AI chatbots spill conversations to ad industry – ChatGPT, Claude, and Gemini also affected

TVS News AI, editorially supervised, presents AI chatbots spill conversations to ad industry – ChatGPT, Claude, and Gemini also affected.

You type a private question into the AI chat one evening – and with six of nine examined chatbots, parts of that conversation land with ad services, according to a study. Golem.de is reporting on it. Among the affected services are ChatGPT, Claude, and Gemini.

Anyone who uses AI assistants for tax questions, health, or personal problems usually trusts that those chats stay between them and the provider. Exactly that picture is now cracking.

Six of nine – and the three big names are among them

The study Golem.de cites examined nine chatbots. Six of them shared chat data with ad services, according to the measurement. Three held back. From the available excerpt, ChatGPT, Claude, and Gemini appear in the group of sharing services – three of the most popular AI assistants anywhere.

Why this is more than a niche data-protection issue

Ad services inside apps and on websites exist to target users precisely. Once chat content reaches those networks, profiles can be built from it: which topics you type into the AI, in what tone, at what hour. Combined with other data, that paints a much sharper picture than many people expect when casually trying out a chatbot.

What you can actually do

Until the full study report is available, three simple steps help:

• Don't type anything sensitive into just any chat: Consciously separate which content you entrust to which provider – not every chatbot needs every sentence from you.
• Check your privacy settings: Many providers offer toggles to disable training on your chats or data sharing – often buried deep in account or privacy settings.
• Block browser trackers: An ad blocker or the tracking-protection feature in modern browsers filters out at least the ad services on the web that might be reading your AI requests in the background.

What remains open

Neither the study itself nor the exact experimental setup is visible from the excerpt. The names of the three chatbots that – according to the report – shared no data are also missing. What already counts: don't treat your own chat behavior naively, and work through your settings consistently.

The detailed report is available at Golem.de.