Google Says DMA Compliance Is Its Worst Ever Search Quality Cut

In the week's single biggest story, Google officially told Reuters that its Digital Markets Act compliance rollout in Europe represents "the largest reduction in quality of service at the world's most popular internet search engine in its 29-year search history." According to Search Engine Roundtable and Search Engine Land, which both covered the statement, this is an extraordinary public admission from a company that rarely speaks this candidly about self-inflicted search degradation.

As reported by Search Engine Journal, the practical changes include a complete redesign of Search results across the European Economic Area — reworking aggregator units and introducing new supplier units for hotel, flight, transport, and product queries to comply with DMA requirements. Google simultaneously published new help documentation detailing regional differences in the Search experience by country, covering availability of aggregator units, supplier units, and carousels — essentially a map of how deeply the SERP has been restructured across Europe. A booking widget disclaimer was also surfaced, noting that displayed prices are now "customized based on factors such as your device type" — raising transparency questions about what other personalization signals influence shown prices.

Following three weeks of coverage — the initial August 18–21 rollout (August 17th issue), post-official-end-date volatility and AI content targeting findings (August 24th issue), and Google VP Rajan Patel's public endorsement of the spam team (August 31st issue) — Kevin Indig's Growth Intelligence Brief #24 now provides the first comprehensive visibility data from the full update window. Over the 21 days ending August 31, Amazon hit its lowest SEO visibility since tracking began in November 2022, dropping 16.9%. U.S. government sites were hit hardest: the IRS dropped 59.6% and the SBA fell 52.8% in organic visibility. The IRS figure is consistent with the August 24th issue's report of Googlebot being inadvertently blocked by aggressive bot-filtering, but the SBA drop appears to be a distinct development. These are some of the sharpest drops recorded in a single update window and represent the first named-winner/loser data from this update cycle.

Google's John Mueller issued one of his strongest-ever warnings about programmatic SEO this week. According to Search Engine Roundtable, Mueller stated that pSEO "often leads to a site that's either spam, borderline spam, or low quality," and that these tactics can cause Google's systems to have "possibly lost faith in your site providing good value to users based on the old pages."

Search Engine Journal expanded on this, reporting that Mueller clarified the damage is not simply undone by removing low-value pages. Recovery requires demonstrated value over time — Google's systems may continue to judge a site by its historical content quality long after the offending pages are gone. This is a direct signal that algorithmic trust rebuilds slowly, and that bulk content operations should be audited proactively rather than reactively.

Search Engine Journal reported a striking real-world case study this week: a UK news publisher suffered a catastrophic 98% drop in news search visibility after migrating from a .co.uk domain to a .com. Critically, the new .com domain did not inherit the lost rankings — the trending-query visibility that had accumulated on the old domain simply evaporated. This illustrates a specific and severe risk for news publishers undertaking domain migrations, where Google News indexing signals appear to be non-transferable.

According to Search Engine Roundtable, new data shows Google AI Overviews now answer nearly 100% of all People Also Ask questions — a staggering leap from just 12% fourteen months ago. This near-total absorption of PAA results by AI-generated answers represents a fundamental shift in how Google's most common SERP feature operates, and has direct implications for organic click-through rates from PAA boxes. Content that previously earned traffic through PAA placement is now at risk of being answered entirely within the SERP.

Two publisher data points this week painted a nuanced picture of AI Overviews' real-world traffic impact. A University of Washington study, reported by Search Engine Journal, estimates that Google's AI Overviews have reduced search referrals to Wikipedia by approximately 5% — a figure Google disputes. In contrast, Le Monde told Search Engine Journal that its audience figures have not declined since AI Overviews launched in France, attributing resilience partly to app traffic growth and subscriber gains.

The divergence suggests AI Overview impact may vary significantly by site type, audience loyalty, and traffic composition. The Wikipedia figure, while disputed, is notable as one of the first peer-research estimates on AI Overview traffic suppression from a known-scale dataset.

