API

The ultimate goal of the Google Natural Language Processing API is to help computers understand language as well as we do. Natural language processing is the driving force behind things like virtual assistants, speech recognition, sentiment analysis, automatic text summarisation, machine translation and much more. It is Google’s proprietary system of understanding a body of text.

In relation to ASO, this would mean that Google can determine which keywords to associate with your listing without the use of keyword density and make better associations with a broad range of terms, rather than those specific keywords targeted. The GNL API allows us to break down the text and how Google understands it, which for ASO marketers means a better understanding of how Google is perceiving your app.

Let’s look at the key areas that GNL will give us insight into, and how this might help with ASO:

Sentiment Analysis

Sentiment analysis inspects the given text and identifies the prevailing emotional opinion within the text, especially to determine a writer’s attitude as positive, negative, or neutral. This feature is probably the least relevant to ASO but can give overall context on how the algorithm is reading your tone of voice.

Entity Recognition & Salience

Entity Recognition is the task of identifying and categorising key information (entities) in text. An entity can be any word or series of words that consistently refers to the same thing. Every detected entity is classified into a predetermined category. For example, an entity recognition machine learning model might detect the words “Yodel Mobile” in a text and classify it as “Company”.

For each entity, you will receive a salience score, which is a way of measuring content relevance. The salience score for an entity provides information about the importance or centrality of that entity to the entire document text. Scores closer to 0 are less salient, while scores closer to 1.0 are highly salient. E.g. If you were creating content about ASO marketing, the word “ASO” might receive a salience score of 1, because of how relevant it is to the content, whereas the words “and”, and “’but” may receive a lower score as they’re not specifically relevant to the content.

Incorporating relevant entities related to your app’s category, niche, or functionality within your metadata can significantly enhance relevance and context for both users and app store algorithms. Entities are recognized content elements like names, places, or specific features, which help algorithms understand the subject matter of your app more deeply.

Entities should be integrated in a way that feels natural and relevant, ensuring that they are recognised by Google’s algorithms. Additionally, placing key entities strategically at the beginning and end of your description could potentially enhance their visibility to both users and search algorithms.

Categories

Within the analysis from GNL, content is categorised into thousands of categories and sub-categories, and we can use it to understand the way that Google reads any kind of copy. For ASO, it’s especially useful for understanding what categories and therefore keyword types Google may associate your app with.

We input text into the Google Natural Language API, and it tells us how confident it is that our text fits into one or multiple categories. By optimising this category confidence score (ranging between 0 and 1), we can increase the field of search terms that Google considers relevant for our apps.

This is perhaps the most important element of GNL in relation to ASO. Optimising metadata to increase confidence scores within a category or across multiple categories is a key technique for improving keyword ranking that we’ve seen success with. A higher confidence score associates you more strongly with that category and allows Google to determine a wide range of potentially associated keywords which you can then rank more highly for.

You can achieve increased confidence scores by making sure to use clear natural language relevant to the categories you’re targeting. Avoid keyword stuffing and make every effort to include useful information about your app’s features and benefits and the ways it can solve a user’s pain points. It would also make sense for you to anticipate the user intent behind those who will be searching for apps like yours and tailoring accordingly. When doing this, think about highlighting unique selling points, key features, and benefits that differentiate your app from competitors.

Case Studies

We have been using these strategies for just over a year now and can share a few anonymised examples of the impact we have seen.

You can see below a keyword ranking chart for one of our clients, where the y-axis indicates Total Ranked Keywords, and the colour coding indicates how highly the app ranks for those keywords. The big dips in ranking signal significant algorithm updates which trigger keyword reindexing, as Google redefines some of the factors in how they determine keyword relevance.

What is crucial is that with GNL optimisation, we were able to counteract and even benefit from these algorithm changes. Over time we saw massive gains in our keyword strategy, rather than being disadvantaged by the turbulence in the store.

This translates into real world impact too. Below you can see an example of one of our other clients – by optimising both the core category score and additional categories, we were able to drive over 500% increase in our organic Explore installs.

By working with the algorithm and trying to understand how we were perceived, we were able to drive significant performance improvements even with major algorithm updates taking place over that period.

Conclusion

In this post we have been able to explore the ways that NLP can be used alongside, or even to the benefit of your ASO efforts. Ultimately, this type of innovation only means new strategies and revised approaches to your app marketing. For more in depth knowledge on these new developments, feel free to reach out to us so we can help with any queries that you may have.

Smokehouse Yard, 44-46, St John Street, London, EC1M 4DF 🇬🇧

Smokehouse Yard, 44-46, St John Street, London, EC1M 4DF 🇬🇧