AI Mode in US Search and Gemini now builds grocery orders for Target, Albertsons, Safeway and Shaw's. Google's own example put two Target private-label SKUs in a $37.49 Nashville pickup cart, turned "two bagels" into two six-packs, and never showed a payment screen. Connected Apps policy bars partners from serving third-party ads in the flow; Target's Roundel booked $296M last quarter.
Why it matters: For CPG and grocer retail-media teams, run an @Target brunch prompt this week and screenshot which SKUs the agent picks; that is the new shelf, not Roundel.
Glenn Gabe and Metehan Yesilyurt: ChatGPT deal cards now show product, price, expiry, Copy and Get deal on hosting coupon prompts (Bluehost, Hostinger, Cloudways, HostPapa). Deal cards on Black Friday product answers went from 3% to 16% in four weeks; Gabe: free ChatGPT shows the cards, Plus does not. One code was BLACKFRIDAYREDDIT scraped from affiliate pages.
Why it matters: For affiliate and promo-code teams, check which of their codes appear on ChatGPT cards this week and push exclusive codes that only work through their own link before Friday.
Cross-channel budgeting now splits by source platform so Meta, TikTok and Google sit on their own lines instead of "paid social." Plans are estimates; nothing changes in connected accounts. The DDA model needs a year of cost from at least two channels, one of them not Google, and it ranks those platforms using Google's own attribution.
Why it matters: If you already import Meta/TikTok cost into GA, open a source-platform scenario this week and treat the ranking as Google's view, not a budget instruction.
Ads in Bing AI Overviews now run through the Search, PMax and AI Max campaigns you already have, in every market except China, with no opt-in. URL inclusions are rolling out so AI Max can pull copy and landings from pages you name, like holiday gift guides.
Why it matters: Pull one Bing AI Overview query this week and check whether the ad is already sitting next to the answer; then add gift-guide URL inclusions before the bidder writes its own landing.
Marshall Evans cut a trophy seller from $2,000/mo to $500 and moved the rest to Tubi, Bing, Reddit and Nextdoor after AI Overviews buried the ads. OnlyFinder's Max Anderson says costs tripled or quadrupled, spend is about half, and there is still no predictable way to buy the AI Mode slot. VML's Valerie Schlosser: the biggest change in searcher behavior.
Why it matters: Treat Google Search as a shrinking slot this week, not a default; if the ad sits under the Overview, split a test cell to Bing/Reddit before Black Friday, not after.
Product leads Shah and Nevarez: 80% for directional budget/target/AI Max moves, 95% for structural ones, and the AI Max for Search experiment template is now the most popular way to test it. A shared-budget and portfolio-bid pilot opened this month; a $100 campaign testing $150 runs as $50 vs $75, so the trial itself spends $125.
Why it matters: If the team is still waiting for 95% to turn on AI Max, use the template this week at 80% and put shared-budget accounts in the new multi-campaign pilot instead of splitting them by hand.
Lluc Penycate's screenshots: a Continue on Google pop-up pushing Expedia at $170 covers the hotel's own $158 rate, then dumps the booking into AI Mode with Google Pay. No loyalty numbers, no promo codes; the partner stays merchant of record.
Why it matters: For hotel and OTA accounts, pull the brand query today and screenshot whether Continue on Google is hiding the cheaper direct rate; that is the new inventory, not a listing tweak.
The Nov 18 model upgrade auto-rolls: bids stick to CPA/ROAS more consistently, adjust nearly twice as fast on a target change, and pace more evenly through the day. Microsoft's August Search + Audience data: 10% more conversion value per click on Max Conversion Value / tROAS at similar ROAS. Navah Hopkins: five account-level conversions is enough to start conversion bidding.
Why it matters: Freeze Microsoft bid-strategy flips this week and watch pacing at campaign level; a CPA you did not touch can already be bidding into new inventory.
Canvas's Talos flags under/overpacing and only pushes after a person at the agency clicks yes. Behavioral signals come through an MCP to Dstillery (renter's insurance, moving trucks) because off-the-shelf multifamily packs were not converting actual movers. Zillow SVP Kandikonda sets the outcome; Canvas turns it into a scorecard.
