Claude Sonnet 5 launch, Marmite rebrand & more
This week's highlights across Search, Social, PR and AI.
Search
1. Google is testing AI-generated summaries inside Search ads
Google has been spotted running a small experiment that places AI-generated summaries directly beneath standard Search ad descriptions, with a disclaimer noting that responses are generated independently and can make mistakes.
If this rolls out broadly, it would mean Google’s AI, rather than the advertiser, decides what context gets surfaced alongside a paid listing. The implications for brand control and messaging accuracy are significant, and it’s another sign that generative AI is slowly reshaping every surface of the Search experience, not just organic results.
2. Google Search Console can now track your Instagram, TikTok, X and YouTube posts
Google is introducing platform properties in Search Console, a new property type that lets creators see how their social media and video posts perform in Google Search and Discover.
Once an account is connected, you get the familiar Performance report showing clicks, impressions and the queries driving traffic, plus an Insights report and milestone-based Achievements. Crucially, you don’t need a verified website to use it, making it available to creators who live entirely on social platforms.
This is a meaningful acknowledgement from Google that social content is now a legitimate part of the search ecosystem. For brands and agencies, it also opens up a cleaner way to report on social content’s contribution to search visibility, something that’s historically been very difficult to measure.
Social
3. Facebook is testing post view counts
Facebook is trialling a post view count display for creators, following the same shift Meta has already made on Instagram and Threads. The metric would initially be visible only to the creator themselves, not the wider public.
Meta has been pushing views as a more meaningful signal than likes or follower counts for some time now, reflecting how algorithm-driven feeds have changed what actually drives reach. Given that Facebook’s daily active user numbers are reportedly falling, more granular data could help creators understand whether their content is landing.
4. One in four Americans now discover products through influencers
New YouGov data shows that 27% of US adults discover new products through social media influencers or bloggers, rising to 41% among Gen Z. That puts influencers level with search engines as a discovery channel for younger consumers, and well ahead of TV and radio advertising, which only 19% of Gen Z cited.
Brands with strong influencer partnerships are seeing measurable uplifts in consideration, with CeraVe, e.l.f. and Dunkin’ all showing higher consideration scores among Gen Z who discover products via creators. The generational gap in how people find things is becoming harder for brands to ignore.
5. Young Brits are concerned about social media misinformation, but using it more than ever anyway
IPA TouchPoints 2026 data shows that 79% of UK 16-34s are concerned about the amount of misinformation on social media, yet they still spend over two and a half hours a day on it. Nearly nine in ten young Brits use social media in a typical week, and it now occupies the top five most-used media properties for younger audiences.
The data underlines what marketers and platforms alike already suspect: awareness of the problem isn’t translating into changed behaviour. Social media’s hold on attention remains as powerful as ever, even as trust continues to erode.
PR
6. Marmite temporarily becomes 'WeMite' for England's World Cup run
Marmite has rebranded to WeMite for the duration of the FIFA World Cup, releasing limited-edition jars scaled by optimism: a small jar for fans expecting a quarter-final exit, a medium one for semi-final believers, and the largest for those confident England can go all the way.
Created by adam&eve\TBWA and available via Heathrow Express, the campaign is designed for the tens of thousands of fans making the trip to the US.
Tying the product into the ritual of travelling support gives it real utility alongside the cultural moment, and the sizing mechanic gives fans a chance to declare their level of belief before they board. Very clever.
7. F1 and Lego bring 22 drivable minicars to the British Grand Prix drivers’ parade
Following the success of last year’s Miami drivers’ parade, Lego and Formula 1 returned at Silverstone with 22 individual drivable minicars, one for each driver, each built from over 28,000 Lego bricks and capable of reaching 25km/h. The cars took 6,400 hours to build and were decorated in each team’s 2026 livery.
The activation has already generated a ton of organic coverage, reinforcing why experiential stunts that create playful moments like this remain some of sport’s most effective brand marketing.
AI
8. Anthropic launches Claude Sonnet 5
Anthropic has released Claude Sonnet 5, its most capable Sonnet model yet, with performance described as close to Opus 4.8 but at a significantly lower price. The model is built for agentic tasks, with notable improvements in reasoning, tool use, coding and multi-step autonomous work.
It launches as the default model for Free and Pro plans, and is available across Claude Code, Claude Cowork, Team and Enterprise. Introductory pricing of $2 per million input tokens runs through the end of August.
9. Fable 5 is back, and Anthropic is proposing an industry jailbreak framework
Claude Fable 5 returned on 1st July following the lifting of US export controls that had suspended access since mid-June. Alongside the redeployment, Anthropic published a detailed account of the timeline and announced a new industry framework for assessing the severity of AI jailbreaks, developed in partnership with Amazon, Microsoft, Google and other partners.
The publication offers an unusually transparent look at how frontier labs are navigating the tension between deploying highly capable models and managing government concerns about dual-use risks. If adopted across the industry, the proposed framework could become a standard for how AI developers communicate risk to both regulators and the public.
10. Google opens up Nano Banana 2 Lite and Gemini Omni Flash to developers
Google has released two new generative media models: Nano Banana 2 Lite, a fast and cost-efficient image generation model producing outputs in around four seconds at $0.034 per image, and Gemini Omni Flash, a model for high-quality video generation and conversational video editing. Both models support SynthID watermarking, and developers can chain them together to take an image concept through to an animated video clip in one workflow.
The launch positions Google as the platform of choice for developers building multimedia AI experiences at scale, and Nano Banana 2 Lite is already rolling out inside consumer surfaces including AI Mode in Search, the Gemini app and Google Photos.
11. Meta quietly launches Pocket, an app that turns prompts into playable mini games
Meta has launched Pocket, a new creative app that lets users generate small interactive experiences, called “gizmos”, from text descriptions. The gizmos can respond to touch and phone tilt, use sound and camera, and be shared in a social discovery feed where other users can play with what you’ve made.
It’s particularly interesting that Pocket treats prompting as a way to make interactive objects, not just content. If audiences start sharing lightweight playable things the way they share Reels templates or AI images, the creative unit for brands and campaigns changes considerably.
12. Meta's internal Watermelon model is reportedly matching GPT-5.5 benchmarks
Meta’s superintelligence chief Alexandr Wang reportedly told employees at an internal town hall that the company’s unreleased model, codenamed Watermelon, has caught up with OpenAI’s GPT-5.5 on key AI benchmarks. Wang reportedly described the model as still in training and using an order of magnitude more compute than Meta’s previous model, Muse Spark.
Internal benchmark claims without published methodology or independent replication are common in the frontier model race, and carry real risk of optimistic framing. That said, the compute trajectory Wang described is the more concrete indicator that Meta is scaling aggressively.
That’s a wrap for this week! See you next Wednesday for another round-up of the best and biggest stories from across the Search-Verse™.