Google's spam update finished this morning after two weeks, so Google AI Overviews readings can be treated as settled again.

The incident now carries a second post, timed 8 October 2026 at 01:37 Pacific: "The rollout was complete as of October 8, 2026." It began on 24 September at 09:15 Pacific, so it ran fourteen days. Barry Schwartz notes the contrast with earlier 2026 spam updates: "this one took two weeks to roll out whereas the others took only two days to roll out." Google has not said what the update targeted, how many searches it touched, or anything about AI Overviews. Separately, Schwartz reports that two consultants (Marie Haynes and Daniel Foley Carter, both on X on 7 October) are seeing more manual penalties, with thin content and scaled content named. That part is practitioner talk with no Google statement.

GPT-6 "replaces the 'Latest' option in the model menu". Paid plans got GPT-6 Sol on 7 October, and the release "will extend to Free and Go users from tomorrow", meaning today, on GPT-6 Luna. Answers can now mix text with charts, comparisons and small interactive tools. On search, Search Engine Journal reports the model "makes better decisions about when to look something up" and that the announcement "doesn't say how sources will appear" in the new formats. Nothing was said about shopping or about how businesses are named.

The deprecations page now lists `claude-haiku-5-5` as Active, "Not sooner than October 7, 2027", and Anthropic's newsroom features "Introducing Claude Haiku 5.5" dated 7 October. The older `claude-haiku-4-5-20251001` is still Active with "Not sooner than October 15, 2026". `claude-sonnet-4-5-20250929` is unchanged: deprecated, retiring 30 November 2026.

In March 2026 the author built a site of 50,000 statistics pages in a day for about $50. In his words: "Every number was totally invented." His results: "Google has forgotten the site exists and deindexed the pages, but ChatGPT hasn't." He reports "more than 600,000 requests from bots" in the last 30 days (about 72,000 from ChatGPT's browsing agent), "10,000+ sessions from ChatGPT alone" since March, and that "its made-up numbers have also ended up in an academic paper and in a bank's investment research." He adds "I absolutely do not recommend coming up with bogus data." Limits: one site, one author who now works for a tracker vendor, his own analytics, ChatGPT only, and no count of how many answers actually repeated a false number.