Nikki Lam and Samanyou Garg analyzed roughly 16 million brand appearances and found ~40% of AI citations link the source without naming the brand in the answer. The split is structural: Perplexity (52%), Google AI Mode (49%) and AI Overviews (41%) are citers that link constantly and omit names, while Gemini (25%) and Copilot (19%) name brands but cite less often.
Why it matters: Citation-count dashboards are measuring the wrong thing. Mentions and citations need to be two separate metrics in any AI-visibility report.
Kevin Indig's H1 report puts ChatGPT at 56% share, down from 78% a year ago, with Gemini tripling to 30% and Claude at 10%. AI Mode is at a billion MAU with queries roughly 3x longer than classic search, and publishers are down ~33% in referral traffic year over year.
Why it matters: An AI-visibility audit scoped to ChatGPT now covers barely half the surface. Gemini and Claude coverage is no longer optional.
Alex Galinos codified a 14-step page-update process into Claude Code skills that enforce the steps in sequence, from a 56-day Search Console baseline through a delta brief that only rewrites changed sections. Across 59 completed tests: 2,284 additional organic clicks, 24.9% more organic purchases, 20.1% revenue lift against locked baselines.
Why it matters: The lesson is where the judgment lives. Enforce the steps in the harness rather than asking for them in the prompt.
Yelp licensed its reviews, ratings, photos and business details to OpenAI, so local recommendations inside ChatGPT now surface Yelp content with Yelp branding and links. A Request a Quote flow is coming.
Why it matters: For local and service businesses, ChatGPT visibility is now downstream of Yelp hygiene. An aggregator nobody controls became a ranking input overnight.
IAB Australia published a bots-and-crawler decision matrix that forces publishers to one of four verdicts per crawler: allow, allow with conditions, require licensing, or block. It cites Cloudflare data showing automated requests hit 57.5% of web-page requests in June 2026, with ~52% of AI crawling feeding training and 2.6% serving real-time queries. A September 15 Cloudflare deadline is forcing the calls.
Why it matters: The blanket block-the-AI-bots instinct is about to cost sites citations. The real question is which crawlers return anything at all.
TAG and ANA's first statistically rigorous measurement puts AI slop at 1.3-2.4% of open-web programmatic spend, and it beats legitimate supply on every verification signal: 0.05% invalid traffic versus 0.32%, 77.2% viewability versus 74.9%, and a $7.08 TrueCPM versus $6.15. 11,552 domains classified, 88% of them also made-for-advertising, with vendors putting 25-40% of social video inventory in the same bucket.
Why it matters: Brand-safety scores have become an anti-signal. An unusually clean programmatic report is the one to go open up.
Knak surveyed 300+ enterprise marketing leaders: 70% run AI in production, 88% say output needs moderate to substantial human editing, and adoption is stacked on copy (64%) rather than the production layer (25% for building emails and landing pages). Approvals (47%), design production (38%) and cross-team coordination (36%) are what actually blows deadlines, with one email still touching four or more people and roughly $300 of internal labor.
Why it matters: The agentic work worth buying is in approval and assembly, not copy generation. That is where the hours and the money sit.