ChatGPT’s search responses now carry a labeled field identifying Bing as the underlying result source, according to a finding that researcher David Konitzny shared publicly and that Search Engine Roundtable reported on July 16, 2026. Konitzny said he found the label while inspecting the Network tab in Chrome DevTools during routine testing, and that he reproduced it across multiple prompts.
The field itself is narrow: a key called “result_source” returns the value “bing” inside ChatGPT’s response payload. That is a technical confirmation, not a product announcement from OpenAI. No new feature was unveiled, and OpenAI has not commented on the finding.
The significance sits in what the label replaces. For more than a year, third-party researchers have inferred a Bing connection by comparing citation sets: pages that rank near the top of Bing’s organic results turn up disproportionately often in ChatGPT and SearchGPT answers, a pattern attributed to ChatGPT search drawing on Bing’s index. That was circumstantial evidence built from output comparison. A named field in the response payload is a direct data point instead.
The distinction matters for how a search team allocates effort. If a site never gets crawled and indexed by Bing, or if Bing ranks it poorly for a query, the same weakness now has a documented, first-party channel through which it can suppress that page’s odds of being cited by ChatGPT. GEO (generative engine optimization, the practice of optimizing for LLM-driven answers) practitioners have often treated Bing as a low-priority secondary engine behind Google. This finding argues for reversing that priority when the specific goal is ChatGPT visibility.
Practically, teams should confirm Bing indexation status through Bing Webmaster Tools, verify that IndexNow submissions are current, and check organic ranking position on target queries directly in Bing search rather than relying on Google Search Console as a proxy. A page that ranks on page one of Google but is unindexed or buried in Bing has a specific, fixable failure mode now worth testing against ChatGPT citation rates.
The caveat is proportional to the evidence. Konitzny’s finding is a single researcher’s reproducible observation of an internal field name, not a disclosure of ChatGPT’s full retrieval architecture. OpenAI has not stated what share of queries route through Bing, whether the field applies uniformly across ChatGPT’s search modes, or whether the labeling is stable rather than an artifact of a specific backend version. Response payload fields also change without notice, and a field observed in July is not guaranteed to persist.
Even with those limits, the finding gives search teams something they lacked: a testable hypothesis rather than a correlation. Teams that already track ChatGPT citation rates for their pages can now cross-reference those same URLs against Bing ranking position and indexation status, turning an assumed relationship into a measurable one over the next reporting cycle.
Search Engine Roundtable reported the finding on July 16, 2026, based on a discovery shared publicly by researcher David Konitzny.