Building on the August 17th issue's report that generative UI in AI Overviews now generates functional tool equivalents directly in the SERP, Google's zero-click encroachment continued this week on two fronts. Search Engine Roundtable reports Google is testing a noticeably thinner citations panel on the right side of AI Overviews, which may reduce the visibility of cited sources and further suppress click-throughs to linked pages. Separately, Google AI Overviews now include an interactive character counter widget directly in search results — another simple utility absorbed into the SERP itself, consistent with the pattern of tool pages losing traffic to AI-generated equivalents documented over prior weeks.

Following Aleyda Solis's 3-layer diagnostic framework for measuring AI search presence (August 24th issue) and Lumar's authority-as-binary-filter finding (August 31st issue), this week's measurement conversation shifted toward correcting flawed data inputs and narrowing tracking scope. Search Engine Journal published a critique arguing that publishers are blocking AI crawlers based on unreliable referral traffic figures — the widely cited data omits an unknown share of referrals from its denominator, making it an unsound basis for crawler access decisions. A separate SEJ piece argued that Perplexity's declining market share warrants reducing its weight in AI visibility tracking frameworks, recommending measurement focus shift to ChatGPT, Gemini, and Claude. Both pieces complement the August 24th and August 31st frameworks by addressing data quality and platform prioritization — the prerequisite steps before applying the measurement layers previously described.

Google Search Console hit a rough patch on multiple fronts this week. First, a brief but alarming indexing reporting bug temporarily showed pages across many sites as "Crawled - currently not indexed." Google's John Mueller confirmed the glitch within hours, calling it a "short blip" that resolved within an hour — reassuring SEOs that no actual indexing had been lost.

A more persistent problem: the Google Search Console Links report has not been updated for a full month, with the last refresh occurring around August 8, 2026, according to Search Engine Roundtable. No official explanation has been offered. Separately, the page indexing report is also missing a significant block of data from June 2026, with Barry Schwartz noting a broader pattern of indexing complaints over recent months.

In a notable admission, Google's John Mueller stated this week that "the old 'position 1–10' is hard to map" for site owners in the context of how impressions are counted for Generative AI results in Google Search Console. According to Search Engine Roundtable, Mueller was responding to questions about the Search Console impression documentation for AI-generated results, but his comment signals a broader truth: classic rank-tracking metrics increasingly fail to reflect real-world search visibility in a SERP where AI Overviews, featured snippets, and AI Mode dominate above traditional blue links.

The first major expert ranking factors survey in some time was published this week by Cyrus Shepard and Dawn Anderson at Zyppy, covered by Search Engine Roundtable. The survey finds that content relevance, backlinks, and content quality remain the top-ranked factors according to polled SEO experts in 2026. While the results reinforce long-standing consensus rather than reveal surprises, the survey provides a useful current-year benchmark for practitioners to weigh their priorities against industry peers.

Search Engine Journal reported that Google DeepMind researchers have developed and tested Autoregressive Ranking (ARR), a unified AI model designed to potentially replace current retrieval and ranking methods in Google Search. Unlike the current pipeline — which runs separate retrieval, ranking, and re-ranking steps — ARR operates as a single unified model. This is a long-horizon research development, but it signals that Google is actively exploring a fundamental rewrite of its ranking architecture driven by large language model design principles.

Two developments this week point to Google's continued experimentation with AI search interface and access. Search Engine Roundtable reports Google is testing paginated follow-up responses in AI Mode, introducing multi-step question-and-answer flows with a skip button — a significant UI shift that could change how users navigate complex queries.

Meanwhile, Search Engine Journal flagged that Google's developer documentation reveals a partner-only Web Search Service API path providing access to full-web search results. The partner-only designation raises questions about tiered data access and what competitive advantages approved partners may hold over third-party SEO and search tool developers.

Two significant e-commerce search developments this week. First, Search Engine Journal reported that ChatGPT's shopping results are increasingly drawing from feed-integrated product sources, with store visibility positioning also shifting alongside this dependency. For e-commerce marketers, product feed optimization is emerging as a distinct AI search ranking lever — separate from traditional SEO.

