Google Adds Top Stories Carousel Inside AI Overviews

One of the biggest stories of the week: fraudulent DMCA takedown requests are being systematically used as a negative SEO weapon, and Google is complying with them even when the claims are fabricated. According to Search Engine Roundtable's Barry Schwartz, legitimate content — owned and actively hosted by the original publisher — is being removed from Google Search because bad actors are filing false copyright complaints. The core problem is structural: Google processes takedown requests before disputes can be resolved, meaning a page can disappear from search results before the rightful owner even knows an attack is underway.

Search Engine Journal's Roger Montti went further, arguing the problem is systemic and will worsen without fundamental changes to how Google handles DMCA abuse. A companion piece by Matt Southern laid out the mechanics: once a takedown is filed, the page is removed first and restored only if a successful counter-notice is filed — a process that takes weeks and requires legal knowledge most small publishers don't have. Search Engine Journal warned SEOs to monitor Lumen Database (where DMCA requests are logged publicly) and set up Google Search Console alerts for manual actions as an early-warning system.

The story was prominent enough to be the lead item in Search Engine Roundtable's weekly video recap, with Schwartz noting that "no one seems to be doing anything" about the abuse. For SEOs managing content-heavy sites, the immediate action is monitoring Lumen Database entries for their domain and ensuring counter-notice processes are understood before an attack occurs.

Google rolled out several notable interface changes to AI Overviews and AI Mode this week. Most visibly, Google began displaying a Top Stories carousel inside AI Overviews — a feature pre-announced last month that is now reaching real users, per Search Engine Roundtable. This is a significant development for news publishers who had feared being cut out of AI-powered results entirely.

On the AI Mode front, Google made another round of improvements to recipe results. Google's Robby Stein officially announced a new visual treatment with prominent links at the top of responses, including creator names, recipe ratings, and ingredient counts. According to Search Engine Roundtable, this is the second such improvement to recipe visibility in AI Mode, suggesting food publishers have genuine traction with Google's product team.

A technical anomaly also emerged: Google AI Overviews were found to be displaying raw markdown file contents directly in snippets. John Mueller described the behavior as "unexpected" and clarified that markdown files receive no special indexing treatment — they are processed the same as any other content format. SEOs exposing markdown files on their sites should note this is a rendering quirk, not a signal to switch content formats.

A new study on Google AI Overviews, covered by Search Engine Journal, delivers a finding that complicates the dominant narrative around AI-driven click loss. The research found no measurable difference in bounce rates, time on site, or return-to-search rates between clicks that occurred with AI Overviews present versus without them. In other words, the clicks being lost to AI summaries were not the shallow, low-intent visits that some had argued were expendable — they were comparable in quality to clicks that still come through.

This directly challenges the argument that AI Overviews are only absorbing "low-quality" traffic and that publishers who lose clicks are not really losing valuable users. The data adds a new dimension to the ongoing debate about AI Overviews' business impact and strengthens the case for publishers who argue the feature is materially damaging their traffic.

Google's Head of Search Liz Reid made a series of public statements this week about the future of publishers in AI-powered search. In an interview on the AI Inside podcast with Jason Howell and Jeff Jarvis, Reid said: "We really do want great content to shine and continue to connect people to it." A related Search Engine Journal post attributed a more specific directive to Google: that publishers seeking visibility in AI search must focus on creating content users genuinely want to read — not content engineered primarily for rankings.

Reid also argued that personalization and preferred source settings could help small publishers get found in AI search, though Search Engine Journal noted she offered no supporting data to back up the claim. The statements come amid widespread concern in the publisher community about declining referral traffic from Google, and the gap between Reid's reassurances and the traffic data many publishers are experiencing was not lost on industry observers.

Building on the Similarweb finding covered in our June 22nd issue — that brands recommended by AI assistants received 2.5x more site visits, primarily via branded search — a fuller version of the study titled *The Downstream Impact of AI Visibility* has now been shared via SparkToro's blog and Amanda Natividad's newsletter, with expanded detail across three consumer industries including finance. The new distribution adds Amanda Natividad's framing of a critical attribution problem: much of this AI-driven traffic does not surface as identifiable AI referrals — it appears later as direct traffic or branded search, making standard attribution models blind to the full impact. This has significant implications for how marketers should measure ROI from AI visibility efforts: the signal exists, but it is hidden in brand metrics rather than channel reports.

Search Engine Journal published a detailed analysis by a practitioner who reverse-engineered ChatGPT's source selection by reading actual network traffic rather than inferring behavior from outputs. The findings challenge generic GEO (Generative Engine Optimization) advice: what matters most is crawlable facts, third-party validation, and whether a query triggers a live search — not superficial content optimization tactics.

