ChatGPT's missing tracking tags were a bug and they are back, so the correction to our analytics post is now about one lost week, not a permanent change.

New since our 14 to 16 October items: the tags returned. Profound's share of cited ChatGPT answers carrying at least one tagged link was about 11% late on 15 October (UTC), then climbed to about 99% by 06:30 UTC on 16 October and held there, including inside the new "Intelligent UI" answers (about 96%). Cedric Huesler of Adobe, who first flagged the drop, updated his post with "It was a bug." and told Search Engine Journal that OpenAI confirmed that to him on 14 October; tagged clicks across the domains he tracks were back to roughly 95 to 100% of the same hour a week earlier. There is still no public OpenAI statement, OpenAI's release notes say nothing about link tracking, and Glenn Gabe had not yet seen the change on his own accounts. Huesler adds that most ChatGPT clicks still arrive without a referrer, so the tag is the thing to count.

The new clause authorises Meta to use "web scrapers, bots, robots, spiders, crawlers, and user agents" to "access, index, cache, analyze, collect, retrieve, or crawl" a business's domains and pages. Per Search Engine Journal, it covers any business use of Meta's products, from ads to simply running a Facebook Page, where the grant used to sit only in the ad terms, and Meta may use what it collects to develop and train its products. Continued use after 30 October counts as acceptance. A Meta spokesperson told SEJ that robots.txt still controls access, but the terms themselves do not mention robots.txt, and Meta's own crawler page says Meta-ExternalFetcher "may bypass robots.txt" for user-requested fetches.

ChatGPT's newest note (16 October) adds Slides and a faster GPT-6 Pro, nothing on search or sources. Google's changelog is unchanged since the 16 October Indexing API quota page. Perplexity's October entries now include Gemini 4 Argon and GPT-6 Astra in its Agent API and still name no Sonar model. Anthropic's newest post is still the 15 October evaluator update. On the deprecations page, `claude-haiku-4-5-20251001` is still Active with "Not sooner than October 15, 2026" and no notice, and `claude-sonnet-4-5-20250929` still retires 30 November.

First test: a UX agency bought an exact-match domain and built a small directory of UX agencies with itself at the top, tracked for 67 days. The site had 6 referring domains and an average Google position of 68.6, yet it was cited for "best UX agency" style prompts within weeks; 91% of its retrievals came from ChatGPT. It appeared in 38.6% of answers to the head-term prompt but under 5% of answers to buyer-style prompts. Second test: one page gained a "Key Takeaways" section of standalone sentences plus schema, with a similar page left alone as a control; over six weeks the test page's citation rate rose 61% (1.45 to 2.33) against about 7% for the control, while its overall visibility fell from 22% to 9%. Limits: vendor case studies, one site and two pages, prompt counts not stated, citation rate is Peec's own metric, and B2B agencies, not coaches.

SEJ gathers the measurements: Graphite saw ChatGPT-attributed sessions fall 22% to 57% across its tracked sites from 9 October (about 45% from baseline, by its estimate), LLM Pulse about 46% with total traffic steady, and Peec found direct links in 42% of responses, up from about 8 to 10%, with 54% of those pointing to homepages. The Graphite and Peec figures were seen only through SEJ, and SEJ notes several sources sell tools and two do not give sample sizes.