AI apps are all the rage in 2025, but a new study looked into exactly how much data they are collecting, and spoiler alert, it’s a lot.
Since ChatGPT, every business under the sun has been jumping on the AI bandwagon to make the technology part of their platform, with the goal of improving productivity and, ultimately, revenue.
However, with the technology advancing so fast, regulations haven’t been able to keep up, which has led to some serious data harvesting practices that may shock some users.
AI Apps Collecting the Most User Data
According to an analysis by Ecommerce Platforms, there are some AI apps that are a lot worse when it comes to protecting user data than others. In fact, almost all of the analyzed apps in the top ten collect more than 50% of the data inputted into the service.
Here are all of the AI apps collecting the most user data in 2025, based on the percentage of data collected versus the data on the app.
- Amazon Alexa – 93%
- Google Assistant – 86%
- Duolingo – 79%
- Canva – 64%
- Otter – 57%
- Poe – 57%
- Facetune – 50%
- Bing – 50%
- DeepSeek – 50%
- Mem – 43%
- ELSA Speak – 43%
- PhotoRoom – 43%
13. Trint – 43%
14. ChatGPT – 36%
15. Perplexity AI – 36%
16. Lensa – 36%
17. StarryAI – 36%
18. Wombo – 36%
19. Youper – 36%
20. FaceApp – 36%
21. Luma AI – 36%
22. Speechify – 29%
23. Pixai – 21%
24. Clipchamp – 14%
As you can see, the evolution of generative AI platforms like ChatGPT has not necessarily lead to worse privacy practices, with long-time stalwarts Google Assistant and Amazon Alexa leading the pack by a substantial margin. However, what that data is actually used for could be a more pressing question.
What Is User Data Being Used For?
In addition to tracking which AI apps track user data in general, the analysis from Ecommerce Platforms also investigated what user data is actually being used for any ranked AI apps on that as well.
For example, the research found that Canva was one of the worst companies when it comes to harvesting user data, not because of how much it collects but because of what it’s used for. 36% of collected data is shared with third-party advertisers and another 43% is specifically used for the company’s benefit.
This just in! View
the top business tech deals for 2025 👨💻
Alternatively, while Google collects 86% of data on the platform, only 21% is shared with third-party advertisers and only 36% is used for the company’s own benefit. Admittedly, neither is great, but there is a clear difference.
AI Terms & Conditions
In addition to investigating the data collection practices of these apps, Ecommerce Platforms also dug into the terms and conditions that exist for these platforms to see how long it took to get through them and how hard they were to read. And, another spoiler alert, the answer is quite long and very difficult.
The worst offender was Clipchamp, which took three hours and 16 minutes to read from beginning to end, mostly due to the jargon-laden copy that was one of the hardest to read. Bing came in second at two hours and 20 minutes, while the term and conditions for Google Assistant comparatively took a breezy 56 minutes total.
Suffice to say, AI hasn’t made the nefarious data collection practices of big tech any better. In fact, given the swaths of data required to train these models into anything that actually works, it’s safe to assume that data collection is only going to get more prevalent in the coming years.