ChatGPT is set to show ads in Germany from late August while becoming more personal and human-like in its interactions. This affects people using the free service in particular, but also anyone relying on ChatGPT for work, education, or private questions. The changes may make the assistant more useful, yet they also blur the boundaries between a tool, a data collector, and an advertising environment.
What changes when ads arrive?
OpenAI has been testing ads in ChatGPT in the United States since February 2026. Canada, Australia, New Zealand, the United Kingdom, Japan, South Korea, Mexico, and Brazil followed. According to the report on the German ad launch, Germany, France, Ireland, and Singapore will be added in late August.
OpenAI says advertising will help fund the high costs of new models, data centers, and additional computing capacity. The company aims to generate $100 billion in advertising revenue by 2030. Market researchers cited in the report doubt that this target is realistic. It should therefore be read as a corporate goal, not as a reliable revenue forecast.
For you, advertising first changes the nature of the service. An answer may no longer appear in a seemingly neutral workspace alone, but alongside commercial messages. The supplied sources do not fully explain how ads will be selected in the European Union (EU), or whether an ad-free experience will be limited to particular plans. The “way out” mentioned in the lead report’s title is not described in the supplied excerpt, so no price or specific subscription can be established from it.
There is also an unresolved tension. OpenAI and Anthropic reportedly say that they do not create advertising profiles from user data. At the same time, OpenAI is building an advertising business inside ChatGPT. Those positions do not necessarily contradict each other because ads can be displayed without a personal profile. The reports do not establish what separation will apply in practice.
What new data could be recorded?
Alongside advertising, OpenAI is expanding the context ChatGPT can use. Its macOS desktop app is receiving an optional feature called Computer History. It turns your activity into a timeline that ChatGPT can reference when handling later requests. The Computer History report says this can include clicks, keystrokes, and other events.
The feature is intended to learn how you work, suggest automations, and resume unfinished tasks. In one demonstration, ChatGPT found the last document a user had edited, checked through Slack whether it had been shared with other people, and summarized how the user had spent the morning. That could be useful at work because you would spend less time reconstructing where a task stopped. It also creates a more detailed account of your working habits.
Computer History is opt-in, according to the report. You can exclude specific apps and websites and delete individual entries. Content in private or incognito browser tabs is supposed to be ignored automatically, according to an OpenAI manager. OpenAI also says the feature does not capture images, video, or audio, relying on “events” instead. These are provider claims and were not independently verified in the supplied sources.
This differs significantly from an ordinary conversation. A regular chat mainly creates a record from material you deliberately type or upload. A computer timeline may additionally show which programs and websites you use, the order in which you work, and which task was left unfinished. More context may improve the assistance, but it also gives you more data to manage.
Why is the assistant becoming more human-like?
The new GPT Live voice feature uses pauses and filler words such as “hmm,” “yeah,” and a drawn-out “okay.” This makes conversations sound less awkward than earlier voice modes. An analysis of the updated speaking style describes these signals as a way to make conversations feel more familiar and human.
That natural delivery can make the feature easier to use. If you ask for the next step while cooking or discuss an outline during your commute, smooth transitions may be more pleasant than long pauses. Filler words also suggest thought and understanding, however, even though the system is not a human conversation partner. A warm voice is not evidence that an answer is correct, confidential, or capable of responsible empathy.
The source refers to the “uncanny valley,” meaning the discomfort people can feel when an artificial figure or voice seems almost, but not completely, human. ChatGPT presents an additional concern: a convincing voice may lower your hesitation about sharing private information. That matters because chatbots are already used for sensitive subjects.
A Bitkom survey of 1,003 people ages 16 and older in Germany found that 34 percent of surveyed Artificial Intelligence (AI) users consulted chatbots about health questions. At the same time, 43 percent of all respondents worried that their personal data was not safe with AI providers. These figures describe attitudes and behavior in Germany; they do not demonstrate the quality of medical answers or prove that a particular data breach occurred.
