SEO Recap covering July 13, 2026

6 min read·8 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)

ChatGPT sent users to outside websites in 5.2% of sessions versus 31.1% for Google searches, per Comscore clickstream data analyzed by Bocconi University researchers.
ChatGPT's outbound clicks favored reference, SaaS, academic, and developer sites; ad-supported sites made up 27.6 percentage points less of its referral traffic than Google's.
Broader ChatGPT Search access cut weekly traditional search queries by 9.4%, reaching 17% after 20 weeks.
Informational searches fell hardest, with academic research referrals down 32.8% and reference queries down 26.5%, while transactional and recreational searches barely moved.
ChatGPT's monthly referral rate rose from about 2.5% to nearly 6.5% during the measurement window, still below Google's per-query rate.

Across 19 markets, the US had the lowest AI-assisted search rate at 48%, versus 89% in India, Indonesia, and the UAE; only 28% of US searchers trust AI assistant information against 70% for search engines.
Search engines lead every task tested: 69% start a specific question there versus 16% for AI, and 62% versus 4% for product research.
Among AI searchers, 22% most often click through to supplied links and only 17% stop at the AI answer; that click-through climbs to 33% for daily AI users.
Clear source links (16%) and official-source framing (15%) move AI searchers most, but 49% of non-AI searchers say no trust signal would change their mind.
Greg Jarboe advises keeping classic fundamentals (on-page SEO, schema, crawlability) primary while structuring pages so click-throughs from AI answers land on more detailed, clearly sourced content.

Chris Green maps protocols like LLMs.txt, ARD, MCP, WebMCP, A2A, OKF, and UCP along action-versus-knowledge and agent-versus-publisher axes.
He frames the roles: ARD helps agents find capabilities, MCP and WebMCP expose or invoke tools, A2A coordinates agents, and UCP applies these to commerce journeys.
The recommended priority for most organizations is understanding the landscape, improving discoverability, then exposing capabilities, rather than chasing every acronym.
Green suggests ecommerce teams watch UCP for agentic checkout, WebMCP and ARD for customer actions, and OKF for large or complex site discovery.
His core point: standards win through ecosystem adoption, not technical superiority, so understanding each one's purpose matters more than betting on a winner.

Marvin says nothing has changed on eligibility: ads reach AI Overviews and AI Mode through Broad Match or keywordless targeting via AI Max, Performance Max, Shopping, and Dynamic Search Ads, paired with Smart Bidding.
She notes the relevance bar is higher in AI Search, with ads matched against both the user query and the content of the AI response, making text customization and Final URL Expansion central.
Google describes Qualified Future Conversions as a predictive metric estimating conversions up to 180 days after an ad interaction, positioned as a supplement to existing attribution, not a replacement.
Google says roughly 70% of standard campaign conversions land within a 30-day click and three-day engaged-view window, dropping to about 50% for Performance Max and 40% for Demand Gen.
AI Brief, which lets advertisers give messaging and matching guidance in natural language, is expected to roll out in English in the coming months.

Google updated its canonical help document to explicitly say to include a rel="canonical" link on the canonical page itself.
The two added lines make clear the same self-referential canonical should appear on the canonical page.
Barry Schwartz notes the advice isn't new: John Mueller said in 2011 that self-referential canonicals help keep things clean.

Shelley Walsh argues the keyword-first evergreen model collapsed after AI Overviews, citing Duane Forrester's point that content fully replaceable by a summary has no moat.
The Reuters Institute's 2026 trends report found publishers deprioritizing evergreen content by 32 percentage points in favor of original investigations.
Named writers including Kevin Indig, Duane Forrester, and Harry Clarkson-Bennett are moving to Substack, reflecting a shift where trust attaches to people rather than mastheads.
Walsh points to Aleyda Solis's method of building realistic prompt sets from sales calls, reviews, and communities, then running them across platforms to see where a brand appears or is missing.
The piece recommends building around owned distribution (newsletters, followers, direct audiences) and firsthand expertise that only one person or dataset could produce.

Barry Schwartz flags a possible unconfirmed Google update around July 11 (dubbed the "7-Eleven update"), with tracking tools showing a volatility spike but unusually quiet SEO community chatter.
Glenn Gabe reported that some sites hit by the January 2026 unconfirmed update, which hammered "commodity content" and self-serving listicles, are now recovering; Lily Ray called January the year's clearest pattern.
WebmasterWorld posters describe clicks trending up, Discover recovering for news sites, and mixed but improving revenue.
Multiple third-party trackers including DataForSEO, Mozcast, AccuRanker, and Semrush registered the July 11 blip.

Microsoft Bing's Krishna Madhavan said Bing does not issue one-off penalties because they aren't scalable, instead building algorithms to detect a spam type and act across all matching sites.
The comments followed Glenn Gabe and Lily Ray flagging a site surging in ChatGPT (via Bing rankings) with no Google visibility.
That flagged site, with roughly 90,000 URLs indexed, was completely deindexed from Bing shortly after.
Schwartz notes Bing has been seen emailing sites when they're no longer blocked, though that can be automated.

Published on July 14, 2026