Confirmed
Google says DMA-mandated changes caused the largest search-quality reduction in its history
Google senior vice president Nick Fox told Reuters that European Digital Markets Act compliance produced 'the largest reduction in quality of service' at Google Search in its 29-year history. The changes give price-comparison sites like Booking.com and Expedia more prominent placement while stripping real-time pricing from Google's own travel carousels. Google says an earlier round of DMA changes already cut free, direct booking traffic to European businesses by 30%.
In plain termsDMA is the EU competition law that already forced Google to show more links to rival services like Booking.com in travel results, and Google is now saying that same law is making its regular search results worse across Europe, not just travel searches.
Why it mattersFor any EU-facing e-commerce or travel brand, Google is now on record saying its own results got worse, which is grounds to audit whether DMA-driven vertical-search placement changed your organic visibility.
Our takeI flagged in 'Google's billions-of-clicks claim' that Nick Fox was one of three Google voices to repeat a stable, unfalsifiable traffic number over eleven months with no published baseline attached. He is now on record with a concrete, admitted decline instead, and that shift from vague reassurance to precise blame is worth watching as more DMA compliance data ships in November.
Confirmed
Apple confirms Applebot-Extended site rules do not affect Apple Search ranking
Apple's official Applebot support documentation, updated September 4, now states plainly that 'site rules for Applebot-Extended are not considered in ranking for Search.' Applebot-Extended controls whether a site's content can train Apple's generative AI and Apple Intelligence features; the main Applebot crawler that feeds search ranking runs independently of that opt-out. The clarification lands weeks ahead of iOS 27's expected public release.
In plain termsApplebot-Extended is a separate crawler from the one that actually decides where your pages show up in Apple's search results, so telling Applebot-Extended to stay out only opts you out of AI training, not out of being found.
Why it mattersA publisher can now block Apple's AI-training crawler without any ranking risk in Apple Search, which removes a guessing game that Google's own Googlebot/Google-Extended split already settled for the bigger search engine.
Our takeI wrote in 'Blocking the AI crawlers did not stop the citations' that publishers keep aiming crawl controls at the wrong bot in the Google pairing, blocking the training crawler while leaving the retrieval crawler that actually creates citations wide open. Apple just codified the same split by name, so the same discipline applies: check which of Applebot and Applebot-Extended is doing the work you're trying to stop before you disallow either one.
Confirmed
Linux kernel maintainer documents crawler load exceeding all legitimate git.kernel.org traffic combined
Kernel.org maintainer Konstantin Ryabitsev reports that abusive scrapers now consume more CPU cycles at git.kernel.org than every legitimate access path combined, including git clones. Across five geographically distributed servers, 14 CPU cores run continuously just to render Git commits as HTML pages for crawlers that ignore rate limits and robots.txt. Ryabitsev's post documents the operational cost of treating a public code archive as an unlimited scraping target.
In plain termsA CPU core is a unit of a server's processing power, so dedicating 14 of them full time just to generate pages for bots means over a dozen computers' worth of capacity is being spent purely on being scraped, not on serving the humans the site exists for.
Why it mattersThis is the infrastructure-cost mirror of the AI-training-data debate: crawler load severe enough to dedicate 14 CPU cores full time is the kind of number that justifies aggressive bot-blocking even on sites with nothing to hide.
Our takeI measured a version of this same imbalance in 'Google Wants a 304' testing conditional requests on retail homepages, where servers that fail to send a 304 burn crawl capacity resending pages that never changed. Ryabitsev's number is a more extreme case of the identical failure mode, a server spending finite capacity on requests that add no value, except here the requester is abuse traffic with no rate limit to respect.
Observed
Google tests a narrower AI Overviews citations panel that could reduce publisher visibility
A design change spotted in Google Search shows the citations panel next to AI Overviews rendered noticeably narrower than the current version, cutting how much source information is visible before a user has to click to expand it. Search Engine Roundtable's Barry Schwartz confirmed he could reproduce the change and posted side-by-side comparisons. Google has not confirmed the test or a rollout timeline.
In plain termsPicture a research paper's works-cited section still being present but printed in a much narrower column, so fewer entries fit on screen at once; most readers never scroll far enough to see who was actually cited.
Why it mattersLess visible citation real estate ahead of any click is one more lever, alongside placement and expansion behavior, that can quietly suppress publisher referral traffic from AI Overviews without a documented policy change to point to.
Our takeA narrower panel is a design lever, not a policy change, so it can ship, get measured internally, and reverse with no announcement either way. The pattern worth tracking is whether citation panel width correlates with click-through to source sites over the coming weeks, since that would turn an easy-to-miss UI test into hard evidence of intent.
Observed
Google tests paginated follow-up responses with a skip button inside AI Mode
Screenshots shared by researcher Sachin Patel show AI Mode generating a new query within the same conversation after a user responds, then presenting the follow-up as a stacked, paginated card the user can page through or bypass with a skip button. The test appeared on both mobile and desktop. Google has not confirmed the feature or said whether it will expand beyond the test group.
In plain termsPaginating a follow-up means splitting what would be one long AI-generated answer into a set of swipeable cards instead, and the skip button lets a user jump past that follow-up entirely if they don't want it.
Why it mattersPagination and a skip option inside a single AI Mode conversation change how many turns a user takes before reaching a brand or product mention, which shifts where in a multi-turn conversation practitioners should expect their content to actually surface.
Our takeAI Mode has already grown ad placements gradually from an initial small test to default behavior, so a UI mechanic this specific is worth tracking rather than dismissing as cosmetic. If pagination becomes the default shape of a multi-turn answer, the follow-up card, not the first response, becomes the surface worth optimizing for.
Observed
Google's EU booking search widgets show prices customized per user by 'the booking partner'
A disclaimer spotted in Google's EU booking search widgets now states that displayed prices are 'customised by the booking partner based on factors such as your device type.' The widgets sit inside the same DMA-mandated vertical-search surfaces Google's own quality complaint elsewhere in this rollout describes. The full list of factors behind the customization, beyond device type, has not been disclosed.
In plain termsThe widget is telling users upfront that the price they see was picked for them individually, based on things like what device they're using, rather than being one fixed price everyone sees for that hotel or flight.
Why it mattersPersonalized pricing inside a search-results widget means two users can see genuinely different offers for the same query, which complicates any competitive price-monitoring or rank-tracking exercise run against those surfaces.
Our takeThis is the same DMA-mandated vertical-search surface Google says is degrading its results, so a disclosed personalization factor here is a small window into how those partner integrations actually work under the hood. Device type is the factor Google chose to disclose, which says nothing about whether location, browsing history or booking urgency also feed the price shown.