App Store Optimization (ASO) is a foundational step in building an effective marketing strategy for your app. Mobtimizers is always looking for new tools and technologies to help with this most crucial aspect of app marketing. Artificial intelligence (A.I.) is becoming a ubiquitous tool in many aspects of marketing. And it is now becoming possible to complete more complicated tasks with regards to apps and app marketing.
In this article we’ll demystify A.I. capabilities by showing specific examples of how Mobtimizers explored using the large language model ChatGPT 3.5 in optimizing their clients apps for ASO. Before we delve into the specific examples, let’s review the many ways one could use A.I. in ASO and app marketing.
As A.I. continues to advance and evolve, its impact on ASO strategies and app marketing techniques is becoming more pronounced. With its ability to analyze vast amounts of data, identify patterns, and help make data-driven conclusions, A.I. empowers marketers and agencies alike to optimize user acquisition and enhance the visibility of their apps. A.I. brings efficiency and speed to the realm of ASO and app marketing.
Here are some ways that A.I. can work for App Store Optimization and app marketing:
A.I. can help in identifying the most relevant keywords to target for a particular app. You can analyze the competition and identify the keywords that have the highest search volume and the lowest competition. It can also suggest related and long-tail keywords that can be used to improve the app’s visibility. Furthermore, artificial intelligence technology can be used to help with localization, translation, metrics, and analysis.
A.I. can be used to produce A/B testing elements for the app store listing, such as conceptual ideas for app icons, screenshots, and somewhat well-structured copy for short and long descriptions. It can analyze the user behavior data and suggest the most effective elements that lead to higher conversion rates.
A.I. can help in identifying the most effective channels for user acquisition. It can analyze the user data to identify the demographics and user behavior patterns that are most likely to convert into app downloads. It can also suggest the most cost-effective channels to target these users.
Theoretically, A.I. can help in predicting the future performance of the app based on the historical data. It can analyze the user behavior patterns and identify the factors that lead to higher engagement, retention, and revenue. Predictive analytics like these, will require A.I. in your tech stack, including a connected long term memory (e.g. Pinecone.io).
But as A.I. continues to develop, we should remain cautious and not give away all the responsibilities to a tool that is not yet an “expert” as humans are. The strategic decisions, task management, the oversight, fact-checking, and more, will remain with us humans for some time. And even the simple tasks still require human oversight, as we shall see below.
And with new technological possibilities there is always the challenge of remaining true to one’s core values. At Mobtimizers, we believe in providing organic growth and delivering personalized solutions for each client. Therefore, we must always balance the use of new tools and technologies with our team’s varied long-term expertise along with our clients goals.
Mobtimizers has been keeping an eye on the developing A.I. capabilities as they pertain to ASO for over a year. And over the last few months we have implemented ways of using A.I. to fine tune and assist us in enhancing deliverables where feasible. But, as of now, A.I. is still just a tool – one of many – and we prefer to use it during developmental and exploratory stages, while ensuring our personal and human expertise remains the final arbiter in delivering analysis, copy, creatives, solutions, and growth opportunities to our clients.
With that squared away, let’s take a look at some examples of how we utilized A.I. as part of our ASO work for our clients.
One of our clients is a European based, global weather app. Below we will delve into detail on the specific tasks we explored in using A.I. for ASO. There are many foundational tasks involved in optimizing apps, from title, subtitle, short description, long description, keywords, graphics, CTAs, videos, etc. Further, there is also the app analysis, Value Proposition Mapping, competitive analysis, market analysis, and strategic decisions based on those. For our research here we explored working with A.I. on a few foundational tasks.
Using A.I. to translate and localize keywords
To begin with, we first input into the A.I. what app user demographics are. We fed the A.I. our optimized keywords in English and asked it to review and optimize as well. Then we tweaked and added competitors to the A.I. request for a more robust competitive keyword list. Finally we reviewed this comprehensive list to determine the keywords relevance to our client.
Next, we asked the A.I. to help translate and localize this list. First from English to Danish, then from Danish to Swedish. We have native speakers on our team so this made sense for us to translate into languages we understand first. Then we fact-checked the A.I.’s work, that it understood the commands and gave relevant results.
