Opening

This week, the thing you should really be watching isn't which ad went viral, it's that your GA4 report might be lying to you: ChatGPT has dropped its outbound tracking parameters, and AI referral traffic has overnight turned into "direct traffic." Meanwhile, Google Ads has rolled out channel priority and budget allocation controls, and the link placement and source rules in AI Mode have been overhauled across the board. I'll get to why your GEO baseline needs a reset in a moment. I'm Muyan, welcome to AI Marketing Weekly.

This Week's AI Marketing Highlights

1. Google Ads opens up channel prioritization and budget allocation controls

• What happened: Starting October 15, Google Ads is rolling out two opt-in controls across Search, Shopping, Performance Max, and Demand Gen: Channel prioritization lets you mark a channel as High or Low priority, so the AI feeds the high-priority channel first within your tCPA or tROAS targets; Budget allocation lets you cap how much of your daily budget a low-priority channel can consume.
• Marketer's take: This is a rare instance of Google handing the steering wheel back after years of automation, and it's useful for addressing the long-standing problem of pure search budget getting eaten up by Display or Search Partner Network.
• What to do now: Check your account to see if these two settings are available yet, mark your most important channel as High priority, and test on a small scale before scaling up.
• Source: Search Engine Land (https://searchengineland.com/google-ads-introduces-channel-prioritization-and-budget-allocation-controls-495946)

2. ChatGPT appears to have dropped outbound tracking parameters, turning AI referral traffic into direct traffic in GA4

• What happened: Since GPT-6 launched, ChatGPT's outbound links no longer carry the utm_source=chatgpt.com parameter it added back in June. Adobe's Cedric Huesler points out that the way links open also strips referrer information entirely. Third-party analysis shows that for the same set of pages, ChatGPT traffic is down roughly 31,000 visits per day, while direct traffic is up about 26,000 visits over the same period. OpenAI has not stated whether this is intentional.
• Marketer's take: This looks like an attribution shift, not an actual traffic collapse, but if your monthly report reads the GA4 numbers at face value, the AI referral line could be misread as shrinking.
• What to do now: Before putting together next week's monthly report, check whether the drop in ChatGPT traffic lines up with the rise in direct traffic, and cross-check using landing page patterns, branded search volume, or server logs.
• Source: Search Engine Land (https://searchengineland.com/chatgpt-appears-to-drop-utm-tracking-making-ai-referral-traffic-harder-to-measure-496552)

3. Google AI Mode / AI Overviews move links higher up, add official website blocks and creator profiles

• What happened: On October 15, Google announced that inline links in AI Mode and AI Overviews answers will now appear earlier in the response more often, so users can click through without reading the whole passage. Google is also testing a dedicated top-of-page block for one or two official websites, showing clickable video creator and publisher profiles in AI Mode, and expanding the real-time news carousel and Preferred Sources to more regions and languages. Google VP of Product Management Hema Budaraju said usage of Preferred Sources has grown more than 7x since the start of the year (per Google's own figures).
• Marketer's take: The GEO (Generative Engine Optimization) battlefield is shifting from "getting cited" to "getting clicked." It's worth shoring up your official brand pages and author identity signals this quarter.
• What to do now: Check whether your site has clear author or publisher markup, and whether Preferred Sources settings have rolled out to your industry yet.
• Source: 9to5Google (https://9to5google.com/2026/10/15/google-ai-mode-overviews-links/)

4. Gemini 4 Argon rolls out globally, lands directly in Search's AI Mode

• What happened: On October 13, Google began rolling out its next-generation model, Gemini 4 Argon, to Google AI Pro and Ultra subscribers worldwide, covering both the Gemini App and AI Mode in Search. The official line is that it's better at handling multi-step tasks that require back-and-forth intent clarification.
• Marketer's take: A new engine means the answer composition changes too. The brand mention rates and citation sources you tracked last month could get completely reshuffled overnight.
• What to do now: Rerun your usual brand and competitor keyword tracking this week and set a new GEO baseline. Don't compare old data against the new engine.
• Source: Search Engine Journal (https://www.searchenginejournal.com/google-gemini-4-argon-ai-mode/592806/)

5. Google's Universal Cart launches in the US, putting a shopping cart directly inside Search and Gemini

• What happened: On October 15, Google launched Universal Cart in the US, a cross-retailer cart that lives in both Google Search (including AI Mode) and the Gemini App. It's built on the Shopping Graph's 60 billion product listings, with price-drop and restock alerts plus Google Wallet deal suggestions. Checkout can happen via Google Pay or route back to the merchant's own site, with the merchant remaining the seller of record. Launch partners include Nike, Sephora, Target, Ulta Beauty, Walmart, Wayfair, and a number of Shopify sellers.
• Marketer's take: For e-commerce sellers, Merchant Center data quality and promotion setup just went from "only affects ad creative" to "determines whether you even get included in Google's cart."
• What to do now: E-commerce brands should check whether Merchant Center inventory, pricing, and promo codes are syncing in real time, since that's what determines Universal Cart eligibility.
• Source: PPC Land (https://ppc.land/googles-universal-cart-reaches-us-search-and-gemini-but-not-youtube/)

6. Kroger × Vidmob × MMA Global: predicting conversions before the creative even runs

• What happened: Kroger, together with Vidmob and MMA Global, published research analyzing 1,934 video and image creatives run on Meta and DV360. The model was trained on full-year 2025 data and validated against Q1 2026 data, and the researchers claim it can predict e-commerce conversions with 81% accuracy before a campaign even launches. Creatives that matched the model's recommendations saw average conversion rates 4x higher and cost per conversion up to 70% lower (all figures as claimed by Kroger, Vidmob, and MMA Global). The study also found that lifestyle-context creative outperformed plain product shots.
• Marketer's take: Now that generative AI has exploded the volume of creative assets, the bottleneck has shifted to filtering out the bad ones first. Predictive scoring is a ready-made way to turn creative debates into a shared language, but since these numbers come from the vendors themselves, treat them as directional, not a guarantee.
• What to do now: Before your next creative review meeting, pull historical data comparing "lifestyle context" versus "pure product shot" creatives and see which one your audience actually responds to.
• Source: Marketing Dive (https://www.marketingdive.com/news/how-kroger-uses-ai-and-predictive-scoring-to-tie-creative-to-conversion/831659/)

7. Fanta's AI Halloween ad "Haunted Universe" gets torn apart

• What happened: Fanta hired AI studio GRAiL to produce a long-term Halloween IP called "Haunted Universe," introducing original characters like a cyborg Frankenstein and a werewolf DJ, with a cameo from Ghostface of the "Scream" franchise. The comments section was overwhelmingly negative, with many viewers flatly saying they'd stop buying Fanta. Criticism centered on the idea that with all the resources of the Coca-Cola system behind it, the visuals still lacked craft and polish.
• Marketer's take: Whatever money AI saves on production can easily get clawed back on the brand goodwill side, especially for projects aiming to build long-term IP around characters and world-building. Audiences have very little tolerance for anything that looks cheap.
• Source: Creative Bloq (https://www.creativebloq.com/ai/fantas-ai-slop-halloween-advert-is-a-real-horror-show)

Closing

The one thing worth remembering this week is that AI lets marketers do more, but it doesn't automatically earn trust. From ChatGPT's traffic attribution to Fanta's comment section, it's really the same reminder. Next week, keep an eye on whether channel prioritization has rolled out in your Google Ads account, and whether your own GA4's AI referral traffic is showing that same "fake drop." See you next Tuesday.