Issue 15 — 14/07/2026

2026 AI Visibility Index: 126 Million U.S. AI Search Prompts Analysed

Semrush · 26/06/2026

Per a study from Semrush, 126 million U.S. AI search prompts issued between January and April 2026 were analysed to map how brands are mentioned, cited, and represented across AI-powered discovery environments — a 50,400× scale-up from the firm's September 2025 baseline of 2,500 prompts. The expanded dataset is among the largest published AI citation corpora to date. Practitioners can use the directional findings to benchmark brand citation share and identify which content surfaces AI engines are drawing from in their vertical.

AEO Periodic Table V4: Citation Factors Drawn from 1.13 Million Prompts

Goodie · 23/06/2026

Per a study from Goodie, 1.13 million prompts run across ChatGPT, Claude, Perplexity, Grok, Gemini, and Google AI Mode produced a fourth-edition framework mapping the factors that earn brand visibility in AI-generated answers. A central finding: AI models treat a brand as the aggregate of everything learned in training and retrieved at query time — including Wikipedia, Reddit, review sites, and forums — surfaces that fall outside the scope of most SEO teams. Practitioners should audit off-site entity presence alongside on-site content.

AIO Impact on Google CTR: 2026 Update — Full-Year Analysis Across 53 Brands

Seer Interactive · 24/04/2026

Per a study from Seer Interactive, a full-year analysis of 2025 spanning 53 brands, 5.47 million tracked queries, and 2.43 billion organic impressions found that the previously predicted continued CTR decline into 2026 has reversed directionally — the declining trend has turned the other way in early 2026. The study also found that AI Overviews now appear on nearly 65% of question-based searches. Practitioners relying on 2025-era CTR decay models should recalibrate their forecasts against Q1 2026 actuals before setting H2 traffic targets.

Google AI Mode Crosses One Billion Monthly Users; AI Overviews Appear on 13.7% of Queries

Perplexity AI Magazine / GoGoChimp · 05/07/2026

Google confirmed at I/O 2026 that AI Mode has crossed one billion monthly users (per GoGoChimp, citing the announcement). Separately, a 55,393-query study cited in Perplexity AI Magazine found AI Overviews appeared in 13.7% of all queries analysed, rising to 64.7% for question-form searches; nearly 30% of AI Overview-cited domains were absent from the co-displayed first results page. For practitioners, traditional rank tracking now systematically undercounts citation exposure — passage-level monitoring is required alongside position tracking.

AI Engines Maintain Distinct Citation Identities: Seven-Month, Seven-Engine Analysis

Conductor · 07/05/2026

Per a seven-month analysis from Conductor tracking 1,056 data points across ChatGPT, ChatGPT Search, Perplexity, Google AIO, Google AI Mode, Gemini, and Claude (September 2025–March 2026), every engine exhibits a persistent editorial identity — a default source type it gravitates toward by intent, time after time. Notable findings: only ChatGPT and ChatGPT Search surface Wikipedia; Perplexity and Gemini both favour YouTube; Google AI Mode routes citations back into the Google ecosystem exclusively. A single cross-engine AEO strategy is insufficient; content format and placement decisions must be made per engine.

OpenAI Launches GPT-5.6 Series: Sol, Terra, and Luna Models

OpenAI Developer Community · 09/07/2026

OpenAI announced the GPT-5.6 model series — Sol, Terra, and Luna — released via API on 9 July 2026. Full capability and context-window specifications were not available in the sourced material. Practitioners building retrieval or citation-monitoring pipelines against the OpenAI API should verify whether GPT-5.6 variants alter web-search grounding behaviour or citation formatting relative to prior model versions.

Google Expands Managed Agents in Gemini API: Background Tasks, Remote MCP, Custom Functions

Google DeepMind Blog · 07/07/2026

Google DeepMind announced expanded capabilities for Managed Agents in the Gemini API, including asynchronous background execution, remote MCP server integration, and custom function calling — all managed within an isolated cloud sandbox. For GEO practitioners, the addition of remote MCP connectivity and web-information retrieval inside agent workflows increases the surface area through which Gemini-powered agents can access and cite external content. Ensuring crawlability and structured markup on priority pages remains directly relevant as agentic retrieval scales.

Google Introduces AI Transparency Labels for Ads: Mandatory Disclosure of Generative AI Use

Google Ads Blog · 09/07/2026

Google announced new mandatory disclosure requirements for ads created or edited with generative AI, surfaced via a 'How this ad was made' panel accessible from Search, YouTube, and Discover ad menus. Advertisers using Google's own AI tools are labelled automatically; those using third-party AI tools must apply labels manually. While primarily an ads-compliance requirement, the move signals that Google is codifying AI-content provenance across its surfaces — a trajectory practitioners should watch as analogous labelling logic could eventually extend to organic AI-generated content.

ChatGPT's Four Data Sources Explained: Training Corpus, Live Web, Licensed Content, and Prompt Context

LLM Pulse · 09/07/2026

A detailed breakdown published by LLM Pulse documents the four distinct data pipelines feeding ChatGPT responses: frozen training data (web crawl, books, code, licensed datasets); live web search when real-time retrieval is triggered; OpenAI-licensed publisher content; and user-supplied prompt context and memory. The article notes that each pipeline has different latency, recency, and hallucination-risk profiles. For practitioners, understanding which pipeline is likely to surface a given answer type is the prerequisite for deciding whether to invest in on-site structural optimisation, third-party citation building, or publisher licensing relationships.

Nvidia Patents Video Data Curation Pipeline for World Model AI Training

Patentlyze (USPTO publication US 2026/0196042 A1) · 09/07/2026

A USPTO patent application (US 2026/0196042 A1, filed April 2025, published July 2026) describes Nvidia's automated system for scoring and filtering raw video for AI world-model training — assessing motion consistency, scene completeness, and annotation quality before data enters training pipelines. While focused on video for physical AI and robotics, the patent illustrates the growing sophistication of pre-training data quality gates across the industry. Practitioners producing video content for brand visibility should note that AI training pipelines increasingly apply structural quality filters; well-captioned, clearly structured video may clear those thresholds more reliably than raw footage.

DuoFlow-KG: A Dual-Modal Evidence Retrieval Framework for High-Density LLM-Augmented KGQA

Not listed in source · https://www.nature.com/articles/s41598-026-59526-3 · 09/07/2026

Published in Scientific Reports, DuoFlow-KG proposes a unified retrieval framework that jointly optimises semantic relevance and structural graph dependencies to build compact, high-density evidence subgraphs for knowledge-graph question answering — addressing the fragmented-evidence problem common in multi-hop reasoning. For GEO practitioners, the research reinforces why building dense entity relationships across content (rather than isolated topic pages) improves an AI engine's ability to construct coherent, well-sourced answers that cite your domain.

Entity-Centric Retrieval-Augmented Generation with Knowledge-Graph Re-ranking for Biomedical Question Answering (ECR-KG)

Che-Ju et al. · https://aisel.aisnet.org/pacis2026/ishealthcare/ishealthcare/4 · 2026

ECR-KG combines entity-centric query rewriting with Graph Edit Distance-based structural re-ranking to improve retrieval robustness in RAG systems, demonstrating gains on MedMCQA and MedQA benchmarks over standard semantic-similarity baselines, particularly at larger retrieval depths. The practical implication for GEO: entity-aware, structurally consistent content — where entity relationships are explicit and internally consistent — is more likely to survive graph-based re-ranking stages that increasingly sit between retrieval and final answer generation.