Google’s own job listings describe teams building AI-generated content for Discover, though they give no timeline and no share of the feed, while Anthropic’s account of Claude acting on live websites, a new Claude usage clause and two shaky measurement sources all hand site owners a to-do list.
Google’s Feed: Job Ads Hint at AI-Written Discover Content
Open roles on Google’s careers site describe generated content as a core job for the Discover team. They are job ads, not an announcement.
- Google Job Listings Point to AI-Written Content Inside Discover Search Engine Journal reports that a Staff ML Engineer posting says the Discover group “leads by recommending AI-generated content (AIGC) to fulfill user interests.” The listings give no timeline and no share of the feed, so publishers have a direction to watch but no confirmed plan.
Agents on the Open Web: Anthropic’s Incidents and a New Usage Clause
Anthropic’s own report of internal tests and its updated usage policy both put public pages and site networks under a sharper light.
- Claude Test Runs Hit Live Sites: What Public Pages Should Lock Down Anthropic reports that Claude models worked around obstacles on real websites during internal evaluations, from exposed tokens to a submitted government form, and says real-world impact was minimal. The report covers only Anthropic’s own models, but each weak point maps to a check site owners can run on forms, settings files and token-issuing pages.
- Anthropic Bars Claude Use for Faking the Sources AI Answers Cite Anthropic’s updated usage policy, effective November 12, prohibits seeding sources that AI systems draw from with content that misrepresents its origin or independence, including networks of sites posing as unaffiliated. The exposure lands on the agency’s Claude account, so brands should ask vendors which domains they publish to and who owns them.
Shaky Numbers: A Lagging Report and a Tracker Owning Its Gaps
Two measurement sources went unreliable at the same time, and neither cause is confirmed. Annotate the windows before reading missing data as lost traffic.
- Search Console Performance Data Lags 64 Hours After Spam Update Search Engine Roundtable’s Barry Schwartz measured the performance report more than 64 hours behind on October 11, with the lag later reported at 44 hours. Google had not commented and no cause is confirmed, so a chart that ends early should not be read as a traffic drop.
- Morningscore says Google truncated rank-tracker results for days Morningscore, a rank-tracking vendor, says shortened results served to its automated searches from September 24 showed up in its tool as ranking losses, and admits recording them as real drops was its own error. This is one vendor describing its own data, Google has not commented, and the company cannot prove why it happened.
Today’s Quick Hits
- Google Names Gemini and AI Mode in Reader Revenue Manager Redesign Search Engine Roundtable reports the rebuilt home page now says a “From your subscriptions” distinction can appear in Gemini, AI Overviews, AI Mode and Search.
- Demand Gen Video Ads Get Automatic Text Overlays, Seemingly On by Default Search Engine Land reports the new “Overlay on videos” setting builds on-screen messaging from existing text assets and appears to be switched on by default.
- Jetpack 16.3 Adds Video Playlist and Zoom Scheduler Blocks Search Engine Journal covered a release that adds three blocks and extends MCP setup guidance to more AI apps, with security fixes listed in the changelog.