The return rate of domestic AI glasses exceeds 30%. Can Meta trigger an "iPhone moment"?

Recently, at the developer conference Meta Connect, Meta launched three AI glasses with different positioning and announced the opening of the developer platform.

"Glasses are the only devices that allow AI to 'see what users see and hear what users hear,'" said Zuckerberg as he entered the venue wearing Meta's first "AI + AR" product, the Meta Ray-Ban Display. As a bellwether for the industry, this product is also regarded as an important example of the consumer-grade implementation of AI and AR technologies.

Image source: Meta

Zuckerberg and Kevin Kelly, the author of Out of Control, share the same view on the next 25 years. Smart glasses will replace most mobile phones and become tools for searching all data, including the data you can see and the data you generate yourself, such as your every expression. Then, a world combining the real and the virtual will be formed. Kevin Kelly defines this as "a transparent mirror world," a world without privacy.

Will you sacrifice your privacy when the mirror world can provide you with customized services? If, as Meta hopes, AI glasses are about to enter a booming "spring," how would you answer this question?

Perhaps the answer should start from this "spring." After all, the return rates of many new products launched this year remain high.

Behind the booming market, the return rate remains high

How popular AI glasses are this year can be illustrated by a set of data.

The semi-annual report on smart glasses released by JD.com on August 26 shows that in the first half of 2025, the sales volume of this category increased by more than 10 times year-on-year, the number of settled brands increased by about 3 times year-on-year, and a new product was launched on average every 9 days. IDC data also shows that in the second quarter of 2025, the global smart glasses market shipped 2.555 million units, a year-on-year increase of 54.9%; the Chinese smart glasses market shipped 664,000 units, a year-on-year increase of 145.5%.

Since this year, major domestic and foreign manufacturers have also intensively announced their smart glasses plans. For example, Amazon plans to launch consumer-grade AR glasses codenamed "Jayhawk" by the end of 2026; during the 2025 World Artificial Intelligence Conference, Alibaba exhibited the Quark AI glasses equipped with a dual-optical engine display system and announced that it will be officially launched within this year.

A series of developments indicate that smart glasses are rapidly developing in multiple dimensions, including technological integration, product innovation, and ecosystem cultivation.

However, behind the impressive sales figures, the high return rate has become an unavoidable reality for the industry. According to statistics from the industry media XR Vision, the return rate of AI glasses on platforms such as JD.com and Tmall is about 30%, and on the Douyin platform, it is as high as 40% - 50%. Among them, "insufficient functional practicality" is the core problem reported by consumers.

Huang Xin (a pseudonym), a consumer, returned the Xiaomi AI glasses after using them for 5 days. He told reporters, "The Xiaomi AI glasses, which only have a camera but no lens display, are more like a simple combination of 'headphones + camera,' and the actual scene experience did not meet expectations." For example, during navigation, due to the lack of visual display, it is difficult to accurately identify the road at intersections relying only on voice announcements. Eventually, he still had to take out his mobile phone to check the map.

Image source: IT Times

Meta has once again become a leader in the industry.

Compared with the previous Ray-Ban Meta, the Meta Ray-Ban Display has two major innovations. One is that a color display is installed inside the right lens, which can complete visual functions such as message viewing, video calls, real-time subtitle translation, and route navigation through Meta AI; the other is that it is equipped with the Meta Neural Band, which allows users to interact with the display using gestures.

With the support of visual functions, the Meta Ray-Ban Display can enrich daily usage scenarios to a certain extent. For example, AR navigation can project route arrows on the road surface; message pop-ups can float in front of the eyes, allowing users to process information without looking down; in the office scenario, the "real-time transcription" function of the AI glasses can automatically organize meeting speeches into written minutes and project them in front of the eyes.

Image source: Meta

"The core of AI glasses is to achieve the in-depth service value of 'data accumulation + decision-making assistance' based on the technology of 'all-day wearing + visual and auditory perception,'" Jiang Gaoping, the general manager of Shenzhen Yitian Technology, told reporters from IT Times. For example, it can recommend suitable stores and products according to the user's past consumption preferences and budget when shopping; combined with the needs recorded in the meeting, it can remind key information during project decision-making, becoming a "private decision-making assistant."

However, most of these functions are just supplements or minor innovations to existing devices. At least for now, AI glasses still lack a core function that can truly impress consumers.

Without regulations, problems can only be solved while developing

The privacy and security issues brought by AI glasses have always been controversial. One of the most criticized application scenarios is secretly recording with the camera.

The China Academy of Information and Communications Technology's Terminal Laboratory once tested the reminder function of the Ray-Ban Meta Wayfarer during shooting and recording to evaluate the device's solutions for privacy protection. Most users thought that the smart glasses performed moderately well in terms of the brightness of the reminder light and the volume of the reminder sound, which was helpful for improving the user experience. However, the reminder light may be too dim in outdoor environments, and the color is not easy to distinguish for some users.

It is reported that the Ray-Ban Display released by Meta this time continues to use the shooting reminder light. There is a sensor on each side of the glasses. If the light intake on both sides is detected to be inconsistent, the camera cannot be started. This is to prevent the behavior of secretly taking pictures by covering the reminder light.

However, a technology blogger previously tested the Ray-Ban Meta and found that by covering the right reminder light with tape and covering the left camera with a hand to keep the light intake on both sides consistent, after hearing the reminder sound for starting shooting, releasing the hand covering the camera, one can take pictures while covering the reminder light.

After the Xiaomi AI glasses were launched, a blogger also tested and found that by simply smearing with a marker, the shooting reminder light could be successfully covered, and it was difficult to tell from the appearance whether the reminder light was on or not. After the shooting function was turned on, people around could hardly notice that the glasses were in the recording state.

Image source: IT Times

Wang Yiru, the financing director of XREAL, said in an interview with the media that the industry is still in the stage of discussion and exploration regarding issues such as regulating AI use and avoiding privacy infringement. Currently, the scale of the smart glasses market is not large enough, and the level of attention is not high enough. We can only predict some risks and solve problems while developing.

AI glasses should have "strong reminders"

Facing the concerns brought by AI glasses, what should consumers do? How can privacy and security be guaranteed while innovating? All these issues have emerged in the consumer market, and it requires the joint efforts of the industry, the legal system, and consumers themselves to install a "safety brake" for the rapid development of AI glasses.

Image source: Meta

Yu Zehui, a senior consultant at Beijing Xingye Law Firm, told reporters from IT Times that AI glasses may bring risks in five aspects: data privacy, notification and consent, intellectual property, product liability, and algorithm ethics. Industry technology empowerment, legal mechanism guarantee, and manufacturers' design concepts are the three cornerstones for ensuring compliance.

First, in terms of data privacy, according to relevant laws and regulations such as the Personal Information Protection Law of the People's Republic of China, enterprises must follow the principle of minimization, the principle of notification and consent when collecting, processing, storing, and using personal information, and take necessary technical and management measures to ensure data security.

This requires enterprises to set the highest level of privacy protection in the factory settings of the product. During use, only the data that is directly related to and necessary for the core functions of the product should be collected. Data such as facial features, fingerprints, voices, and physiological characteristics, which belong to sensitive personal information, must obtain the user's separate consent before shooting and collection, and clearly inform the purpose, method, and scope of processing.

The public has the right to stop AI glasses from shooting.

Yu Zehui suggested that AI glasses manufacturers should make "strong reminders" a design aesthetic on the one hand. It can be an elegant halo around the frame or a unique and non-disabled sound reminder; on the other hand, specific gestures can be set. For example, if the other person raises their palm, the AI should immediately stop recording once it recognizes the gesture.

"The potential of AI glasses is far beyond shooting, but lies in its AI analysis ability behind. If functions such as face recognition, emotion analysis, or identity tagging are integrated in the future, new risks will emerge one after another," Yu Zehui further suggested. Before the algorithm model is launched, an algorithm review mechanism should be established for strict ethical review and bias testing, and records should be kept; in product design, a clear and convenient switch should also be provided to allow users to turn off all personalized recommendation and analysis functions based on personal information with one click.

This article is from the WeChat public account "IT Times" (ID: vittimes). Author: Shen Yibin, Editors: Hao Junhui, Sun Yan. Republished by 36Kr with permission.