# Ahrefs: 14.4% of sampled ChatGPT responses contained fan-out queries

Published: 2026-10-12T09:05:19.552Z · Source: Ahrefs Blog (https://ahrefs.com/blog/query-fan-out-chatgpt/)
Modified: 2026-11-07T04:24:48.000Z
Source date: 2026-10-12
Flags: Ahrefs study — vendor-published
Entities: chatgpt, openai

> **Correction, 2026-11-07:** Ahrefs corrected its source in October: 14.4% of responses contained fan-out queries, rather than 14.4% searching the web. Our headline and inference that six in seven answers relied only on prior associations were too broad and are corrected. Reddit’s separate figure measures answers citing Reddit, not its share of citation links.

Ahrefs analyzed 66.6 million free, logged-out ChatGPT-5.6 Luna responses from August and September 2026 and found 14.4% contained fan-out queries, averaging 1.6 fan-out queries — the sub-queries ChatGPT runs while answering — each. Of the resulting 15 million fan-out queries, 6.8% used the site: operator to search one domain directly, spread across roughly 116,000 sites; apple.com, ikea.com and nike.com topped the list by mentions. That 6.8% sits far below the 76% that a smaller Seer Interactive study measured on a narrower prompt set in September — Ahrefs' own comparison table attributes the gap to sample scope and prompt mix, not a methodology error in either study.

A second, separate Ahrefs dataset tracking actual citations (2.4 to 3.8 million ChatGPT answers a month, via Brand Radar) confirms the citation collapse this wire has tracked since August at larger scale: the share of answers citing Reddit fell from roughly 1 in 5 answers in March to 1 in 43 by late September, an 88% drop, and Wikipedia fell from 1 in 7 to 1 in 35, down 80%. Over the same stretch, the combined citation share of OpenAI's 17 licensed content partners more than doubled from 7.5% in March to a July peak of 15.8%, before easing to 12.5% by late September. Reddit itself still ranks ninth among the domains ChatGPT searches directly with the site: operator — Ahrefs reads this as the model still fetching Reddit for opinions while citing it less.

Why it matters: Retrieval measures need precise definitions: a fan-out query count cannot establish that an answer used no web sources.

## What this answers

**What does Ahrefs’ 14.4% measure?**

The share of its sampled ChatGPT responses containing fan-out queries. Ahrefs corrected its original description of this as the share searching the web; the figure cannot establish how often ChatGPT uses any web retrieval.

**How often does ChatGPT use the site: search operator?**

6.8% of its 15 million fan-out queries used the site: operator to search one domain directly, spread across roughly 116,000 domains, per Ahrefs — far below the 76% a smaller, narrower Seer Interactive study found in September, a gap Ahrefs attributes to sample scope rather than a wrong number.

**Is ChatGPT citing Reddit and Wikipedia less often?**

Yes. Per Ahrefs' Brand Radar citation data, the share of ChatGPT answers citing Reddit fell from about 1 in 5 answers in March 2026 to 1 in 43 by late September (down 88%), and Wikipedia fell from 1 in 7 to 1 in 35 (down 80%), while OpenAI's licensed content partners rose from a 7.5% combined citation share to a July peak of 15.8%, easing to 12.5% by late September.


Canonical: https://anythingengineoptimization.com/item/2026-10-12-ahrefs-chatgpt-triggers-a-web-search-on-just-14-of-prompts/
From Anything Engine Optimization (AEO Wire) — https://anythingengineoptimization.com/ · Standards: https://anythingengineoptimization.com/standards/
