In this week’s digital news, Microsoft Advertising has reported that ad delivery was impacted for advertisers last week. Google Ads API v24.1 adds deeper reporting, expanded AI campaign testing, and new security updates for advertisers.
Google’s new AI Search guide says AEO and GEO are still SEO and names tactics site owners can ignore, including llms.txt, chunking, and special schema. Schema isn’t dead, but its pitch as an AI citation shortcut is weaker after Google’s FAQ removal and new Ahrefs data.
Microsoft Advertising confirms Ad delivery disruption
Microsoft Advertising has reported that ad delivery was impacted for advertisers between May 14, 2026, 9:20 PM UTC and May 14, 10:30 PM UTC. During this window, many experienced drops in impressions, clicks, and spend due to a platform-related issue.
Read more here.
Google expands Ads API testing tools in v24.1 release
Google released version 24.1 of the Google Ads API, introducing deeper reporting segmentation, expanded experiment support, and new security features as advertisers continue adapting to increasingly automated campaign environments. The update also prepares developers for Google’s upcoming data retention policy changes, which take effect next year.
Read more here.
Google’s official guidance on AEO and GEO
The impact of FAQ removal on schema value
It works until it doesn’t: AI content strategies that backfire
AI content at scale often drives fast SEO gains, then sharp declines once Google catches on. Analysis of 220+ sites shows a clear pattern. Over half lost at least 30% of peak traffic, many losing far more. The issue is not AI itself but how it is used. Templated, repetitive content built for rankings gets demoted. Risky formats include comparison pages, listicles, glossaries, FAQs, and programmatic pages. Sustainable performance comes from original, expert-led content with real value, not volume.
Read more here.
Inside ChatGPT Search: how web.run and fan-out queries shape AI visibility
New research reverse-engineering ChatGPT’s internal browsing system found that a March model switch reduced cited domains per response by over 20%, concentrating visibility on fewer high-authority sources. The newer model distributes queries across 10 or more “fan-out queries” per response, each targeting specific trusted domains, while product searches now get individual retrieval commands per product rather than bundled calls. The study distinguishes between parametric visibility, built through training data including press coverage and Wikipedia presence, and dynamic visibility from real-time search, noting that a brand unknown to the model’s training data won’t surface even when search is enabled.
Read more here.
If you need more information on any of these stories, or you’d like support on any of your digital marketing campaigns. Get in touch with our team of experts by dropping us an email to [email protected]