The methodology itself is notable. By examining the actual HTTP requests ChatGPT makes during a session, the author identified when ChatGPT performs real-time web lookups versus when it draws from training data, and which signals correlate with citation. The conclusion: sites that earn third-party mentions and publish crawlable, verifiable facts are structurally better positioned than sites following generic AI-optimization checklists.

Two pieces this week converged on a similar strategic conclusion: the most defensible path to AI citation is authenticity and originality. Kevin Indig's Growth Memo (sent to 27,558 subscribers) argued that business-generated proprietary data is the single most reliable lever for page originality — and that it outperforms commissioned surveys or assembled datasets because it cannot be replicated. The old PR-research play of buying a loosely product-tied survey, Indig wrote, no longer compounds the way it once did.

Search Engine Journal's Slobodan Manic took a complementary angle, arguing that buying Reddit upvotes and citations to game AI results is the modern equivalent of link farming — and will likely be filtered out the same way Google's Penguin update neutralized bought links. The only durable strategy in an agentic web, the piece argued, is being genuinely real: real products, real community presence, real third-party mentions. Together, these pieces sketch out a consistent framework — AI citations reward what was always rewarded in traditional SEO: genuine authority and original evidence.

Search Engine Journal issued a critical warning this week: Cloudflare's new AI crawler management system can inadvertently block Googlebot if site owners choose to block "training" crawlers without understanding how Cloudflare categorizes bots. Cloudflare's system classifies crawlers as Search, Agent, or Training — and Google's crawlers, including Googlebot, can fall under the Training category in certain configurations. Default settings are scheduled to take effect on September 15, meaning site owners who have not actively reviewed their Cloudflare crawler rules could unknowingly exclude Google from crawling their sites.

This is an urgent heads-up for SEOs managing sites on Cloudflare. The recommended action is to audit current crawler management settings in the Cloudflare dashboard before the September 15 defaults activate, and explicitly whitelist Googlebot if training crawler blocking is enabled.

Google has made a significant infrastructure change to how it handles AMP in Search: users are now sent directly to publisher-hosted AMP pages rather than Google's cached AMP viewer. AMP content continues to rank normally in search results, and the change does not affect eligibility for AMP-associated features. Google stated the shift was made to "simplify and reduce maintenance efforts for publishers who are creating AMP content."

Both Search Engine Journal and Search Engine Roundtable confirmed the change. For publishers who have built around the AMP ecosystem, the practical effect is that traffic now lands on their own domain rather than Google's `amp.google.com` viewer — which is generally a positive for first-party data collection and user experience, though it removes the performance guarantee that Google's CDN delivery previously provided.

Two significant Google Search Console developments occurred this week in quick succession. First, Google's newly launched Generative AI performance reports — which give SEOs data on how their content appears in AI-generated search experiences — were immediately hit by a data logging bug affecting Discover and Generative AI in Discover data, starting June 24. Google documented the issue, but the fact that a brand-new reporting surface broke almost instantly means any data collected in that window should be treated with skepticism.

Separately, Google fixed the Page Indexing report after it had been broken and delayed for over three weeks — a significant outage for a fundamental SEO monitoring tool. Search Engine Roundtable's Barry Schwartz confirmed the fix went live in real time as he was writing about it. For SEOs who had been flying blind on indexing status during that period, the report is now reliable again. The coincidence of one report being fixed while a newer one immediately broke is a notable contrast in Google Search Console's reliability this week.

The line between AI-generated content and paid advertising continued to blur this week across both Google and OpenAI. Google is testing AI-generated summaries placed directly beneath ad descriptions in sponsored search results, with a disclaimer reading: *"Google AI responses are generated independently and can make mistakes, so double-check responses."* Both Search Engine Land and Search Engine Roundtable noted the test, which raises questions about ad performance and user trust when AI-hallucinated content sits beneath an advertiser's official copy.

Simultaneously, Google is integrating Gemini-generated content into AdSense Ad Intents units — meaning user interactions with ad anchors or chips now surface both ads and Gemini AI responses in the same dialog. Meanwhile, OpenAI job listings indicate ChatGPT is preparing to launch image, video, native, and conversational ad formats, signaling a major monetization push that could reshape the paid search competitive landscape. ChatGPT has already quietly updated its ad labeling, swapping "Sponsored" for the simpler "Ad" label while repositioning it within the ad unit.

