> ## Documentation Index
> Fetch the complete documentation index at: https://docs.peec.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# ChatGPT fan-out targeting — domains persist, URLs churn

> David Konitzny tracked 500 prompts for 17 days after GPT-5.6 launched (Aug 2026) — 91.9% of site: targets are bare domains, 18.1% of domains recur almost daily, only 3.5% of URLs repeat.

Domain-level visibility is a more stable signal than URL-level visibility in ChatGPT's fan-outs. David Konitzny tracked 500 prompts (no brands or domains mentioned) daily for the 17 days after GPT-5.6's release ([full article](https://www.linkedin.com/pulse/how-unique-domains-urls-chatgpt-query-fan-outs-david-konitzny-aqmze/), Aug 2026), analyzing every `site:` occurrence:

* **91.9% of site: cases target only a domain;** roughly 9% include a full URL path — and those paths point at pricing, feature, or individual product pages.
* **Domains show persistence:** 45% are one-offs, 55% appear multiple times, and **18.1% recur almost every day** — a stable hard core, even as ChatGPT keeps discovering new domains at a slowing rate.
* **URLs churn:** 64.9% are one-offs and only **3.5% repeat across the dataset**; the cumulative distinct-URL curve keeps climbing with no plateau.

Interpretation: ChatGPT converges on a trusted set of domains while continuously exploring different pages within them. For measurement, domain-level visibility is the more meaningful signal; counting raw site:-operator occurrences without separating domain vs URL targeting misreads the behavior. How the operator is used qualitatively: [ChatGPT site-operator fan-outs](/research/chatgpt-site-operator-fanouts).

**Source:** [How unique are domains and URLs in ChatGPT query fan-outs?](https://www.linkedin.com/pulse/how-unique-domains-urls-chatgpt-query-fan-outs-david-konitzny-aqmze/), David Konitzny, LinkedIn, Aug 2026.