Pages prepared for machines
According to Futurism, Time publishes plain-text copies of its pages alongside the HTML ones readers see, because AI systems process the plain versions more easily. The magazine places ads in those copies that gather a brand's information as questions and answers, and marks them as sponsored content. Ally Bank and the Project Management Institute are the named clients so far. Time reports that on most days bot traffic to the site exceeds human traffic.[1]
The artifact is easy to name: a second version of a magazine page that no human reader opens. The label sits on that version. Futurism's report does not describe the route by which the marker travels into an answer a chat assistant produces. Mobian cofounder Jonah Goodhart states the advertising expectation plainly: shifting a brand's standing inside ChatGPT means reaching the whole system at once.[1]
How much does the label change a reader?
The Decoder reports research by Sydney Sears and Deena Skolnick Weisberg, published in Judgment and Decision Making, in which more than 2,500 people took part across three experiments. In the first, 1,682 participants read six stories of about 1,000 words each — three from literary magazines, three generated with ChatGPT 4.0 — and half were told the true source while half were told the wrong one. On a scale running from minus 3 to plus 3, the AI stories scored 1.54 for quality against 0.97 for the human ones, and 1.42 for immersion against 1.00. In the second and third experiments, 905 participants read the two kinds side by side and tried to pick the source; their hit rate did not rise above chance.[2]
Read together, the two reports point to one constraint: knowing the source does change how a reader judges a text, and it does so only where the reader meets the label. In the Time programme the marker sits on a page no human opens, while the text a reader meets is an answer an assistant produced. Another explanation stays open: assistants may already carry the sponsorship marker through into their answers. Neither source examines an assistant's output, so a marker lost along the way remains a possibility the sources leave open rather than an observed outcome.[1], [2]
Which measurement would close the gap?
What would close the gap is concrete: a measurement that collects the answers an assistant produces from these Time pages and counts how often the sponsorship marker appears in them, with its sample and method stated. The Sears and Weisberg study shows that disclosing the source changes a reader's judgement; the matching measurement, of whether that same disclosure survives the machine-mediated path, has yet to be published.[1], [2]
The test can be kept narrow. Put a question to an assistant with one of the plain-text Time pages as its source, look at whether the sponsorship marker appears in the answer that comes back, and repeat that over a sample whose size is stated. Until such a measurement is published, whether a reader knows the origin of the text in front of them stays an open question for the Ally Bank and Project Management Institute ads. The Sears and Weisberg finding is what makes the question matter: a reader's judgement really does move once the origin is known.[1], [2]