Reviewing the translated lists, we then asked it to remove one keyword from the Swedish list that was not relevant, and the A.I. apologized, then regenerated an updated list without that keyword.
We fed the A.I. a list of Dutch competitors, and requested a list of keywords in Dutch based on the competitors. We then translated the Dutch back to English to fact-check, and had a Dutch native speaking consultant review the final list. We then removed irrelevant keywords.
Continuing to fine-tune the Dutch keywords, we then asked the A.I. to “ improve this list targeting a weather app in the Netherlands and competitors, keep it relevant and add search volumes.” The A.I. gave relevant results ordered by search volume with highest volume first. We then fact-checked in an ASO tool to ensure the volumes were correct and the keywords given were relevant.
Using A.I. to translate and localize short descriptions
We also tested using A.I. in exploring various translations of our optimized app short descriptions. Translating into certain languages (for example, Danish, Swedish, and Dutch) creates a challenge with the 80 character limit of a short description. In this case, the A.I. was fast and helpful in trying out various translations, while swapping out key terms for others that would fit the limit in several translations we were working on. It also helped us keep true to the original optimized version and text that we wanted to maintain.
Specific examples of directions given to the A.I.
For the weather app being translated into several languages, we asked the A.I. to review our optimized short description for localization. We asked: “If this sentence has to be 80 characters, which word would you take out and why?” The A.I. responded: “I would take out the word “onze” (meaning “our” in English) as it is not necessary to convey the main message of the sentence and removing it brings the character count down to exactly 80 characters.”
Furthermore, we asked the A.I. to take our Swedish localized short description and make it easier to read. It went from “Få exakta väderprognoser och live väderkartor med vår väderradar app” to “Få exakta väderprognoser och se live väderkartor med hjälp av vår väderradar-app.” We then asked for it to be translated back to English so we could review it again.
We asked the A.I. what changes it had made. It is interesting to see how the A.I. responds and notes the edits it made. It said “I replaced ‘vår väderradar app’ with ‘med hjälp av vår väderradar-app’ to make the copy flow better and sound more natural in Swedish.” Does the A.I. really understand how natural Swedish should flow? To be determined.
We also asked the A.I. to double check the character length by asking: “Is the new sentence 80 characters max?” It replied, “No, the new sentence is 81 characters long.” And then it took out 2 words and added the word radar instead. Good progress! However, it’s interesting to note that the A.I. “forgot” the original instructions that the description maximum is 80 characters. A good reminder for users to continue to fact-check responses when using A.I.
Using A.I. for market analysis
After some basic exploration we further tested the A.I. capabilities in more complex tasks for market research. We asked the A.I. “What is the demographic profile of users for this app?” It gave us a caveat first, “As an A.I. language model, I do not have access to specific information about you or your business unless you provide it.” True, A.I. must be given certain data points and information before processing internal analysis.
For this request only one specific result was given: that weather app users were more likely to be between the ages of 25 and 54. However, this was not necessarily specific to our client’s app. It further made more generalizations such as, “Users of a weather-related app could be of any age, gender, and location.” It went on to “list” results for each key demographic, however the results were all vague, without any specifics. For example, “weather app users could be male or female, users could have any income, live anywhere,” and so on. So this A.I. task was not successful.
Conclusion
As with all new tools and technologies we must explore the possibilities with patience and caution. Over the last few months of testing A.I. capabilities on ASO and market research tasks we clearly saw some benefits and limitations.
A.I. is certainly capable of delivering on simple tasks to aid us humans in saving time and some research. With specific data fed to it, and small tasks asked of it, the A.I. helped us better understand our target audience, helped us save time, and be more efficient in certain tasks.
However, these basic tasks continue to require oversight and much guidance from human users. Even in simple tasks there were often errors introduced in the initial results. We must continue to be diligent in reviewing, fact-checking, and synthesizing results. And while it’s a helpful tool, A.I. is a long way away from being the expert in this dynamic.
We are excited to work with the new A.I. technology as it matures. And with the continued integration of A.I. into our solutions toolbox and our clients’ tech stacks, experimentation with the technology is in our best interest. We’ll continue to explore the formation of A.I. tech stacks in a future article.
*This article was written by a human, with examples of tasks produced by A.I.
