Super GEO Agent Readiness

A Claude Skill that covers Generative Engine Optimization (GEO) and agent readiness in one place. SEO targeted blue links. This skill targets the answers AI engines produce, and the agents that fetch and act on your content.

Recent updates

Network-traffic analysis of ChatGPT source selection (latest)

Integrated findings from Suganthan Mohanadasan's June 2026 analysis of ChatGPT's raw browser network traffic into the platform, content, technical, and measurement references.

1. Four-tier source pipeline. Every web result ChatGPT pulls is stamped internally with a result_source field carrying one of four values: serp (open web baseline), labrador (licensed allowlist of national publishers like Reuters, WSJ, Wikipedia, arXiv), bright (Bright Data scraper, dominant for shopping, finance, weather, local), and oxylabs (Oxylabs scraper, regional press). Most sites compete in the scraped tiers, so clean scrapability plus third-party coverage is the lever; the licensed tier is largely shut.
2. Query triage. Before searching, ChatGPT files the question into one of six turn_use_case buckets. The text bucket triggers no web search at all and answers from training, including some current, high-stakes queries. Confirm a target query actually triggers retrieval before investing in a page.
3. Fetched, cited, and mentioned are three outcomes. A page can be pulled into context, attached to a specific sentence, or named as a brand chip, and you can win or lose each independently. This independently confirms the retrieval-gate finding from the Ahrefs pass below: retrieval is not citation. Two unrelated methods now reach the same conclusion.
4. The JavaScript pricing trap. When pricing or specs sit behind JavaScript, the model cannot parse them and sources your numbers to a third-party aggregator instead. Captured in the model's own recorded reasoning. Facts that you want cited must sit in plain server-rendered HTML.
5. One strong page per claim. ChatGPT dedupes retrieval by domain, so volume content collapses. One comprehensive page that owns a claim beats many thin ones.
6. DIY verification. A free, reproducible method to read the pipeline behind any ChatGPT answer from your own browser's Network panel, added to the measurement reference.

These are one researcher's single-account observations. Structural findings (which fields exist, that text queries skip the web) are firm; frequency findings (which domain dominates, percentage splits) are directional and vertical-skewed. Verify before quoting.

14 Ahrefs studies (2025 to 2026)

Integrated 14 Ahrefs studies published between October 2025 and May 2026 (roughly 1 billion data points across ChatGPT, Google AI Overviews, and AI Mode) into an evidence base at references/ahrefs-2026-studies.md, with the findings threaded through the content, platform, structured-data, measurement, and audit references.

Three findings reorder the playbook: