# Pulse ยท September 23, 2026 > Google auto-enrolls Shopify merchants into AI Mode checkout, the UK proposes AI-assistant search choice screens, and Databricks ships Genie One MCP to GA. Canonical: https://brandonlazovic.dev/pulse/2026-09-23/ Author: Brandon Lazovic Published: 2026-09-23 ## UK Regulator Proposes Choice Screens That Let AI Assistants Compete With Google as Default Search Status: Confirmed | Topics: organic-search-core, ai-overviews-ai-mode The UK's Competition and Markets Authority published strengthened proposals requiring Google to show Android and Chrome users a choice screen for their default search service, with annual reminders to reconsider. AI assistants such as ChatGPT and Perplexity could appear on that screen if they meet technical and security criteria, and any featured provider would have to attribute publisher content properly. Public feedback closes October 9, with a final decision expected by year-end 2026. In plain terms: A 'choice screen' is the prompt a phone or browser shows before a default is set, the same pattern Windows and Android already use for browser choice in the EU, here proposed for search and AI-assistant defaults instead. Why it matters: A choice screen that seats an AI assistant next to Google as a default option is a distribution channel practitioners have never had to compete for, since 'default search engine' has meant one company for most of the mobile era. Our take: If this clears in its current form, 'default search provider' stops being a single line item and becomes a placement AI assistants can win the same way a search engine would, on a screen most users tap through once and never revisit. The detail worth tracking is the 'technical and security requirements' bar for an AI assistant to qualify, since that clause decides whether this opens real competition or quietly protects whichever incumbents can already clear it. - [GOV.UK: CMA strengthens proposals allowing people choice over their search service](https://www.gov.uk/government/news/cma-strengthens-proposals-allowing-people-choice-over-their-search-service) - [Search Engine Roundtable: CMA Wants Google To Make It Easier To Choose Other Search Engines](https://www.seroundtable.com/cma-android-chrome-search-engine-42148.html) ## Google Auto-Enrolls Eligible Shopify Merchants Into AI Mode and Gemini Checkout, Opt-Out Only Status: Confirmed | Topics: agentic-commerce, product-feeds-shopping Google began automatically enabling UCP-powered native checkout for eligible U.S. Shopify merchants in Google AI Mode and Gemini, per Shopify's own help documentation and emails Google sent merchants on September 22. Products synced to Merchant Center and eligible for Shopify Catalog are included by default with no setup required; merchants must opt out by hand in Shopify admin under Sales channels > Agentic to keep purchases on their own domain. In plain terms: UCP, the Universal Commerce Protocol, is the shared technical standard Google, Shopify and Stripe built so a shopper can complete a purchase inside an AI tool's chat window, with the payment routed through Google Pay instead of the merchant's own checkout page. Why it matters: This flips UCP checkout from an opt-in a merchant had to request to a default a merchant has to notice and turn off, the kind of change that gets missed until a support ticket asks why an order shows Google Pay instead of the store's own checkout. Our take: I wrote in 'Shopify's catalog landgrab: build vs. syndicate for agentic commerce' that syndicating your catalog to a platform means accepting that platform's defaults, not just its distribution. Auto-enrollment into native checkout is that argument made concrete: a merchant who synced a feed for visibility now has to actively opt out of losing the checkout relationship, not opt in to gaining one. - [Shopify Help Center: Selling on Google AI Mode and Gemini](https://help.shopify.com/en/manual/online-sales-channels/agentic-storefronts/google) - [Search Engine Roundtable: Google Merchant Center Notifications For Native Checkout](https://www.seroundtable.com/google-native-checkout-emails-42140.html) ## Spotted: ChatGPT Ad Density in Gaming Conversations Jumped From 3% to 47% in a Quarter Status: Observed | Topics: chatgpt-assistants, ads-paid Comscore's Q2 2026 AI Intelligence Report found sponsored-ad presence in ChatGPT conversations about video games and consoles rose from 3% in March to 47% in June, per Search Engine Journal's review of the report. The data comes from Comscore's opt-in panel of U.S. PC activity only; Comscore's own September 3 release publishes the same tracking methodology for travel searches (6% to 24% over the same window) but not