Why it matters: If you still buy generic in-market packs, the next brief should name the intent signal and require human approve-before-push on the agent.
Lily Ray (Amsive) logged 4 clicks from 25 mentions on her consultancy dashboard because a mention only counts when the website is in the answer. Daily Mail's Carly Steven wants the data to validate GEO pitches; Shehata says a publisher's own citation count is not proof of a brand's GEO return. Half of six large publishers Digiday spoke with are still setting up; legal teams stalled on the ToS.
Why it matters: Claim the Hub this week, but split URL-mentions/clicks from third-party brand-mention scores and never treat a publisher citation as the brand's GEO ROI.
Sonicare used Amazon Ads Agent to buy Paramount+, Roku, Tubi, Peacock, TelevisaUnivision and WBD as one buy instead of six publisher deals. Priyanti Telang: 24% incremental reach on in-market oral-care audiences, 75% lower CPA, and a 15x purchase rate vs Sponsored Ads-only. Amazon replaced DSP with DVA+.
Why it matters: For retail/CTV accounts already in Amazon Ads, the agent is the consolidation layer; ask for reach and CPA vs the last publisher-by-publisher buy.
Google's Nov 17 allow-list tells sites to skip CAPTCHAs and per-IP limits for Google-Agent, including on-site search, GraphQL, add-to-cart and booking, because robots.txt generally does not apply and many users share one egress pool. Blend already saw bot protection block its shopping agent at 29 of 200 UK/US retailers.
Why it matters: Pull CDN and WAF rules this week and whitelist Google-Agent through checkout APIs, or agent shoppers never reach the cart.
Task mode dropped the $19.99 paywall in the US, Japan, Korea, Canada, Australia and India, running on Hydrogen. In the US it can call businesses (it discloses it is an AI; businesses opt out in Google Business Profile), sign into sites with Password Manager, apply a discount code, and pay with Visa/Mastercard AP2 tokens; on some sites it still hands the last click back.
Why it matters: Local and retail teams need a GBP automated-call setting and a promo-code policy for agent checkout before this hits volume.
Google's new Search Journey Awareness help page: tCPA, tROAS and Max Conversions on Search (including AI Max) and PMax Search inventory automatically learn from secondary actions if they sit in a real goal category, not Default. Changing secondaries does not swing live bidding; changing the primary goal does.
Why it matters: Audit every secondary conversion category before Black Friday week; Default-bucketed add-to-carts are invisible to the model that is already bidding.
Ribar (ex-Optus EGM, JPMorgan): keep north of 80% of customers on the champion, because every extra live test pulls people off it and puts revenue at risk. Agents, simulation and live tests are one loop; "do nothing" is a valid next-best action, and a 12-month universal control is how you stop experiment spaghetti from contaminating the P&L.
Why it matters: Before the next lifecycle test calendar, put a universal holdout in place and run the long list of ideas in simulation, not on the paying base.
Howes' ECHO stack (Profound plus in-house agents) briefs Bustle editors; Innisfree's visibility on 200 skincare prompts rose 30% Aug 11–Sep 30 while Bustle went from #709 to #12, and Pampers fragrance-free moved #2 to #1 after a Scary Mommy explainer. At least three brands have paid, Reddit paid still feeds citations, and Blyskal says the influence layer still does not net to the transaction layer.
Why it matters: If you are buying "AI visibility," demand prompt-level before/after and a named citation source, not a mention score.
Peec's Konitzny: in-answer clickable links (not sidebar citations) went from about 10% of chats to 47.8% by Nov 14, averaging four links instead of two. 63% of linked URLs are missing from that chat's cited sources; they are mostly brand homepage or product pages. Citation dashboards are counting a different object than the click.
Why it matters: Split GEO reporting into cited vs linked this week; a homepage link in the answer is not a citation win and will land as Direct.
Google's new help page, spotted by Hana Kobzová: URL Contains is token-not-substring, underscores are not separators, spaces are AND, and exclusions only fire when Text customization and Final URL expansion are both on. DSA rules go read-only after the upgrade. `/sale` does not block `/resale`.
