Search is splitting into two tracks simultaneously: AI Overviews are measurably shrinking the click-through value of ranking, while Google and OpenAI are both building new monetisation surfaces that reward being followed or embedded rather than simply found. This week’s data makes that shift impossible to ignore.
AI Overviews Slashed Publisher CTR by 23%
Ahrefs tracked 963 domains in Google Search Console across the 28 days before and 9 days after France received AI Overviews on July 22, and found the most exposed domains lost 23.1% of their click-through rate — the first hard number behind what had been largely anecdotal anxiety. The timing matters: this isn’t a projection or a lab test, it’s live Search Console data from a real rollout, giving SEOs a concrete benchmark to stress-test their own traffic models against.
Audit your highest-impression, lowest-click queries now — those are the exact pages AI Overviews are already answering for your audience.
OpenAI Puts Sponsored Actions Inside ChatGPT Chats
OpenAI launched Sponsored Agents with direct integrations into HubSpot and Shopify, letting brands place sponsored actions inside a live ChatGPT conversation rather than a static ad unit — putting OpenAI in direct competition with Google’s and Meta’s ad stacks for the first time. This shifts ad budget toward conversational commerce and gives ecommerce brands a genuinely new placement to test, one tied to intent signals that are far richer than a keyword match.
If you manage ecommerce ad spend, request early access to OpenAI’s advertiser programme before your competitors lock in the learning curve advantage.
Google Lets Shoppers Transfer Their Cart Straight to Checkout
Google is rolling out cart transfer and checkout testing inside Merchant Center’s UCP hub, closing the gap where a shopper discovers a product on Google but abandons before reaching a merchant’s own checkout — a friction point that has quietly bled conversions for years. The update also includes a YouTube ads beta and wider AI insights, signalling that Google is building Merchant Center into a full commerce operating layer rather than just a feed manager.
Enable cart transfer in Merchant Center as soon as it appears in your UCP dashboard — early adopters will have cleaner conversion data before the feature is widely benchmarked.
Reuters Proves Dynamic Paywalls Can Grow Subs and Ads Together
Reuters reports that its dynamic paywall — which adjusts the subscription threshold individually per visitor — is growing subscriptions while simultaneously strengthening its ads business, rather than trading one off against the other. That directly undercuts the long-held assumption that a harder paywall always costs ad reach, and gives publishers a live reference point instead of a theoretical model to argue over in strategy meetings.
If you run a content subscription model, test a metered or behaviour-triggered paywall before assuming a hard gate is your only path to reader revenue.
Google Search Profiles Now Open at 10,000 Followers
Google dropped Search profile eligibility to 10,000 followers for U.S. sites and creators, adding new account tools and updated article displays — turning Search itself into a following-based discovery surface at a far lower threshold than before. The timing is pointed: AI Overviews are actively shrinking the referral value of a ranked result while Google simultaneously builds a retention layer where being followed inside Search compensates for lost clicks.
Claim and optimise your Google Search profile today, because follower-based visibility is becoming the hedge against AI Overview cannibalisation.
A 200x Cheaper Model Just for Classify-and-Route Tasks
TypeSafe released Jev, a “System One” model that classifies, routes, and scores using reinforcement learning rather than generating text token by token — running more than 100 times faster and roughly 200 times cheaper than small frontier LLMs on those specific tasks. Any routing, moderation, or scoring step that currently burns a full LLM call inside an agent pipeline is now a candidate for replacement with a dedicated model at a fraction of the cost, without touching the parts of the pipeline that actually need language generation.
Map your agent pipeline and identify every classify-or-route step that doesn’t require text generation — each one is a direct cost-reduction target using models like Jev.
Read the full story →
Join the discussion →
More from Rafal Reyzer
For deeper dives on AI and marketing strategy, visit my YouTube channel →
Rafal Reyzer
Hey there, welcome to my blog! I'm a full-time entrepreneur building two companies, a digital marketer, and a content creator with 10+ years of experience. I started RafalReyzer.com to provide you with great tools and strategies you can use to become a proficient digital marketer and achieve freedom through online creativity. My site is a one-stop shop for digital marketers, and content enthusiasts who want to be independent, earn more money, and create beautiful things. Explore my journey here, and don't forget to get in touch if you need help with digital marketing.