Following last week's full coverage of the June 2026 spam update's rollout and its extension to AI-generated answers, this week's digests confirmed the update's final standing: a two-day rollout that Barry Schwartz (Search Engine Roundtable) and Ahrefs Digest #279 both flagged as unusually fast and substantially impactful. No new enforcement signals have emerged beyond what was reported last week. SEOs who saw traffic changes around the rollout window should audit against Google's spam policies, particularly around scaled content, spammy structured data, and manipulative link practices.

Search Engine Journal drew a critical distinction between the two major phone AI agents shipped this month: Google's Chrome auto-browse agent actively visits and interacts with websites, executing JavaScript and triggering server requests, while Apple's Siri AI only reads content from a snapshot without live interaction. For website owners, this difference is not cosmetic — Google's agent behaves more like a real user (and will count in analytics and server logs), while Apple's approach is more akin to a passive screen reader.

The technical implications include ensuring sites render correctly for automated browsing agents, that key content isn't locked behind interactions requiring human gestures, and that server infrastructure can handle an additional class of bot traffic.

Two technical findings this week highlight the growing importance of AI agent readiness. A report from Siteline tested a Claude AI agent against top B2B product websites and found it frequently failed to locate pricing information on-site, often falling back on third-party sources like G2 or Capterra instead. The implication for B2B marketers is that pricing pages may need to be restructured — more prominently accessible, clearly labeled, and ideally marked up with structured data — to be machine-readable by AI shopping agents.

Separately, Lighthouse 13.3 introduced an Agentic Browsing audit category that specifically checks `llms.txt` files. The audit fails sites whose `llms.txt` lacks properly formatted markdown links. According to Search Engine Journal, the fix is simple — properly formatted markdown links — but the introduction of a formal Lighthouse audit signals that `llms.txt` compliance is becoming part of mainstream technical SEO hygiene.

Following last week's tribute to Bruce Clay — the Father of SEO whose passing was covered extensively in our June 22nd issue — SEO Notebook's Steve Toth added a personal remembrance this week, noting he had "been sitting with the news for a couple of days before writing anything," and compiled Clay's top 25 most-linked articles as a legacy resource for the community.

New this week: Fabrice Canel announced his retirement from Microsoft after nearly 30 years, closing out a career defined by his leadership of Bing's crawling and indexing infrastructure and his role as the primary champion of the IndexNow protocol. Both Search Engine Journal and Search Engine Roundtable covered the retirement, noting Canel's outsized influence on how web crawling and real-time indexing evolved during his tenure.

Multiple pieces this week examined the structural decline of publisher traffic in the AI search era. Search Engine Journal's analysis found that habitual readership — users who regularly return to publisher sites directly — is collapsing as AI tools intercept demand that previously drove repeat visits. A companion piece found that younger audience share is declining faster than surface metrics suggest, with publishers bearing a disproportionately larger share of the audience loss compared to social platforms.

Google's own AI Mode data, covered by Search Engine Journal, shows a measurable behavioral shift: conversational, intent-driven queries now outpace traditional keyword searches, meaning content strategies built around head and mid-tail keywords are increasingly misaligned with how people actually search. The cumulative picture these stories paint is one of structural disruption for the content publishing business model — not a temporary dip but an accelerating erosion of the traffic patterns publishers have relied on for over a decade.

Local SEOs had two separate issues to track this week. First, a wave of complaints flooded the Google Business Profile support forums as Google reviews began disappearing from business listings, with no official explanation from Google. Search Engine Roundtable flagged the influx of complaints and noted Google had not responded publicly — leaving businesses uncertain whether a bug, a policy change, or a review removal sweep was responsible.

Separately, Google was spotted testing a redesigned local pack layout that moves the map to the top of results (rather than the right side) and repositions action buttons. UI tests in local results are common, but map placement has historically influenced how much attention users pay to the listing cluster. Local SEOs should monitor both situations: the review disappearance may require proactive client communication, and the layout test could signal upcoming SERP structure changes.

Google's AI governance paper, reported by Search Engine Journal, publicly defends the use of public web content for AI training as fair use under copyright law. The paper points to opt-out controls and paid licensing deals for specialized content as the appropriate avenues for publishers who object, rather than treating training use as requiring opt-in consent.

The policy position is significant for several reasons: it signals Google's legal strategy ahead of potential publisher lawsuits or regulatory action, it frames the opt-out burden on publishers rather than requiring Google to seek permission, and it comes at a moment when publishers are simultaneously losing referral traffic to AI Overviews and having their content used to train the systems responsible. The combination of the DMCA abuse story and this fair-use defense paints a complex picture of publishers' dwindling leverage over how their content is used in the AI era.