Week in Digital Marketing 2026-10-22

The most important shift this week: a customer can choose a product in one place, buy it somewhere else, and have analytics attribute the sale to yet another channel. Last week, we examined the economics of AI intermediation. Now, PYMNTS research reveals a clear distinction: consumers are moving product search and comparison to assistants faster than they are abandoning familiar shopping platforms. A change in how people discover products does not automatically change who makes money from the transaction. For growth leaders, this is a warning: allocating budgets solely on attributed conversions can reward capturing existing demand at the expense of activities that actually change customer decisions. My argument: competitive advantage will come from factoring these distinctions into business decisions, not simply increasing visibility in another channel.

TL;DR

• AI is changing where the buying journey starts faster than where purchases happen. In the PYMNTS study, 23.2% of US consumers identify AI as their primary starting point for online shopping research. Meanwhile, 59.7% of respondents describing their most recent AI-assisted purchase completed it on Amazon. These percentages refer to different groups, not successive stages of a single funnel.

• Direct audience relationships require organizational changes. Digiday describes how The Verge is replacing general audience growth roles with specializations in newsletters, community, and audience analytics. The point is clear ownership of relationship development, not simply having owned channels.

• Proximity to the transaction does not prove incrementality. A SalesTechStar analysis reminds us that attributing a sale to a retail media ad does not establish whether the campaign caused the purchase. I recommend increasing investment only when measurement shows incremental impact against a credible no-advertising baseline.

• Human attention is becoming the bottleneck in automation. At Anthropic, around 80% of the duration of interactive agent sessions is spent waiting for user input. This is a case study of one company using AI extensively, but it sends an important signal to those designing RevOps processes.

• More efficient creative production is not enough without faster decisions. In a Forrester Consulting study commissioned by Smartly, only 25% of respondents say they can produce a CTV ad variant within two days. Before increasing budgets, it is worth shortening approval workflows and using measurement findings to inform subsequent creative changes.

• GEO requires evaluating recommendation quality, not just counting mentions. Delante’s BrightonSEO recap highlights the importance of entire conversations and the arguments that determine whether a brand is recommended or rejected. I suggest checking which customer requirements cause a brand to drop out of recommendations, rather than reporting only its presence in answers.

Summary of Changes

The first shift concerns the buying journey. The PYMNTS study suggests that an assistant may help build a product shortlist, while the customer completes the purchase with a familiar retailer because of price, delivery, and confidence in payment security. However, self-reported data does not prove that AI caused Amazon’s sales to grow. The Reuters Institute, meanwhile, describes publishers that view inclusion in AI answers as a way to build awareness and authority, rather than a direct replacement for revenue from website traffic. My takeaway: we need to separate influence on customer choice from the ability to turn that influence into revenue.

Last week, we considered the value of registrations; now, The Verge’s example shows what that approach means for team structure. Specialists with distinct growth goals are responsible for newsletters, community, and analytics. According to its director of audience and subscriptions, more than half of new subscribers arrive directly at the sign-up page, although that share varies from month to month. This should not be taken to mean that earlier brand interactions did not matter. I recommend assigning teams goals for progression between relationship stages, without treating the final visit as a complete explanation of customer acquisition.

The second shift exposes a measurement problem shared by AI search, CTV, and retail media: influence on a decision may occur before the conversion visible in analytics. PPC Land cites an analysis of Similarweb’s panel of US desktop users. In this sample, only 8.8% of visits linked to an earlier AI recommendation came from a direct assistant referral, while 55.9% came from search engines. This is the finding of a specific study, not a universal multiplier for estimating your own AI-assisted traffic. In the study cited by Smartly, 62% of respondents say they cannot measure CTV’s indirect impact on search and social media. SalesTechStar describes the opposite side of the problem: a sponsored product ad can receive credit for a purchase that would have happened without it.

I see a risk here: automated budget allocation can reinforce the channels that are easiest to measure while cutting funding for activities that build future demand. I suggest combining attribution reports with incrementality tests and marketing mix modeling, accounting for each method’s limitations. The aim is not to find one metric that explains everything, but to check whether different methods lead to consistent budget decisions.

The third shift concerns an organization’s ability to make decisions. Last week, we asked when an agent should pause; now, Anthropic’s research shows the burden on the human side. For the average employee, the share of interactive sessions devoted to orchestrating agents and similar tasks rose from 10% in fall 2025 to 22% in spring 2026. The authors note that tool use was not random, and the company’s rapid growth makes it difficult to separate AI’s effects from business changes.

Osnat Lidor describes a related mechanism in B2B: additional tools increase administrative overhead when information about customer needs and interaction history is lost during lead handoffs between teams. Yoni Tserruya, co-founder and CEO of Lusha, also highlights data freshness: information about people, roles, and job changes must keep pace with sales activity. My recommendation for RevOps: measure the time from a verified signal to a decision that enables action, not the number of sequences launched. In practice, this means clear ownership of exception handling, up-to-date decision documentation, and fewer unnecessary escalations—not another layer of notifications.

Patterns of Change

In recent weeks, I have seen attention shift from competing for audience access to identifying where value is actually created. Last week’s discussion of platform intermediation now leads to questions of accountability: who is responsible for changing preferences, who for developing the relationship, and who for monetizing it? The Verge’s example shows that the answer may require restructuring teams and responsibilities, not just changing reporting.

