Shared AI training opt-out 🤖, Reddit blames AI Mode 📉, AI Overviews CTR data 📊

Issue 27, 2 November 2026

What changed in search and AI answers, for marketing teams, in five minutes.

Headlines & Launches

Bing splits AI training controls from AI search grounding (4 minute read)

Microsoft Bing now lets site owners separately allow or block AI training, AI grounding (use in AI answers) and limited grounding for Microsoft's own products, while staying eligible for normal Bing rankings. Google, OpenAI and Anthropic are also on board. This means you could block your content from training a model while still letting Copilot or other AI answers cite it, so it is worth reviewing your robots.txt settings before the December rollout.

Google, Microsoft and OpenAI plan a shared AI training opt-out (6 minute read)

The three companies say they will honour one shared robots.txt line so sites can opt out of AI training across providers in a single step, based on an Anthropic-submitted IETF draft. None of the three currently supports it yet, and Google says it will treat the shared line as a Google-Extended block by default. If you want to keep content out of AI training without touching AI answer visibility, you will likely need to set this separately once it ships.

Google opens a preview for apps that run inside AI Mode and Gemini (4 minute read)

Google opened public access for developers to connect booking and ordering apps to AI Mode and Gemini via MCP servers, with a policy against misleading descriptions written to win selection. If you sell bookings or products, this is a new channel where an AI agent, not a search results page, decides what a user sees and buys from.

Shopify pays AI shopping agent developers a commission on sales (3 minute read)

Shopify now pays the company behind an AI shopping agent 0.3% of orders on promoted products, rising to 1.5% in Shopify's own example if a store adds extra promotion, and some of those sales route through shop.com rather than the store's own site. This gives agents a financial incentive that is separate from what is actually the best match for the shopper, so if you sell on Shopify, check how your promoted placements and campaign settings affect where the traffic and the sale land.

Deep Dives & Analysis

Google's Q3 earnings call shows query growth but less advertising per query (5 minute read)

Google reported its strongest year-over-year query growth in five years, which it credits to AI Mode, including more commercial queries, but acknowledged those AI interactions generate less advertising revenue per query than classic search. Sundar Pichai described a shift toward AI-first search experiences. For the reader, this is Google's own evidence that more searches are happening inside AI answers, where the value to advertisers, and the click-through to your site, is lower than in the old blue links.

Reddit's CEO blames falling search referrals on Google's AI features (5 minute read)

Reddit told shareholders its search referral traffic fell by more than 1 million visits per week on average in October, after a third quarter already described as down meaningfully, and CEO Steve Huffman named Google's AI Overviews and AI Mode as the leading suspected cause. Reddit will test requiring logins outside the US, which would change what a logged-out searcher sees after clicking through from Google. This is a concrete, named example of a major publisher losing search traffic to AI answers, worth watching if you rely on Google referrals too.

Big travel brands dominate AI search answers, but niches vary (4 minute read)

BrightEdge research shows well known travel brands such as American Airlines, Royal Caribbean and Expedia lead their categories in both ChatGPT and Google AI Overviews, though the top brand differs by engine in some sectors, for example Hilton leads hotels on AI Overviews while Marriott leads on ChatGPT. Only 9% of tracked AI Overviews prompts had booking intent, but booking platforms appeared in 78% of budget-related answers. If you compete against bigger brands, this suggests AI answers favour established names overall but leave openings in specific sub-niches or query types.

Small study puts AI Overviews and AI Mode click-through at about 0.9% (5 minute read)

Cyrus Shepard's small case study, shared on LinkedIn, used Google's opt-out controls to compare a site's clicks with and without AI Overviews and AI Mode visibility, estimating roughly 0.9% click-through from AI impressions, and found that opting out also reduced clicks in regular web results. Barry Schwartz, who reported it, cautions this is based on a small sample. It is an early, rough benchmark for how much traffic you might lose by opting a site out of AI Overviews, but treat it as directional, not definitive.

Tactics & Tools

A real example of fixing alt text at scale lifted image relevance from 30% to 84% (4 minute read)

Dan Petrovic's team scanned Google Store's Pixel Watch 5 page, found 47 of 109 images with no alt text and others with placeholder or mismatched labels, then rewrote them with an AI tool, raising the average match between image and alt text from 30% to 84% across the page. If your product pages have thin or templated alt text, this is a concrete case for auditing and rewriting it, since it affects both accessibility and how well AI systems understand your images.

RankZero's free AI SEO audit (Sponsor)

See whether ChatGPT, Perplexity and Google AI Overviews can find and cite your brand. Free, results in 30 seconds, no signup.

Google tells platforms to noindex pages built for sharing, not search (7 minute read)

Google's spam guide, updated 26 October, now recommends that platforms apply noindex or block crawling by default on user-generated pages meant for direct sharing, collaboration or utility rather than public search discovery. If your site lets users create shareable pages, such as saved carts or shared documents, check whether they are unintentionally indexed and diluting your site's quality signals in Google's eyes.

Google plans a single AI training control across providers (6 minute read)

Google said it will voluntarily adopt a shared AI training preference standard being developed at the IETF, letting sites set one policy token in robots.txt instead of listing rules for each AI provider separately, with implementation planned for 2027.

Gemini stopped adding UTM tags to the links it cites (3 minute read)

Google quietly removed the utm_source=gemini parameter it had been adding to cited links since early September, so any traffic trend you were tracking by that tag will look different from 28 September onward even though referrer data is apparently unaffected on web.

Google may be bringing back the 100 results per page option (3 minute read)

Searchers are spotting the num=100 parameter working again over a year after Google removed it to block scraping, though Search Engine Roundtable could not reproduce it consistently, so treat this as an early, unconfirmed sighting rather than a rollout.

ChatGPT Ads now edits ad copy and links automatically by default (5 minute read)

OpenAI turned on creative enhancements for all ChatGPT Ads campaigns from 30 October, letting its system rewrite headlines, translate text and pick a different landing page on the fly without advertiser approval of each version; advertisers can opt out in Ads Manager.