Reverse-Engineering ChatGPT: Source Selection, Runner-ups, and Hidden Prompts

ArticleSuganthan MohanadasanResearch

Suganthan Mohanadasan examines ChatGPT's network traffic payloads, revealing how source retrieval operates under the hood. Beyond cohort-gated fetch pipelines, an undocumented payload field exposes which competing URLs were considered as runner-up citations for specific claims.

Key points

  1. OpenAI removed the result_source field shortly after the initial teardown, regrouping sources under search_result_groups.
  2. The supporting_websites field reveals runner-up sources per claim that supported the statement but lost the visible citation slot.
  3. Retrieval pipelines differ across account tiers and regions, routing identical URLs through Bing, Bright Data, or Labrador.
  4. Conversations contain hidden system prompts such as sources_and_filters_prompt; query fan-outs actively append 'official' to factual lookups.

Why it matters

For Generative Engine Optimization (GEO), this shows that AI citation is dictated by backend policies rather than static ranking factors. The runner-up data provides an unprecedented, claim-level competitive audit of which sources narrowly missed out on citations.

Analysis

Mohanadasan's findings demonstrate the rapid churn of OpenAI's frontend retrieval plumbing. The supporting_websites property provides actionable intelligence: it exposes both competitor runner-ups and internal domain cannibalisation, where sibling URLs lose out to a primary page. Additionally, observing ChatGPT append 'official' to factual fan-out queries underscores the necessity of maintaining authoritative, plain-HTML specifications and pricing directly on primary brand domains.

Mentioned links

Found in

Summary generated by AI (Gemini 3.8 Flash, 27 September 2026). It may contain errors – the original source is authoritative.

Related entries

Similar in content, found by the AI search.

  • X post:Lily Ray

    Shifts in ChatGPT Citations Coincide with Bing Retrieval Rollout

    SEO specialist Lily Ray reports a trend reversal in ChatGPT citations across many websites starting around 4-5 July 2026, coinciding with Bing being identified as a retrieval source for ChatGPT.

    30bookmarks7.3kviews

    AI & LLMs· Opinion

  • Article:Gintare Rimolaityte

    Analysis of 107 Million AI Answers: Where AI Search Engines Pull Citations

    An analysis of 107 million AI answers across ChatGPT, Perplexity, Gemini, and Google AI Overviews reveals the primary citation sources per industry. While user-generated content like Reddit and YouTube accounts for roughly half of all citations across the board, the remaining half differs sharply between B2B, consumer goods, and communications.

    AI & LLMs· Research

  • Link:Josh Blyskal

    AEO and GEO Research by Josh Blyskal

    Josh Blyskal, Head of AI Strategy & Research at Profound, shares data-driven research exploring how AI search engines like ChatGPT, Claude, and Gemini source, evaluate, and cite brands.

    AI & LLMs· Resource

  • Article:Jen Cornwell

    Tracking AI Platform Citation Patterns in 2026: Tinuiti and Profound Report

    A study by Tinuiti and Profound examines how seven major AI platforms cite web sources across commercial search queries. The findings reveal sharp divergences among platforms, distinct behaviors across Google surfaces, and the strategic implications of platform crawler blocks by players like Amazon.

    AI & LLMs· Research

Back to the knowledge base

Let's talk about your project

Get in touch

Locations

  • Mattersburg
    Johann Nepomuk Bergerstraße 7/2/14
    7210 Mattersburg, Austria
  • Vienna
    Ungargasse 64-66/3/404
    1030 Wien, Austria

Parts of this content were created with the assistance of AI.