In this week’s digital news, Google Ads has expanded New Customer Acquisition with a new feature called Prospect Mode. OpenAI launches conversion-optimised ChatGPT ads June 5. US programmatic CPMs surge 34% year-over-year. Pichai admits Google Search has become too opinionated.
Google launched the May 2026 core update, redesigned Search around AI at I/O, released its first AI Mode usage data, and sent mixed signals on llms.txt. Your brand name in Google no longer guarantees a fast click. New data reveals what AI Overviews are doing to navigational search behaviour.
New: Google Ads Prospects Mode
Google Ads has introduced Prospect Mode, an expansion of its New Customer Acquisition feature. Unlike the standard mode, which targets anyone who hasn’t previously made a purchase, Prospect Mode goes further by focusing exclusively on brand-unaware users, people who have never searched for, visited, or engaged with your business in any way.
When enabled via the new “Only bid for new prospects (BETA)” setting in the Bidding section, the feature automatically excludes users who have purchased from you, searched for your brand, visited your website or app, or interacted with your content on Google or YouTube. Essentially, it targets entirely cold audiences, helping advertisers reach genuinely new prospects rather than those already familiar with their brand.
ChatGPT launches conversion ads June 5. US programmatic CPMs up 34%
OpenAI is rolling out conversion-optimised advertising in ChatGPT from 5 June, but only for accounts with active conversion tracking already in place via its JavaScript Pixel or Conversions API. The measurement setup closely mirrors Meta’s infrastructure. Cost-per-action ads, where brands pay only on clicks, sign-ups, or purchases, were quietly activated for select advertisers in late May. The platform has moved quickly: six major product updates in 26 days, with annualised ad revenue reportedly exceeding $100 million within six weeks. Despite OpenAI positioning ChatGPT as a test channel, its growing performance toolkit tells a different story.
Google launches Core Update amid I/O AI search overhaul
Study of 846,000 sessions reveals changing search behaviour
Can we take the doom out of scrolling?
Researchers at Kellogg tested an alternative social media algorithm on 2,000 Bluesky users around the 2024 US election and found it reduced toxic and politically extreme content without moderating any specific posts. The algorithm worked by simply limiting the outsized influence of a small number of highly engaged accounts, making feeds more representative of the broader conversation on the platform. Users on engagement-based feeds saw 57.3% more political posts after the election compared to those on chronological feeds, and they significantly overestimated partisan animosity in their networks. Participants using the alternative algorithm reported enjoying the platform more, challenging the industry assumption that engagement-driven content is necessary to retain users.
Read more here.
Claude Opus 4.8: “a modest but tangible improvement”
Claude Opus 4.8’s most prominent improvement is its ability to be honest. It is more likely than previous models to flag uncertainties about its work and less likely to make unsupported claims. The model has a new feature that allows mid-conversation system messages. This means users can append updated instructions later in long-running conversations without restating the full system prompt.
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]
