The September change hits Search and PMax, reversing Google's 2025 line that non-Search campaign types would stay untouched. Accounts segmented by language lose the lever and inherit Google's inference instead.
Why it matters: Every "your controls stay" promise from an ad platform has a shelf life, so build reporting that survives the lever disappearing.
Microsoft's free Bot Analytics added a scrape-to-referral ratio, and the example in the wild is roughly 6,000 scrapes per single referral. It turns a vague AEO grievance into a number you can put in a deck.
Why it matters: A free tool that quantifies extraction versus return gives you an AI visibility baseline without buying another monitoring seat.
AI-referred sessions rose 197% year over year in Q2 while organic search still grew 12% on a much larger base, and AI shoppers converted at roughly twice organic in research-heavy categories. The displacement story is wrong; the conversion-quality story is real.
Why it matters: Best counter yet to panic about AI killing traffic, and it reframes AEO work as incremental high-intent volume rather than defense.
Semrush, Profound and Scrunch measure presence, not outcome, so operators are triangulating with first-party data and statistical modeling. ChatGPT referrals to B2B brands are up 303% year over year, which makes the attribution gap more expensive every quarter.
Why it matters: The gap between visibility tooling and revenue proof is where the real work is, so lead with the modeling and not the dashboard.
The indie agency built Orion, an in-house layer over licensed LLM APIs, running internal agents for contractor management and workday prioritization plus a client-facing workspace with search visibility and ad testing. Buying Copilot-class seats for 500-plus staff would have run about $3M, so the build cleared on procurement math alone.
Why it matters: The durable play is owning the orchestration and the data while licensing the models underneath.
Kochava's StationOne Workspaces moves campaign creation, bulk edits, creative QA, reporting and troubleshooting into conversational commands against Google, Meta, Reddit, Snap and TikTok. The tell is what went unstated: whether each workspace can actually write, or stops at reporting.
Why it matters: Read-only "agentic" tooling is the default failure mode, so the qualifying question for any vendor is which platforms it can write to.