Why it matters: Pull every AI Max exclusion list today and re-test underscored and partial-path rules before the next DSA-to-AI-Max migration.
Appointment-finding rolled out Nov 16: Ask Maps checks online availability and can call beauty, pet, auto and home-repair shops on the user's behalf. Google has not said how the business identifies the caller, whether the booking completes, or whether paid listings follow. Local operators just inherited inbound from an agent.
Why it matters: Local and franchise teams need a call-tagging rule for Google-originated appointment hunts this week, because those leads will not show as Maps clicks.
Glenn Gabe's FootJoy query showed a "local retailers" sponsored pack from DICK'S, Golf Galaxy, Golf Headquarters and Worldwide Golf under the answer, then Sponsored stores once you open the Shopping Graph citation; Juozas counted four paid slots in a single response, and there is still no AI Mode opt-out.
Why it matters: Pull one commercial AI Mode query this week and check whether the brand is in the answer text or only in a sponsored pack.
Google confirmed a logging error on the Search Generative AI performance report for Nov 9–11; real AI Mode and AI Overview impressions were not affected, but the chart dipped in the middle of the November core update.
Why it matters: Do not report a mid-week AI visibility crash until you exclude Nov 9–11 from the Generative AI report.
Google made Gemini 4 Hydrogen the default for every AI Mode user and for the hardest AI Overview questions on Nov 12, two days into the core update, and claims more query fan-out and more links without publishing counts.
Why it matters: Split every AI Mode and AI Overview citation pull at Nov 10 (core update) and Nov 12 (Hydrogen) or you will blame the wrong change.
Anthropic's rewritten usage policy went live today: using Claude to run fake-review networks or sites that pose as unaffiliated sources corroborating the same claims, for search or AI answers, is a named ban, and Anthropic can throttle on suspicion rather than proof.
Why it matters: If any GEO workflow uses Claude to generate corroborating pages or expired-domain lists, kill it today; the model is now the enforcement layer.
Natural Intelligence moved the channel from test to growth budget after sending conversion events back to the algorithm; Adthena's 30-brand Q3 pull says 52% still cannot recoup spend, and Kaiser/Schulze put US revenue per session at $0.89 vs $2.45 on Google Search. Twilio is dabbling, Intuit is experimenting, and Qualcomm's CMO says targeting barely exists unless the ask is super broad.
Why it matters: Keep ChatGPT Ads in a tagged test cell until conversion events are firing; do not copy a growth-budget move on click metrics.
{searchtermmatchtype} now returns e/p/b/a, and a is the AI Max expansion and keywordless match. Pair it with {matchtype} so a click that looks exact in the keyword column but a in search-term match is expansion traffic you can score in analytics, not a blended number inside Ads. Spotted by Hana Kobzová.
Why it matters: Add the parameter to every Search and AI Max tracking template this week and split landing-page conversion by a vs not-a before the next AI Max budget argument.
Global CCO Neil Reynolds stood up One Demand squads (brand, insights, KAM, category, agency) and is buying LLM visibility data so M&M's and Dove show up when a shopping agent builds the weekly run. Publicis now owns connected commerce, and the next retailer meeting is data-team to Walmart data-team.
Why it matters: CPG and retail teams need an LLM-source audit and a named commerce owner before the agent starts substituting the impulse SKU.
The AI-Era Attribution Blueprint, out this morning, splits AI-influenced (human still buys) from AI-recommended (the model chooses), then asks three questions at conversion: what created intent, who decided, who executed. A claim is eligibility plus evidence (referrals, seller records, surveys), not a 30% default.
Why it matters: Stop taking platform-reported ChatGPT or AI Mode conversions at face value; score each as influenced vs recommended and demand the evidence trail.
Supermetrics Campaign Manager, live Nov 10, writes to Google, Meta, LinkedIn, Microsoft, TikTok, Snap, and ChatGPT Ads from Claude or ChatGPT with per-account write switches, budget caps, and a change log. Premier Cru's Matthew Thompson used it on Google and Meta and is not taking more clients with the time; he is using it for answers the account never had.
Why it matters: If the team already chats performance in Claude, this is the write path to test with paused creates and a hard budget cap, not a new dashboard.