the gaming figure directly. In plain terms: An 'opt-in panel' means Comscore's numbers come from a fixed group of users who agreed to have their browsing tracked, not from ChatGPT's own usage logs, so the sample size and how representative it is of all ChatGPT users are both unknown from what's published. Why it matters: A near-15x jump in ad density inside one conversation category is the kind of number that changes how a brand reads 'share of voice' inside ChatGPT, from a ranking question into a media-buy question. Our take: Ad density inside an AI assistant is the same metric marketers already track for a SERP, applied to a surface with no equivalent of a rank tracker yet. Watch whether Comscore or a competitor publishes the full category breakdown, since a 47% figure in one vertical says little about the categories that matter to most e-commerce brands until it does. - [Search Engine Journal: ChatGPT Ads Appeared In 47% Of Video Gaming Chats, Data Shows](https://www.searchenginejournal.com/chatgpt-ads-appeared-in-47-of-video-gaming-chats-data-shows/590481/) - [Comscore: Comscore Expands Its AI Intelligence to Measure Sponsored Chat Advertising](https://www.comscore.com/Insights/Press-Releases/2026/9/Comscore-Expands-Its-AI-Intelligence-to-Measure-Sponsored-Chat-Advertising-and-Its-Business-Impact) ## Databricks Takes Genie One MCP to General Availability Status: Confirmed | Topics: ai-data-stack, llm-models-agents Databricks moved the Genie One Model Context Protocol server from limited to general availability on September 22, 2026, giving any AI agent, including Claude, ChatGPT, Copilot and Cursor, a single governed interface into Genie One's business context. The server runs inside Unity Gateway as a managed service with centralized governance, fine-grained policies and audit logging, and exposes five tools including genie_ask and genie_get_query_result. In plain terms: Genie One is Databricks' business-data assistant; MCP, Model Context Protocol, is the standard wire format that lets an outside AI tool like Claude or ChatGPT ask Genie One a question and get a governed answer back, instead of each vendor building its own one-off connector. Why it matters: Any agent an SEO or e-commerce team wires up can pull governed, permission-scoped business metrics through one interface instead of a custom connector per tool, the piece that decides whether an agent's numbers actually match the dashboard's. Our take: I argued in 'An MCP Server Is a Supply-Chain Dependency With Reach Into Your Data' that the protocol standardizes the wire format, not the trustworthiness of what sits behind it. Genie One MCP's audit logging and permission-aware access are the right shape of answer to that risk, but the practical test is still whether a team can see, in one place, which of its agents queried what and under whose credentials. - [Databricks Blog: Genie One MCP: Give any AI Agent the Right Business Context](https://www.databricks.com/blog/genie-one-mcp-give-any-ai-agent-right-business-context) - [Databricks Blog: The Genie One MCP is now Generally Available](https://www.databricks.com/blog/genie-one-mcp-now-generally-available) ## Snowflake Adds Agent Observability to Trace Cost and Quality Across Multi-Step AI Workflows Status: Confirmed | Topics: ai-data-stack, llm-models-agents Snowflake announced on September 22 that Agent Observability is coming to its Observe platform in private preview, with a public launch event set for October 22. The feature traces each step of an agent's workflow, model calls, retrievals, tool invocations, plus token usage, cost, latency and errors, and adds LLM-as-judge scoring to flag quality problems across multi-step AI applications. In plain terms: 'LLM-as-judge' means using a second AI model to grade the first model's output for quality, a stand-in for the human reviewer that would otherwise have to check every response by hand. Why it matters: After a team starts running agents against production data, the first practical question is which step burned the budget or returned a wrong answer, and that is exactly the trace Observe is built to produce. Our take: Observability tooling for agents is following the curve APM tooling took for microservices: instrumentation arrives after enough production traffic makes the blind spots expensive. The number worth watching after this ships is whether a team can attribute a specific bad answer to a specific retrieval step, since that is the harder half of the promise. - [Snowflake Blog: AI Agent Observability: Monitor, Debug, Optimize LLM Applications](https://www.snowflake.com/content/snowflake-site/global/en/blog/ai-agent-observability-monitor-debug-optimize-llm-applications)