a second independent dataset says the AI referral is almost invisible. Scrunch found 1.1% of post-chat visits to news sites carried an AI referrer, while about three quarters arrived as direct navigation.

This corrects our 10 August issue, which reported that publishers cannot cleanly opt out of AI Overviews and that Google-Extended is training-only. The first half is now wrong. Google has a documented control under Settings > Search generative AI in Search Console that excludes a property from AI Overviews, AI Mode, and AI Overviews in Discover, while leaving ordinary Search results alone. Google's own words: "this control isn't used as a ranking or inclusion signal." Content is excluded within one to two days. Inclusion is the default for all properties, child properties inherit the parent's setting, and Google says it is rolling out "to a subset of website owners" for testing. The background is regulatory: the UK Competition and Markets Authority opened a consultation on Google Search, and this is the response.

Clarity's Bot Analytics now shows how much each AI operator crawls a site against how much referral traffic it sends back, broken out per operator. The example in Schwartz's write-up is roughly 6,000 scrapes for every 1 referral, on 41 referrals. We could not find a Microsoft primary announcement to read, so this is reported rather than verified at source.

Millions of searches, AI conversations and publisher visits from an opt-in panel, February to June 2026. Of visits to news publishers that followed an AI conversation, about 75% arrived as direct navigation, about 9% through traditional search, and 1.1% carried an AI referrer. Readers were 20.5 percentage points more likely to visit a news publisher during the week after a news-related AI conversation. Publishers the assistant actually named saw a 10.6 point jump by the next day and 19.7 points by the end of the week. Concentration is brutal: major publishers took 82% of the names volunteered by assistants and 97% of the follow-through visits, against 17% of names for mid-sized outlets.

Judge it honestly. Scrunch sells AI visibility software and was acquired by Sitecore in June 2026, so this is a vendor study with an interest in the conclusion. Panel size, provider and countries are not disclosed. It measures observed page visits only, not subscriptions or revenue, and Scrunch says so. One number in the coverage needs care: news searches with AI Overviews produced publisher clicks around 20% of the time against 30% without, but the within-query comparison narrows that to roughly 2 percentage points. Use the within-query figure. The headline gap is mostly different queries, not the same query with and without.

Q2 2026 against Q2 2025, across Shopify storefronts. AI-referred sessions grew 197%, AI orders roughly 3x, and organic search grew 12% on a much larger base. AI-referred shoppers converted about 80% better than organic overall, and about 2x better in spec-led categories, with watches at 2.4x and necklaces at 2.3x. AI brought 1.3x more first-time customers in taste-led categories. Half of AI sessions landed directly on a product page. The number that matters most to us: when the AI was working from structured first-party Shopify Catalog data, the shoppers it sent converted twice as well as those from scraped or third-party feeds.

Judge it honestly. First-party platform data, which is a real strength, but Shopify does not disclose merchant count, transaction volume, or method, and it has a commercial interest in merchants believing structured Catalog data matters. It is ecommerce. A coach is not a product page.