In my view, this will increase pressure for shared management of the customer experience across channel boundaries. I recommend that AEO and GEO programs involve not just someone responsible for visibility, but also someone accountable for the offering and commercial results. Otherwise, presence in AI answers may grow while the reasons customers ultimately buy elsewhere remain unchanged.

The second pattern is the growing importance of choosing which relationships to invest in, which recommendations to improve, and which decisions genuinely require a human. Articles on B2B trust and AI-assisted creator selection point to the need to distinguish mere presence from evidence of expertise and influence on choice. Anthropic’s research reveals a similar problem inside organizations: more agent-generated content and outputs mean additional work to organize and oversee them.

That is why I prioritize investments that shorten the path from reliable information to a decision. In content marketing, I suggest supplementing lead measurement with an assessment of changes in preference. In GEO, it is worth testing recommendation criteria; in automation, reducing the decision-making burden. This is a better foundation for scaling than simply increasing the number of publications, ad variants, and dashboards.

Topic Groups

Product Discovery and Purchase Completion

• From Prompt to Purchase: Who Wins the AI Shopper?

  This study of 2,191 US consumers separates changes in how people search for products from changes in where they buy. For growth strategy, the key is to build presence in AI recommendations while keeping the offering and purchase terms competitive.

• Agentic shopping and the end of the ad-as-interruption

  Adswerve argues that AI-assisted shopping does not necessarily mean a simpler customer journey. This vendor-authored piece highlights the need to connect product data, media campaigns, and measurement, but does not establish that purchases completed directly in assistant conversations are widespread.

• Explaining AI referral traffic

  This overview explains why direct assistant referrals reveal only part of AI’s influence on traffic and sales. Its comparison of divergent conversion research findings cautions against assigning a single, universal value to this traffic.

Audience Relationships, Trust, and Owned Channels

• Life after SEO: How newsrooms are adapting to the AI search shift

  The Reuters Institute’s conversations with publishing representatives suggest that AI citations build awareness and authority, but do not directly replace revenue-generating traffic. Decisions about allowing bots to access content should take into account its uniqueness and the publisher’s business model.

• The Verge switches up its audience team to prepare for Google Zero

  Digiday describes how The Verge divides its audience team’s work across newsletters, community engagement, and audience analytics. It is an example of translating a direct-relationship strategy into specific roles, responsibilities, and operational goals.

• B2B leaders in APAC see an influence gap as brands prioritise content over trust: study

  This coverage of AIM B2B’s research highlights the gap between sharing information and influencing decision-makers’ preferences. It makes the case for supplementing lead measurement with assessments of changes in brand perception and relationship quality, while accounting for the study’s regional scope.

• AI Influencer Discovery: How 1 in 3 Brands Choose Creators

  5WPR’s analysis describes the use of AI to find and vet creators, and the importance of available evidence of expertise. It also distinguishes specialized platforms from general-purpose assistants. The extent to which the latter are used to select partners remains less well documented.

RevOps, Data Freshness, and Decision-Making Efficiency

• Build “Pain Backward,” Not “Tool Forward:” How B2B Leaders Stop Tech-Driven Revenue Drag

  Osnat Lidor suggests identifying where customer information gets lost and testing the process without a new tool before investing. Technology’s value should be judged by whether it reduces administrative work and improves collaboration between teams, not by its feature count.

• What Good Looks Like in B2B Sales Data in 2026

  Lusha’s CEO uses his company’s data to demonstrate the importance of up-to-date contacts, access to buying decision participants, and rapid responses to signals. The proposed benchmarks reflect a vendor’s perspective, not an independent standard for all B2B teams.

• How employment is changing at Anthropic

  This internal study describes a shift in work toward managing agents and organizing their outputs. For marketing, it offers a useful case study of automation’s organizational limitations, but it is neither a universal employment forecast nor an experiment that can definitively establish AI’s impact.

Incrementality, Retail Media, and Creative Production Workflows

• CMO Round Table on MENA retail media: From shelf space to measurable media

  This discussion among industry representatives from the Middle East and North Africa reveals the obstacles to moving from negotiating retail placement to planning a measurable advertising channel. Clear budget ownership and access to data while campaigns are still running are just as important as ad inventory.

• Retail Media’s Next Problem Isn’t Growth. It’s Proving Incrementality

  This article separates attributing sales to advertising from demonstrating a campaign’s incremental impact. It helps define requirements for retail media partners: transparent measurement methods, the ability to run experiments, and consideration of other channels’ influence.

• Smartly’s Oli Marlow-Thomas on Why CTV Needs a Creative Upgrade: The B2BMX Insights Q&A

  Drawing on research commissioned by Smartly, this interview highlights delays in producing, approving, and evaluating CTV creative. The author identifies two ways to improve workflows: creative built from interchangeable components, and joint performance analysis by creative, media, and analytics teams.

GEO: Recommendation Quality Rather Than Mere Presence

• BrightonSEO October 2026 Recap

  Delante’s recap illustrates the shift from monitoring individual mentions to analyzing conversations, purchase intent, and the arguments that determine recommendations. Speakers’ experiments are best treated as sources of hypotheses for your own testing, not verified rules governing every model.

• Generative Engine Optimization (GEO): What It Is & How to Do It

  This guide covers consistency in brand information, content accessibility, and presence beyond a brand’s own domain, while accounting for the instability of AI-generated results. The variability of cited sources described here supports repeated measurement, but does not in itself prove GEO’s impact on revenue.