Time sells invisible ads that manipulate AI chatbots

In a move that redefines digital advertising, Time Magazine has begun serving ads invisible to humans but perfectly visible to the artificial intelligence crawlers that feed chatbots like ChatGPT, Claude, and Perplexity. Detected by independent developer Vincent Schmalbach and confirmed by The Register, this case marks a milestone in the evolution of the web: ads are no longer aimed at people, but at the machines that influence human decisions. This practice, which some already call 'advertising for AI,' raises unprecedented ethical and regulatory questions, and could transform the digital economy as we know it.

What exactly happened?

Schmalbach discovered that Time serves markdown versions of some of its pages to certain AI bots, with embedded sponsored content. This content, labeled as sponsored, includes frequently asked questions about the online bank Ally Bank, with claims such as being 'the only bank built for today's life' or being in 'a category of its own.' The ads are designed to be regurgitated by chatbots when users ask about banking services. According to Schmalbach's analysis, the FAQs include 'brand facts' that are easily reproducible by language models, such as the number of fee-free ATMs associated with the bank, and answer common questions like 'Is Ally good for everyday banking?' or 'Can you deposit cash at Ally Bank?'.

The bots receiving this version include ClaudeBot (from Anthropic), OAI-SearchBot (from OpenAI), and PerplexityBot. In contrast, Google's indexing bot receives the same HTML that humans see. The publication confirmed the findings using a crawler simulator, and verified that the ads appear on evergreen articles, such as lists of best inventions, but not on breaking news. Additionally, it was observed that GPTBot and ChatGPT-User, used by OpenAI for training and retrieval by users, return HTTP 406 errors in Schmalbach's tests, suggesting that Time selectively blocks certain crawlers while serving sponsored content to others.

Why is this important?

This case is the first documented example of an AI-targeted advertising strategy. With more than 50% of US consumers using AI to search the web, influencing chatbot responses has become a key advertising goal. If a chatbot recommends Ally Bank, the user is likely to consider it an objective suggestion, when in reality it is a paid ad. The practice resembles traditional native advertising, but with a crucial difference: the audience is machines, and the deception occurs in the conversational interface, where users tend to trust AI-generated responses more.

Moreover, AI crawler traffic already exceeds human visits on many websites, according to data from Cloudflare and other analytics companies. This makes it economically attractive to optimize content for these bots, since a single mention in a chatbot can reach millions of users indirectly. This practice could quickly spread to other media, transforming the nature of the web: from a space for humans to an ecosystem where machines are the primary audience. In fact, similar cases have already been detected on other sites, though less sophisticated, such as including sponsored links in 404 error pages aimed at bots.

Consequences for the digital ecosystem

• For users: chatbot responses could be biased by commercial interests without the user knowing, eroding trust in these tools. A recent Pew Research Center study indicates that 62% of users believe chatbots are impartial, making them vulnerable to this type of manipulation.
• For brands: companies could pay to influence AI recommendations, creating a new form of native advertising that is difficult to detect. This opens a potential multi-billion dollar market, similar to SEO but for language models.
• For publishers: media outlets could generate additional revenue by selling ad space to bots, but at the cost of their credibility. Trust in media is already at historic lows, and this practice could accelerate its decline if perceived as deceptive.
• For regulators: the FTC and other agencies could consider this practice deceptive advertising, since consumers cannot distinguish between organic and sponsored content. In 2023, the FTC already issued guidelines on social media advertising, and could extend them to AI. Additionally, the EU is developing the AI Act, which could include provisions on transparency in AI-generated content.

What should readers know?

It is crucial for users to be aware that chatbots are not neutral sources. Responses may be influenced by invisible ads. To protect themselves, it is recommended to cross-check information with direct sources and be skeptical of specific brand recommendations. Chatbot developers also have a role to play: they should implement mechanisms to detect and label sponsored content in training data, as some search engines already do with ads.

Time has not responded to The Register's questions, but pointed to a Digiday article that could shed light on its strategy. Meanwhile, this case sets a precedent that could change digital advertising forever. The question is no longer whether AI will influence our decisions, but who will pay to control that influence.

This is the first clear glimpse of what the web is beginning to look like when the primary audience is AI models.

In a broader context, this practice resembles the early days of SEO, when webmasters hid white text on white backgrounds to deceive search engines. The difference is that now machines are the end audience, and the manipulation can be much more subtle and difficult to detect. The advertising industry should self-regulate before regulators intervene, as already happened with cookies and GDPR.