Pros and Cons of a more personal ChatGPT
Pros:
- Greater continuity – An activity timeline can retrieve unfinished tasks and reduce the need to explain the same context again.
- More natural interaction – Smoother voice conversations may be more comfortable during hands-free use or extended discussions.
- Practical time savings – Usage data indicates that ChatGPT is often used alongside education sites, job portals, and government information services.
- Basic controls – Computer History is reportedly optional, while apps, websites, and individual entries can be excluded or deleted.
Cons:
- A larger data trail – Work patterns, clicks, keystrokes, and services used may become part of the available context alongside chat messages.
- Commercial influence – Advertising creates a financial incentive to retain your attention within the conversation.
- Misleading familiarity – Human-like pauses and filler words can convey competence or understanding that is produced through speech patterns.
- Unclear separation – The reports do not conclusively explain how ad selection, user data, and model training will be kept apart.
Claims of greater productivity also require context. An analysis of real web usage examined more than 200,000 US households between 2021 and 2024. The researchers associated ChatGPT use primarily with productive tasks and calculated productivity gains ranging from 76 to 176 percent.
According to the study, however, users spent almost all of the time they saved on services such as TikTok and Instagram. Assistance with a tax return is a concrete everyday example: ChatGPT may speed up the chore, but that does not automatically turn the free time into another useful activity. The research covered US households during a period before the features discussed here. It provides context about behavior, not evidence of the effects of the new ads or Computer History.
What does this mean for your use?
If you are a beginner, a sensible first step is to take stock of your existing conversations. Swiss provider Proton has introduced AI Paper Trail, a tool that analyzes exported conversations from ChatGPT and Claude and summarizes what the services may have learned about you. According to a hands-on report about AI Paper Trail, it works only with those two chatbots.
Step 1: Export your ChatGPT data
- Open your profile in the lower-left corner of ChatGPT, then open the settings.
- Select “Data Controls,” followed by the data export option under “Export Data.”
- During the later review, pay particular attention to conversations about health, work, finances, and personal relationships rather than searching only for obvious names.
Step 2: Limit additional activity data
- Enable Computer History only if resuming tasks justifies the added visibility into your workflow.
- Exclude apps and websites where you handle sensitive personal or professional information.
- Delete individual timeline entries when they no longer offer value for future tasks.
If you are an advanced user, you can separate context according to the task. A limited activity history may be helpful for a work project, while health questions or private conversations do not need to occupy the same data space. You can also review the settings for history, data export, and model training separately. The sources note that user data may be used for training depending on the selected settings.
Important questions remain unanswered for Switzerland. The reports list Germany and several other countries for the advertising launch but do not state whether ads or Computer History will be available in Switzerland. The direct Swiss connection is Proton’s role as the provider of AI Paper Trail. The sources provide no details about supported languages or specific Swiss data protection conditions.
ChatGPT becomes more convenient when it has additional context and speaks more naturally, while advertising introduces another commercial layer. The available controls are useful, but they do not replace the decision about which information belongs in a chat or an activity history in the first place. The unresolved risk lies less in one “hmm” or one ad than in the subtle combination of familiarity, data collection, and commercial interests.
Sources
- ChatGPT zeigt bald auch in Deutschland Werbung – doch es gibt einen Ausweg – Medium unknown, 2026-08-17
- ChatGPT’s Computer History tracks your clicks and keystrokes – Medium unknown, 2026-08-16
- Was weiss ChatGPT über mich? Dieses Tool wertet deine Chats aus – Medium unknown, 2026-08-17
- ChatGPT macht jetzt “Hmm” und “okay” – warum diese Füllwörter kein Zufall sind – Medium unknown, 2026-08-16
- Stanford-Studie belegt: ChatGPT verlagert produktive Bildschirmzeit zu TikTok – Medium unknown, 2026-08-14
