TIME.com sends Googlebot the same HTML page a browser gets. Three other crawlers get something else: a stripped markdown file, a fraction of the size, carrying sponsored FAQ content that no human visitor on that same page ever sees. Google’s spam policy bans manipulating AI answers inside Google Search. As written, it doesn’t reach a page that Googlebot and human browsers already see identically.
What Vincent Schmalbach Measured
Vincent Schmalbach, a Germany-based freelance software developer, fetched a single TIME article, “The Morning Light Habit Sleep Experts Swear By,” repeatedly, changing only the User-Agent request header each time, and published the results on August 5. Same URL, same moment, three different outcomes.
| User-Agent | Response | Size |
|---|---|---|
| Chrome, Safari, Googlebot | 200 OK, text/html |
303,235 bytes |
| ClaudeBot, PerplexityBot, OAI-SearchBot | 200 OK, text/markdown |
13,409 bytes |
| GPTBot, ChatGPT-User | 406, blocked entirely | no body |
303,235 bytes against 13,409 bytes: the markdown response is roughly 22 times smaller, and identical across all three bots that receive it. Schmalbach’s own words: “Same URL, same second, one twenty-third of the size, and a completely different format.” His conclusion goes further: “There is now a layer of TIME.com written entirely for machines, and the humans who read the site have no idea it exists.”
What Is Inside the Markdown Layer?
The markdown itself is clean: no HTML, no layout, just text. Inside it sits a full Ally Bank FAQ block, questions including “Who is Ally Bank?” and “Can you deposit cash at Ally Bank?”, carrying tracking links tagged campaign="ally-2026-q3". Load the identical URL as a browser and none of it is there: no Ally Bank content, no mention of Mobian, TIME’s ad-tech partner.
What Does TIME Serve to AI Crawlers?
TIME.com returns a different response depending on the User-Agent header. Human browsers and Googlebot receive the full HTML page; ClaudeBot, PerplexityBot, and OAI-SearchBot receive a markdown-only version roughly 22 times smaller, containing sponsored FAQ blocks that don’t appear on the human-facing page at all.
The Business Case, Dated Six Days Earlier
Digiday reported the programme on July 30, 2026, six days before Schmalbach published his test, and dated TIME’s conversion of “all its webpages” to the previous month. The measurement is new; the programme is not. Digiday describes the ad unit as an FAQ carrying a brand’s information and messaging, labeled as sponsored content and converted to PDF for human review before a client signs off. Ally Bank and the Project Management Institute are named as among the first advertisers to buy an ad meant to be read by a machine rather than a person. TIME’s ad-tech partner on the programme is Mobian, described as an AI-powered ad tech platform.
TIME COO Mark Howard: “This is a growing traffic source, and therefore a growing source of inventory that we believe has huge value if we can continue to develop this product.” Digiday reports that TIME sees more bot traffic than human traffic on most days. TIME charges a premium for the placement, and Howard declined to share pricing. Howard also said: “This is an obvious extension of list sponsorship or franchise sponsorship.”
Where Google’s Rule Actually Reaches
Google’s Search spam policies define spam as, among other things, “attempting to manipulate generative AI responses in Google Search.” That clause explicitly covers AI Overviews and AI Mode. Google’s cloaking definition turns on “presenting different content to users and search engines,” and the policies state they apply to “all web search results, including those from Google’s own properties.” TIME serves Googlebot the same page it serves a user. What changes is which non-Google crawler sits on the other end of the request.
OpenAI’s own crawlers split down the middle on that same server. OAI-SearchBot gets the ad-bearing markdown. GPTBot and ChatGPT-User get a flat 406, blocked outright. No source explains why; the split is a fact of the response logs. What it shows a publisher is that the crawler answering a question and the crawler training a model are two separate commercial decisions, and a server can price and permission them independently.
None of this requires TIME’s permission, or Google’s, to check on your own server. A publisher can audit which AI crawlers your robots.txt actually admits.
