Key points
- MIT Technology Review reports agents like Meta's Muse and OpenAI's Dots are logging record downloads and getting access to bank accounts and inboxes.
- Time is publishing agent-friendly article versions and selling ads aimed at agents; no one yet knows if those ads influence agent behavior enough to justify spend.
- A University of Toronto-led study found agents entirely ignore items tagged "sponsored" in search results, undercutting paid placement.
- Pew found people are about 40% less likely to click through from Google results when an AI summary appears.
Why it matters for marketers: If agents ignore sponsored tags and skip site visits, paid placement and click-based traffic lose value, so marketers need to test how agents find and judge their content.
MIT Technology Review argues that the agentic web, a version of the internet used heavily by autonomous software, is already taking shape. Tools such as Meta's Muse and OpenAI's Dots are logging record downloads, and Anthropic launched a research preview of Claude Duet in October. Unlike the tag-dependent Semantic Web Tim Berners-Lee proposed in 2001, large language model agents can read ordinary pages and, in some cases, click through them like a person.
Some publishers and retailers are already adapting. Time is serving agent-friendly versions of its articles and has started selling ads aimed at agents, delivered through the search crawlers that feed the databases agents query. Etsy and Expedia have partnered with major AI firms on tools that let agents work with their services. Amazon, meanwhile, blocked Muse after its release.
For marketers, the early signals are uncomfortable. A study led by Akshit Kumar at the University of Toronto found agents entirely ignore items carrying a "sponsored" tag, which raises doubts about selling favorable positions on results pages. Pew research found people are about 40% less likely to click through when a Google AI summary appears, and one analysis found some publishers' ad revenue fell about 40% between 2025 and 2026.
Design for agents is also unsettled. Wharton professor Ethan Mollick said newer models viewed his agent-directed book page as an attempt to manipulate them rather than recommending the book. Watch how agent-targeted ads perform, whether platforms block or allow agents, and test how your own content is summarized and recommended before committing budget.
Sources
This brief is our own summary. Read the full reporting at the source.
