Week of September 8–14, 2026
Four things moved in AI this week that touch how visible you are in search, how safe your data is with the AI tools you already use, and how freight gets quoted. Here’s what changed and what to do about it.
A new study of 300 real searches found Google shows an AI Overview 93% of the time, and pages ranked below it get almost no clicks.
On September 10, AI marketing consultant Lilach Bullock published a SERP study built from 300 real search queries, the ones with the most impressions on her own site over the prior 28 days, fetched live from Google’s US desktop results that same day. An AI Overview appeared on 280 of the 300 queries, 93%. Coverage varied by topic, from 100% down to 77% depending on the category, and YouTube was the most-cited source inside those Overviews at 57%, with Reddit second at 20%.
The click data is the part worth sitting with. Pages ranking positions 4 through 6, below the AI Overview, pulled just 0.7% click-through. Positions 7 through 10 did slightly better at 1.7%. Those are the ranking spots most small businesses have spent years fighting for, and they’re now converting at a fraction of what they used to.
This is one site’s dataset, not a Google-wide audit, so treat the exact numbers as directional rather than universal. But it lines up with the broader pattern most marketers have already seen this year: ranking below the AI Overview isn’t the same as being visible.
If your whole SEO strategy has been "get to page 1," you may already be there and still not showing up to the person searching, because they never scroll past the AI Overview. The more valuable spot now is getting cited inside that Overview, not just ranked underneath it.
Search a handful of your own top keywords in an incognito browser and check two things: does an AI Overview show up, and is your business cited in it, not just ranked below it. If you're not cited, the fix is usually content that answers the actual question directly and clearly, not more keywords.
Google, Anthropic, and OpenAI all shipped dedicated cybersecurity AI models and new data-handling tiers in the same week.
On September 9, The Hacker News reported that all three major AI labs released dedicated cybersecurity models and access programs within days of each other. Google announced Gemini 3.8 Flash Cyber, its most capable security-focused model, available through a “Fairwind Program” that gives priority access to defenders like governments, healthcare systems, and telecoms. Anthropic released Claude Fable 5.1 and a more restricted Claude Mythos 5.1, alongside Enterprise Frontier Safeguards (EFS), which lets an organization keep its AI usage data in infrastructure it controls rather than the provider’s, while still getting misuse detection across sessions. OpenAI said its Astra model meets the “Critical” cybersecurity capability threshold under its own safety framework, and acknowledged its safeguards can sometimes flag legitimate work as suspicious.
The cybersecurity models themselves are built for large defenders, not small businesses. What’s more relevant here is the data-handling piece: these providers are now actively splitting their plans into tiers with different levels of data retention and privacy control, on top of whatever “Business” or “Enterprise” label you’re already paying for.
Healthcare practices, law-adjacent firms, and anyone handling client financial data through ChatGPT, Claude, or Gemini should check which specific data-retention tier their plan is actually on. Paying for the top-tier plan doesn't automatically mean you're on the strictest privacy setting anymore.
If your team uses paid AI tools on sensitive client data, check your plan's data retention and safeguard settings this week rather than assuming the default is the safest option.
A new report predicts AI risk will factor into 60-80% of liability and cyber insurance underwriting by 2028.
IT research firm ScienceSoft released a forecast on September 10 projecting that by 2028, 60-80% of new policies and renewals at midsize US insurers across four lines, errors and omissions, directors and officers, employment practices liability, and cyber insurance, will explicitly factor in AI-related risk. The forecast is built on a documented 262% rise in publicly reported AI-related incidents between 2022 and 2025. Most insurers are expected to fold AI risk into existing policy types rather than sell separate standalone “AI insurance” products.
Nothing changes on your policy this year. But the direction is the same one cybersecurity questionnaires took a decade ago: a niche technical question that became a standard part of every renewal application.
If you're a contractor, property manager, or healthcare practice already carrying E&O or cyber coverage, this isn't urgent yet. It's worth a five-minute conversation with your broker before it shows up as a surprise question on next year's renewal.
Ask your insurance broker at your next renewal whether they're already asking about AI tool usage in your business, and if not, when they expect to start.
DHL is wiring freight quotes and booking directly into Alibaba.com's AI agent platform for small importers and exporters.
At Alibaba.com’s CoCreate 2026 conference in Los Angeles, DHL Group and Alibaba.com signed a memorandum of understanding to connect DHL Global Forwarding’s freight quoting and booking tools directly into Accio, Alibaba.com’s AI agent platform for small and medium-sized businesses. The idea is that a business sourcing or selling internationally through Accio can get a real-time freight quote and book the shipment through the same AI agent handling the rest of the transaction, instead of a separate call or email to a freight broker. Payload Asia reported it’s the first of a planned set of DHL capabilities the two companies intend to build into Accio.
This is an MOU and early integration, not a live feature yet, so there’s nothing to act on today. It does signal where freight quoting is headed: another manual step getting folded into an AI agent alongside sourcing and payment.
Distributors and retailers here who source building materials, medical supplies, or retail goods through Asia-Pacific routes should expect the quoting and booking layer of that process to keep consolidating into fewer, AI-driven steps over the next year or two.
If you already source through Alibaba.com and ship with DHL, nothing to do yet. Watch for this integration going live; it may replace a manual freight-quote step you're doing today.
What we're watching next week
-
What actually gets a business cited in an AI Overview. This week’s study showed the click gap; we’re watching for more data on what separates a cited source from one that’s merely ranked underneath.
-
Anthropic’s Enterprise Frontier Safeguards rollout. Anthropic said EFS reaches customers in phases “later this fall.” We’re tracking when it’s actually available to smaller Claude for Business accounts, not just large enterprise contracts.
-
The DHL-Accio integration going live. The MOU signed this week has no shipping date. We’re watching for when quoting and booking through Accio actually launches.