# ChatGPT-5.6 Luna shifts to site: search, cites fewer sources

Published: 2026-09-24T21:49:27.935Z · Source: Seer Interactive (https://www.seerinteractive.com/insights/gpt-5.5-vs-gpt-5.6-luna)
Modified: 2026-09-29T03:55:58.000Z
Source date: 2026-09-24
Flags: Seer Interactive study — vendor-published
Entities: chatgpt, openai

> **Correction, 2026-09-28:** This item originally said Luna used the site: operator in "76% of searches" and that "total search volume held roughly steady." Seer measured fan-out queries — the sub-queries ChatGPT runs while answering — not searches: 76% of Luna's fan-out queries used site:, against 21.6% for GPT-5.5, and the total number of fan-out queries held steady (729 versus 733).

ChatGPT-5.6 Luna, now OpenAI's default ChatGPT model, runs far more site-targeted searches than its predecessor GPT-5.5, per a Seer Interactive analysis of 346 matched prompts across six industries. Luna used the site: search operator — limiting a query to one website — in 76% of its fan-out queries, up from 21.6% for GPT-5.5, while consulting 43% fewer sources overall (3,481 versus 6,134) and generating about 13% fewer inline citations, even as the total number of fan-out queries held steady (729 versus 733). Seer notes a brand Luna doesn't search for directly by name may see fewer citations under the new model, and that most AI-visibility trackers measure only citations — the last step in that retrieval process.

Why it matters: As ChatGPT's default retrieval shifts from open web search toward targeted site: queries, showing up among the specific domains a model searches directly — not just ranking broadly on the open web — becomes the lever for AI citation visibility.

## What this answers

**Does ChatGPT-5.6 Luna search differently than GPT-5.5?**

Yes — per a Seer Interactive analysis of 346 matched prompts, Luna used the site: search operator (limiting a search to one website) in 76% of its fan-out queries, up from 21.6% for GPT-5.5, while consulting 43% fewer sources and producing about 13% fewer inline citations.

**Does ChatGPT cite fewer sources with newer models?**

In Seer Interactive's comparison, ChatGPT-5.6 Luna generated about 13% fewer inline citations than GPT-5.5 across the same 346 matched prompts, even though it ran a similar number of searches overall.


Canonical: https://anythingengineoptimization.com/item/2026-09-24-chatgpt-5-6-luna-shifts-to-site-search-cites-fewer-sources/
From Anything Engine Optimization (AEO Wire) — https://anythingengineoptimization.com/ · Standards: https://anythingengineoptimization.com/standards/
