Emblematic of Apple App Store has recently been updated to include summaries generated by user AI, and now we know how it all works.
Apple explained how its AI generates summaries of the opinions of the App Store.
In October 2024, an article in the Unclosed App Store revealed That Apple wanted to summarize the reviews of user applications using artificial intelligence. Months later, in March 2025, the functionality became available for the general public with the release of iOS 18.4.
While we already had some details on the summaries of the revision generated by Ai-Ai, a new message On Apple’s automatic learning blog explains the subtleties and the details of the functionality.
The characteristics and objectives of the examination summaries generated by the AI
The ultimate objective of these summaries is to provide users with a clear image of the reviews of an application, so that they can more easily decide to buy or not install a particular application. However, summarizing user reviews, Apple had to make sure that the IA release was up to date and that it did not include off -topic or offensive information.
App Store applications often receive updates, and changes such as new features, bug corrections or integrated elements often influence user reviews. Application criticisms themselves also vary according to style, length and even relevance. Apple’s summary of AIA was to take into account all these factors, so that the company has implemented a process in several stages.
How Apple’s AI summarizes user reviews
First, user reviews with spam and blasphemies are filtered. Eligible criticisms then went through a series of different LLMs or large language models, which extract key information from user reviews. After that, the common themes are aggregated and the feeling of users is balanced. The result is a summary generated by AI which reflects a wide feeling of users, with a length of 100 to 300 words.
During the first phase of the process, known as “extraction of information”, user reviews are summarized to separate information. Apple says that these ideas encapsulate “a specific aspect of the review, articulated in standardized natural language, and confined to a single subject and feeling”.
“Modeling of dynamic subjects” allows AI to compare relevant subjects on different criticisms, so that the software can identify the most important subjects discussed. Approach and terminology resemble AI test applicationsthat we described in 2024.
For each application, a set of subjects, as well as the “most representative” information for these subjects, is used by AI in the creation of summaries. The specially designed LLM assured that the feeling of users was balanced and that the summaries maintained the required shape and length.
During development, the summaries generated by the Apple were evaluated for characteristics such as earthing, composition, utility, etc. This part of the process involved human examiners, which serves as a indication of the seriousness of its development of AI summary.
Apple’s blog details All the steps mentioned here, with more specific information on the technology used during each part of the process. Overall, the iphone The manufacturer’s approach guarantees that the summaries generated by the AI of user reviews are accurate, useful, without spam and up to date.