Google called yesterday's missing citations a bug and fixed it inside 36 hours, and in the same 48 hours OpenAI shipped a new top model, which means two of our five engines changed the software behind their answers this week and our pipeline recorded neither.

Yesterday this radar led with Gemini 3.8 Flash answering questions in AI Mode without citing or linking anything. Later the same evening Robby Stein, VP of Product at Google Search, replied to Barry Schwartz on X, thanking him for flagging it and adding: "This isn't working as intended, and we'll roll out a fix soon." Google gave no date, scope or mechanism. The next morning Glenn Gabe posted on X, 4 September: "OK, looks like Gemini 3.8 Flash is now providing links. When it first rolled out, there was a bug and there were barely any links being provided." Search Engine Land added a "Fixed" note carrying that post. Note what is and is not confirmed. Google confirmed the bug. The fix landing is one named practitioner's observation, not a Google statement.

Yesterday's brief drew the conclusion that "the citation layer we sell is a model setting Google can move in three weeks." That was wrong, and it is worth saying so plainly rather than quietly dropping it. It was a defect, it lasted about 36 hours, and it only ever reached AI Pro and Ultra subscribers who had selected 3.8 Flash from the model dropdown. Free-tier AI Mode was never affected.

We did not carry this yesterday and should have. OpenAI's own wording: "We're introducing GPT-6 Astra, the world's most intelligent and aligned model", and "Astra is state-of-the-art on computer use, browsing, software engineering, cybersecurity, science, and professional work." On rollout: "GPT-6 Astra is rolling out today to a limited set of organizations and over the coming days will become available to all ChatGPT Plus, Pro, Business, and Enterprise users, as well as through the OpenAI API, Microsoft Azure, and AWS Bedrock."

Read it for what it is. This is a model release, not a retrieval or citation policy change. The announcement says nothing about how ChatGPT selects or attributes web sources, and the free tier is not named in the rollout list. The word that matters to us is "browsing", which OpenAI claims as a state-of-the-art capability, and the phrase "through the OpenAI API", which is the door our own collection code walks through.