Second, Google Merchant Center's AI Performance Insights report gained three new metrics: AI Search Intent, AI Search Terms, and AI Attributes, expanding a feature that launched in July. These additions give Google Shopping advertisers deeper visibility into how AI-driven search is influencing product discovery and purchase intent.

New data reported by Search Engine Roundtable suggests Google Shopping Ads are experiencing a "reverse crocodile effect" since AI Overviews launched — meaning impressions and clicks are moving in the opposite direction compared to organic results, which have seen impressions decouple from clicks (the classic "crocodile" divergence). This is a notable counter-trend for e-commerce advertisers.

In related Shopping news, Google is now displaying social proof metrics such as "200+ sold last week" as ad extensions in Shopping Ads, and separately testing a countdown attribute showing how many days remain in a sale — both aimed at driving urgency and CTR in sponsored results.

Local SEOs faced two separate Google Business Profile problems this week. First, Search Engine Roundtable reported that GBP insights data is missing for the entire month of September 2026 — with the gap persisting more than 8 days into October and no official Google acknowledgment.

More urgently, a widescale scam was documented where fraudsters are exploiting the Google Maps contributor program to upload AI-generated photos containing fake phone numbers, redirecting customers away from legitimate listings. The scam reportedly spans dozens if not hundreds of GBP listings, and according to Search Engine Roundtable, reporting them to Google has had little visible effect. Businesses and agencies managing GBP listings should audit contributed photos immediately.

Following the August 31st issue's report that Google was testing a fifth photo-based GBP verification method, two further GBP updates landed this week. Google is now rolling out post view counts globally for Google Business Profile posts published in the past 18 months, combining view data from both Search and Maps — a concrete performance signal for GBP content that was previously unavailable, and a fuller rollout of what the August 24th issue identified as 'Posts Insights' returning under the name 'people viewed.' Separately, Google is sending email notifications to GBP owners prompting them to confirm or deny user-suggested edits to their listings — a data-accuracy mechanism that also requires business owners to stay vigilant about unsolicited listing changes.

Following the August 17th issue's coverage of the llms.txt v2 specification update adding formal link relations for AI agent navigation, new empirical data from Common Crawl's analysis of 584,107 llms.txt files in the wild reveals a significant implementation quality gap. Most files were generated from templates, many contained no links at all, and a meaningful share of site owners are incorrectly attempting to use the format to enforce crawler access rules — a function llms.txt was never designed to serve. The research suggests the standard's real-world adoption is largely superficial relative to its intended purpose. Separately, Apple updated Applebot documentation to explicitly clarify that rules set for Applebot-Extended do not influence ranking in Apple Search — mirroring Google's equivalent distinction and important for SEOs who may have assumed AI training opt-outs carry ranking weight.

Several pieces this week advanced the practical playbook for Generative Engine Optimization. A Search Engine Journal article argued that the biggest AI search risk for brands is not a lack of content but conflicting information spread across the web — publishing more content won't fix what AI models get wrong if multiple inconsistent versions of brand facts exist online.

A study analyzing 120,000 AI mentions across five major AI models (from Uberall, covered by SEJ) identified four key signals driving multi-location business visibility in AI search results including ChatGPT and Gemini. Separately, a Freshpet/Intero Digital case study shared a framework for auditing AI presence, restructuring content for citation, and tying GEO work to measurable business outcomes — with the core insight that page-one Google rankings do not guarantee AI citation.

Search Engine Journal reported that Automattic's board of directors has placed WordPress co-founder Matt Mullenweg on paid leave, effectively removing him as CEO. Mullenweg has publicly claimed a conspiracy against him. The upheaval at the company behind WordPress — which powers a significant share of the web's SEO infrastructure, including countless plugins and themes — carries potential long-term implications for WordPress development velocity, plugin ecosystems, and the broader CMS landscape that SEO professionals depend on.

A Harvard study that scored 940 occupations on the moral acceptability of AI automation found that search marketing ranked low at 2.31 out of 7 — meaning the public has little moral objection to AI replacing search marketers, as reported by Search Engine Journal. The finding positions SEO among the occupations the public views as most automatable without ethical concern, adding a societal dimension to the ongoing industry conversation about AI's impact on SEO jobs.