Shoppers who use AI on the way to a purchase interact with or notice 12.1 brand touchpoints. Shoppers who do not use AI: 4.2. That is Google's opening number for the holiday season, from a Google-commissioned Ipsos survey run in December 2025 across 25 countries.
Think with Google's sentence: “Online shoppers who use AI platforms or features in their purchase journey interact with or notice 12.1 touchpoints, on average, compared to only 4.2 for shoppers who did not use AI.”
Two things to keep beside it. It is a global average, not a US figure. And it compares people who chose to use AI with people who did not, so it describes a difference, not a cause.
That figure was the backdrop for Rethink 2026, Google's free virtual advertiser event, held in two sessions: Rethink Retail on September 16, 2026 and Rethink ROI on September 17. The pitch was YouTube as a Q4 sales engine. Below are the 6 takeaways I kept, each with its source, what it means for an account, and one thing to try this quarter. Google's claims are presented as Google's. Where the reasoning is mine, it says so.
1. Shoppers are relying on AI to decide
Date and source: September 16, 2026. Courtney Rose, VP of Retail at Google, in the Rethink Retail keynote.
The takeaway. Google calls this the first AI-powered holiday season. Rose, from the stage: “But the shift is how those shoppers will find you because this is the first truly AI powered holiday season. And shoppers aren't just trying AI, they are relying on it.” The punctuation is YouTube's transcript, not hers.
The published companion, from the same Ipsos survey: “3 in 4 users say Google AI Mode and/or AI Overviews help them make faster and more confident decisions.” AI Mode is the chat-style view of Google Search; AI Overviews are the AI-written summaries at the top of results. The 3 in 4 is among people who used those two features.
What it means for your account. Google's framing: speed and confidence used to be a trade-off, and AI removes it. My read: the AI surfaces are now a place a buyer meets your product before your ad does, and most accounts have never looked at what they say.
Try this. Run your 10 biggest revenue searches through AI Mode and AI Overviews. Note where you are absent, and where the product details are thin or wrong. It costs 30 minutes and no budget.
2. Every brand dollar cut costs $1.92 to win back
Date and source: December 12, 2025. BCG, Building Lasting Brand Equity in the Age of AI. Google cites it as the reason brand versus performance is a false choice.
The takeaway. BCG's sentence: “Cutting $1 of brand spending today costs companies $1.92 in future investment to regain lost share, up from $1.85 in 2022.” The trend is part of the claim: BCG says the penalty is growing.
Scope, briefly. The $1.92 is BCG's modeled estimate from one analysis, with no confidence interval and no stated category or country. BCG's article never mentions YouTube; placing them side by side was Google's choice. And Google's own Rethink deck lists a BCG/Google consumer survey among the inputs, so the figure is not fully independent of Google.
What it means for your account. My read: this is the number to bring to the meeting where Q4 trims brand to feed performance. It turns a cost line into a deferred liability.
Try this. Take what you cut from brand in the last 12 months, multiply by 1.92, and show finance the result. Label it as BCG's modeled ratio, because that is what it is.
3. YouTube's case: brand spend that also performs
Date and source: Google's analysis of a Circana meta-analysis covering 2022 to 2024, published by Google on Think with Google in April 2026 and again at Rethink.
The takeaway. Google's sentence: “It's where you can build long-term brand value, drive durable growth, and achieve unmatched results, including an 86% higher incremental long-term ROAS than paid social.” ROAS is return on ad spend. Incremental means the sales that would not have happened without the ads. The footnote says the figure comes from “40 MMM models run for 10 brands across five categories between 2022–2024.” MMM is a marketing mix model, a statistical model that estimates what each channel contributed to sales from spend and sales history. So the base is 10 US brands in 5 unnamed categories, from before the AI shopping behavior the rest of the event was about. Circana supplied the models; Google selected and presented the analysis.
What it means for your account. Google's claim, Google's commissioned study. My read: 10 brands is a small base, so this is a reason to test, not a forecast for your account.
Try this. Run one geo holdout on YouTube this quarter. A geo holdout runs the ads in some regions, withholds them in matched regions, and compares sales. It costs real money, the media in the test regions plus any sales you forgo in the holdout regions, so size it small and agree how you will read it before it starts. Whatever number comes back is yours.
