SEO Recap covering June 19, 2026: Bing ships AI citation share as llms.txt doubts grow

Built by Stephanie Chung·4 min read·5 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)

When ChatGPT shipped model 5 in August 2025, almost every AI citation tracker showed a drop. The cause was fewer citation links in ChatGPT's HTML, not worse optimization.
Third-party tools only see so much. One project site showed 1 to 3 Copilot citations in Ahrefs but more than 36,000 by Copilot's own count.
The author recommends tracking on 2 axes, volatility (how stable your presence is over time) and average response (sentiment and inclusion across related prompts), rather than chasing rank tracking precision.
The advice for stakeholders:
- trade hockey-stick vanity metrics for risk mitigation
- track steadier brand sentiment
- correct how the model misrepresents you

Bing Webmaster Tools rolled out 4 preview features:
- Citation Share reports your AI citation percentage against competitors, using Bing and Copilot data only
- Intents
- Topics
- Compare
Google's John Mueller said llms.txt can't help LLMs tell sites apart because it's self-reported. Ahrefs data across 137,000 domains found 97% of llms.txt files got zero requests, and citation bots accounted for just 1% of fetches.
Google Cloud published the Open Knowledge Format (v0.1), and a coalition of Google, Microsoft, GitHub, and Hugging Face released the Agentic Resource Discovery draft spec (v0.9).
The UK CMA ordered Google to rank UK organic results, including AI Overviews, on objective criteria, and to give advance notice before significant ranking changes.

Google's first AI Mode data (May 20, 2026) cited 1 billion monthly active users, queries 3 times the length of traditional search, and planning queries growing 80% faster than AI Mode queries overall.
Adobe's Q2 2026 report found AI-referred retail traffic converts 42% above non-AI traffic, a reversal of the usual pattern, because those visitors arrive pre-qualified after researching inside the AI.
The author recommends auditing your top 10 AI-referred landing pages, asking whether a visitor can finish their task in 30 seconds, and moving the surfaces that complete a task (booking, pricing, CTA) to the top.
GA4 tracks referrals from chat.openai.com and gemini.google.com, while Search Console folds AI Mode clicks into overall metrics with no separate filter.

15% of daily Google searches are completely new. Complex queries trigger a "fan-out" that expands into parallel sub-searches.
Google said forcing paragraph "chunking" for AI is pointless. Content has to read well for humans, and there is no algorithmic reward for formal HTML validation.
Subscription linking through Reader Revenue Manager produced a 34% engagement boost in internal case studies and surfaces "From your subscription" labels in the SERPs.
The quality guidance stressed unique, specific, and authentic first-hand content over commodity or programmatic text, what Google calls Scaled Content Abuse.
GSC's AI Reporting (Beta) is rolling out to isolate AI Overviews, AI Mode, and Discover impressions and clicks, with an AI Settings panel to include or exclude a site.

Google's Scalable Cluster Termination System (S-CTS) goes after coordinated generative-AI spam by spotting clusters of accounts that reuse the same semantic narrative templates, rather than judging each piece of content on its own.
It uses Sentence-BERT (SBERT) to find semantically similar AI-generated text through mathematical "text embeddings," an approach that rarely comes up in SEO circles.
It leans on LoRA and Automatic Prompt Optimization to adapt quickly to new generative models without the cost of fully retraining large models like Gemini 2.0 Flash.
The research focused on video spam, but the text-based detection it describes could plausibly extend to fingerprinting web content spam.

Published on June 20, 2026·Last updated on July 25, 2026