StephEO

SEO Recap covering July 28, 2026: Indexed Claude chats and AI opt-out publisher risks

Built by Stephanie Chung·9 min read·12 stories

Daily summary of what matters in SEO, GEO, AEO, and AI search generated with Claude Code (beware of hallucinations)

Generative Engine Optimization (GEO) & Answer Engine Optimization (AEO)

How creator content shapes AI answers and where citations come from
Source: Search Engine Land

  • Tinuiti's Q1 2026 AI Citation Trends Report found roughly 82% of AI citations point to earned media rather than a brand's own site, with creators a fast-rising share.
  • Social citation share varies sharply by category. Social platforms drove about 13% of AI citations on apparel prompts but only 3% on over-the-counter health, and Perplexity's social citation share fell from 31% to 13% in one quarter as it pulled back from Reddit.
  • OtterlyAI's YouTube Citation Study 2026 found long-form video accounts for 94% of AI citations and 40.83% of cited videos had fewer than 1,000 views, indicating structure and clear formatting matter more than popularity.
  • Google introduced platform properties in Search Console in July, letting brands track how Instagram, TikTok, X and YouTube content performs in Google Search, which the author suggests aligning influencer and SEO teams around.

Why data integrity is the new technical SEO, from crawling to trust
Source: Search Engine Journal

  • The author notes Google has dropped support for 9 ItemTypes from its rich results gallery in two years, most recently FAQ in June 2026, but argues the deprecation is a display feature, not schema itself, which serves as a comprehension layer.
  • The piece outlines 5 layers of data integrity SEOs should manage:
    • Entities stabilized with @ids tied to Wikidata, GS1, ISNI or ORCID
    • Relationships via sameAs and RDF
    • Format such as JSON-LD, markdown and llms.txt
    • Actions declared to agents like BuyAction, WebMCP, ACP and UCP
    • Perception including third-party sentiment and grounding
  • There is no consortium creating a single agentic standard the way Google, Bing and Yahoo built schema.org and XML sitemaps, leaving SEOs without consensus on what to implement.
  • The suggested priority order is stabilizing @ids and sameAs first, then testing interpretation with NLWeb, auditing ecommerce product feeds, attempting WebMCP, and treating markdown serving and OKF/ARD as later steps.

Yelp licenses reviews and local business data to ChatGPT
Source: MarTech

  • Yelp is licensing its reviews, ratings, photos and local business data to OpenAI, so ChatGPT responses to local queries can include Yelp content with Yelp branding and links, according to Axios.
  • The companies plan to add Yelp's Request a Quote feature, letting users contact local service providers directly from ChatGPT.
  • The deal mirrors OpenAI's licensing agreements with publishers and Reddit, and does not prevent Yelp from licensing to other AI companies. Financial terms were not disclosed.

AI in Search / AI Overviews

AI opt-out may cost publishers a Top Stories spot
Source: Search Engine Journal

  • NewzDash tracking found Google placing the Top Stories carousel inside AI Overviews in 15.5% of tracked US news searches and 17.46% in the UK, with entertainment highest. The two placements never appeared together.
  • Google began honoring its Search Console AI opt-out setting on June 17, initially for a subset of UK site owners, covering AI Overviews, AI Mode and Top Stories placements.
  • NewzDash CEO John Shehata reads it as high confidence that opting out would drop a publisher from a Top Stories carousel inside an AI Overview, though Google has not confirmed and NewzDash plans to test it.
  • Shehata warns publishers not to confuse Google-Extended, which governs Gemini training and grounding, with a search opt-out. Blocking it alone does not remove a site from AI Overviews, AI Mode or either Top Stories placement.

What Google's latest earnings call signals for search and AI
Source: Search Engine Journal

  • Gemini 3.5 Flash-Lite is now the primary model routing queries for AI Overviews, which the author says will likely make AI Overviews better at answering directly and could reduce traditional search traffic.
  • Sundar Pichai emphasized agentic search, with planned in-search connections to apps like Instacart and Canva so agents can complete tasks rather than just return information.
  • Google revenue rose 24% year over year, but the quarter went cash flow negative after $6 billion in physical assets spend (60% servers, 40% data centers). Network ad revenue fell 1% while Cloud rose 87%.
  • The author recommends making websites actionable for agents and watching commerce protocols like the Universal Commerce Protocol, seen so far on Wayfair and some Walmart products.

Technical SEO

Google indexed Claude chats because Anthropic did not block them from search engines
Source: Search Engine Land

  • Wired reported that private Claude chats shared via public snapshot URLs (claude.ai/share) were indexed across Google, Bing and other engines, including sensitive political and health discussions.
  • The core technical error is blocking a page in robots.txt while also using noindex. Crawlers cannot see the noindex tag if robots.txt prevents them from accessing the page, so it can still appear in results.
  • Glenn Gabe criticized the coverage for missing this, and Google's own documentation warns in bold that noindex only works if the page is not blocked by robots.txt.
  • Google's Ned Adriance said neither Google nor any search engine controls what pages are made public, and that Google respects site owners' crawl and index directives.