4. Connected data is what the bidding runs on
Date and source: September 10, 2026, Google Ads blog, and Google's Performance Four page.
The takeaway. Google puts data strength first among its 4 pillars for Demand Gen, its campaign type that runs image and video ads across YouTube (including Shorts), Discover, Gmail, Maps and the Google Display Network. Google's page: “Data strength is the foundation for driving performance with Demand Gen. The campaign is only as powerful as the quality of the data fueling it.”
The numbers Google attaches: “Advertisers who build their data strength with Google tag gateway observe on average a 14% conversion uplift, and over 20% uplift for their Demand Gen campaigns.” Google tag gateway is a setup where the Google tag loads through your own site's hosting or CDN provider rather than straight from Google. Both figures are Google's internal data, a before-and-after observation rather than a controlled test.
What it means for your account. Google's own uplift figures. My read: without CRM stages and offline sales connected, the algorithm optimizes to form fills, not revenue, because form fills are all it can see.
Try this. Connect offline sales and CRM stages before you scale Demand Gen. Sharing customer data with Google needs a sign-off from your privacy and legal people, so get that first. Then watch the new Data Strength Uplift metric in Google Ads, which Google says counts the additional conversions recovered by your first-party data setup.
5. Plan YouTube as a share of growth, not a line item
Date and source: September 16, 2026. Khartoon Weiss, VP, US Mid-Market Sales, Commerce at Google, in the Rethink Retail keynote.
The takeaway. The reframe of the day, in her spoken words: “Or can we actually think about asking how much of my growth should YouTube be responsible for. It doesn't deserve another line on your media plan any longer. With all this technology and power. It deserves a pedestal in your growth plan.”
One third-party figure sits near this, on Google's Performance Four page: “According to a report from Measured in the US, 46% of Demand Gen conversions came from net-new customers, outperforming the paid social average.” The paid social average is not quantified.
What it means for your account. My read: a growth share forces a measurable target, such as the percentage of net-new customers, instead of a budget percentage. A budget share is an input. A customer share is an outcome you can miss.
Try this. Set a net-new customer share for YouTube for Q4 and report against it weekly. Check it against your own new-versus-returning split before you check it against Measured's 46 percent.
6. Demand Gen is where the YouTube push lands
Date and source: 2026, Google Ads Help. The Display migration tool began rolling out in June 2026.
The takeaway. Everything above lands in one campaign type. Google's definition: “Demand Gen campaigns are Ad campaigns that capture engagement and action across YouTube (including Shorts), Discover, Gmail, Maps, and the Google Display Network.” Maps is in beta. Google is also moving Display campaigns into Demand Gen: “June 2026: Google is launching a phased rollout of our migration tool.” The remaining campaigns will be migrated automatically on a date Google has not given. And: “Note: Campaigns that are migrated to Demand Gen can't be reverted back to Google Display Ads campaigns.”
What it means for your account. The Display migration is a one-way door. Per Google's feature table, Manual CPC, Viewable Impressions bidding, pay for conversions, seasonality adjustments and portfolio bidding do not carry over, and HTML5 display ads are listed as coming in late 2026. Display inventory itself is unchanged.
Try this. Filter your campaigns by type Display and audit each one for settings that will not survive before you accept a migration. If you migrate, use Google's tool rather than waiting; Google says it carries 42 days of performance history across. There is no way back, so do the audit first.
Test as you implement
Google's numbers are Google's. Your account is yours. Try each one small before it gets a budget.
- Run one YouTube geo holdout and judge it in your own numbers, not the slide.
- Connect offline sales and CRM before scaling Demand Gen, then read the Data Strength Uplift metric.
- Set a net-new customer share for YouTube and review it every week.
None of these 3 needs Google's numbers to be right. They only need yours to be readable.
Sources: verified September 21, 2026 against the Rethink 2026 watch hub, the Rethink Retail keynote transcript, 2 Think with Google articles, BCG's December 2025 publication, Google's Performance Four article, the Google Ads blog of September 10, 2026, and 2 Google Ads Help pages. Quotations are exact; the keynote quotations follow YouTube's transcript. Lines marked my read are mine, not Google's.
Working out where AI actually fits in your marketing?
I write these while building the systems behind them: measurement, creative pipelines, and agents that do real work. Connect on LinkedIn and tell me what you are working on. That is where these conversations start.
Connect on LinkedIn