Google's Mueller on URLs injected into HTML by CMS platforms
Source: Search Engine Journal

  • John Mueller told a Reddit user that platform-generated internal links, like a Squarespace category URL blocked by robots.txt, have no impact on search or SEO and can be ignored.
  • The odd Squarespace URL is the platform's internal database identifier for a category, which lets users rename categories freely while the underlying ID stays fixed. WordPress does the same with term_id and post_id.
  • Adding ?format=json-pretty to any Squarespace URL outputs the page's JSON code and reveals the underlying database IDs.
  • The takeaway for practitioners is to understand how a CMS works before making changes for SEO, since crawls with Screaming Frog can surface artifacts that are normal and safe to leave alone.

Google SEO test debunks the 5-second rendering window myth
Source: Search Engine Journal

  • Dave Smart of Tame the Bots tested the widely held belief that Google's Web Rendering Service stops rendering after 5 seconds and found no such limit.
  • His test page made two server-side API calls each delayed 3 to 6 seconds, totaling up to 12 seconds, and the WRS still waited for the delayed content and included it in the rendered DOM.
  • The WRS uses a virtual clock it can pause during network requests, so more real-world time can pass than a JavaScript timer reports.
  • Martin Splitt's original comments described the median 5-second wait in the rendering queue before rendering begins, not a rendering time limit, aligning with Smart's findings.

Google Search Console page indexing report delayed for weeks
Source: Search Engine Roundtable

  • Since June 11, the Search Console page indexing report has been stuck, showing data grouped by weeks rather than days despite Google fixing it several times.
  • Brodie Clark documented only 3 unique data points since June 13, grouped into periods of roughly 14, 10 and 18 days.
  • The delay does not affect indexing itself but makes debugging harder, since results appear more dramatic when changes only surface after long gaps.

Organic Search & Algorithm Updates

Why SEO should be measured against blended customer acquisition cost
Source: Search Engine Land

  • The author argues channel-level CAC misrepresents SEO because organic discovery often gets stripped from attribution when conversions land on email, paid search, or owned channels.
  • SparkToro's analysis of Similarweb clickstream data found 68.01% of US Google searches ended without a click in the first 4 months of 2026, up from 60.45% in 2024, making SEO's influence harder to measure.
  • The piece lists ways SEO leans out blended CAC:
    • Creating nonpaid entry points into the funnel
    • Capturing demand paid would otherwise buy
    • Supporting paid search and paid social conversion
    • Growing branded and direct demand over time
    • Feeding owned channels like email
  • The recommendation is to reframe the conversation around how much more a business must spend on other channels for every dollar not spent on SEO, and to report channel CAC alongside influenced pipeline and blended efficiency.

Heavily AI-flagged pages still rank across Google's top 10
Source: Search Engine Journal

  • Using Ahrefs' own AI detector on about 150,000 pages from 100,000 searches in June 2026, 54.7% of top-three results scored under 20% AI-generated content, and pages under 50% AI text took 82.2% of top-three spots.
  • Pages scored at a full 100% AI were rare but real at 5.3% of top-three results, and average AI score rose only slightly from 27.1% at position 1 to 30.9% at position 10.
  • Ahrefs found 49.28% of the least AI-heavy pages in Google's index versus 40.35% of pages scoring 80% or higher, a difference the authors call meaningful but far from disqualifying.
  • The 2025 version across 600,000 pages put the AI-score-to-ranking correlation at effectively zero, so the two studies used different samples and depths and the clearest change is the wording, not proof Google penalizes AI text.

Google lost its scraping case against SerpApi and what it means for the open web
Source: Search Engine Journal

  • On July 20, Chief Judge Yvonne Gonzalez Rogers dismissed Google's DMCA claims against SerpApi, which scrapes and resells Google results, ruling that Google's SearchGuard anti-bot system protects ad revenue rather than a copyrighted work.
  • The claim was thrown out with prejudice where no copyrighted content was involved, with Google given 21 days to return with a narrow version about Knowledge Panel images.
  • The author frames it as the founding question of the agentic web, whether an automated visitor is allowed onto public content, and links it to the Amazon v. Perplexity case testing the CFAA.
  • The practical takeaway is that an anti-bot wall guarding revenue rather than copyright does not buy DMCA protection, so sites should decide access per crawler using host or CDN controls rather than waiting on courts.
Published on
geoai searchtechnical seoorganic searchdigest