top of page
AI processor with icons for security, growth, and marketing

7 eye-catchers in display advertising, i.e. how to create graphics that will catch the customer’s eye

Week in Digital Marketing 2026-10-15

Oct 15
7 min read

The most important shift this week is not about what AI agents can do, but who profits from their role as intermediaries. Last week, we examined how platforms shape and monetize purchasing decisions. Now it is clear that generative AI is also stepping between brands and users who are already searching for them: Ahrefs research shows how often AI Overviews appear in branded search results. Investing in brand awareness therefore does not guarantee direct contact with your audience, even when someone is looking for a specific company. (ahrefs.com)

Growth leaders therefore need to assess customer acquisition effectiveness, the profitability of direct relationships, and actual automation savings separately. My thesis this week: the advantage will go not to organizations that delegate the most tasks, but to those that can demonstrate where AI intermediation creates value—and where it means handing margin, attention, or control to platforms.


TL;DR


  • A branded search does not guarantee a visit to the brand’s website. AEO and GEO should cover not just inclusion in recommendations, but also the accuracy of information presented to people who already know the company. (ahrefs.com)

  • An owned audience needs a steady flow of new users. INMA’s analysis challenges the strategy of relying on retention improvements alone to offset losses from shrinking reach. Funnel optimization buys time, but it does not replace audience acquisition. (inma.org)

  • Registration can deliver value independently of subscriptions. Registered users can generate more advertising revenue and enable more precise segmentation. But you need to measure results across the entire cohort, not just the registration rate. (digiday.com)

  • Agent deployment plans are outpacing implementation. According to research cited by MarTech, 95.3% of surveyed organizations have AI agents on their roadmaps, but only 10.8% of initiatives have reached full-scale production. That is an argument for phased investment, not for funding promises. (martech.org)

  • A good agent must know when to withhold a recommendation. Incomplete conversion data, recent campaign changes, and incompatible definitions of results should prevent action rather than trigger another budget adjustment. (supermetrics.com)

  • Infrastructure control comes at a cost, too. Interest in Trusted Server following Safari’s blocking measures shows that greater control over data requires technical expertise, reliable measurement, and advertiser acceptance. (digiday.com)


Summary of Changes


The first shift concerns the scope of GEO: it is no longer just about building visibility among people discovering a brand, but also reaching those already searching for it. Ahrefs analyzed 232.8 million observations of US desktop Google search results from July through September 2026. In the September 29 snapshot, AI Overviews appeared in 82.91% of the branded search results studied; the September average was 73.3%. In a separate sample of searches for 100 brand names, most of the summaries displayed occupied the top position. These findings describe specific samples, not the share of all user searches. (ahrefs.com)

In my view, more is at stake than a lost click: third-party publications can help shape the answer even when someone is researching a specific company rather than comparing options within a category. Instead of focusing on increasing citation counts, I suggest checking whether answers accurately reflect the current product, pricing, target audience, and limitations of the offering. Source citations, brand mentions, and purchase recommendations need to be measured separately. Otherwise, visibility can easily become another metric that looks good but says little about results. (ahrefs.com)

The second shift concerns the approach to owned channels. Registration does not have to be merely a step toward payment: the publishers covered by Digiday also use registration data to segment audiences and increase the value of their advertising offerings. That does not mean, however, that every registration requirement improves business performance, or that subscriptions automatically offset losses from declining traffic. (digiday.com)

INMA adds an important warning: a loyal audience is no substitute for a pipeline of future customers. In a model of a typical publisher, improving acquisition to upper-quartile performance delivers greater subscription revenue growth in year three than an equivalent improvement in retention and pricing. This is the result of a modeled scenario, not a universal forecast. (inma.org)

I therefore recommend funding audience acquisition and the monetization of direct audience relationships in parallel. Registration requirements should be assessed based on how they change results across the entire test cohort. The calculation needs to account for lost visits, registrations, consents obtained, subsequent return visits, and margin. Higher revenue per registered user is not enough if we ignore the cost of losing the rest of the audience.

The third shift is from broad calls for AI oversight to specific conditions that must be met before acting on a recommendation. In Alexander Group research covered by Demand Gen Report, 96% of respondents report using AI, but only 41% can demonstrate ROI. A separate study cited by MarTech shows that only a small share of agent initiatives have reached full-scale deployment. These survey populations should not be combined, but both findings challenge the assumption that deployment itself equates to financial impact. (demandgenreport.com)

Supermetrics explains what this oversight requires in practice: an agent needs current settings, stable campaign identifiers, a change history, and data that accounts for conversion lag. It should also be able to withhold a recommendation when data is insufficient or too little time has passed since the last change. (supermetrics.com)

Sponsored content from PropellerAds also highlights security: signals related to malware, cloaking, and suspicious redirects should inform media-buying decisions alongside CPA and ROAS. This is a vendor’s position, not an independent standard. It does, however, expose a significant gap: a spending cap is not enough if an agent scales purchases of seemingly cheap but risky traffic. (searchengineland.com)


Patterns of Change


These articles reveal a tension: a brand needs independent sources to substantiate information about it, while also building direct relationships to monetize audience interest. That is my interpretation of the AI visibility analyses alongside publishers’ experiences, not proof of a single universal growth model. (ahrefs.com)

I suggest we stop treating earned media and owned channels as competing budget lines. They should complement each other: the former provides external sources of information about the brand, while the latter helps build direct relationships and encourage return visits. Alongside the brand’s share of AI answers, it is therefore worth measuring the rate of acquisition of new audience groups and their subsequent contribution to revenue. Protecting the existing audience must not mean giving up on future customers.

The second pattern is the shift in the cost of control from platforms to organizations. Trusted Server gives publishers more control over data processing, but publishers must implement it, validate measurement, and secure buyers’ acceptance of the results. With AI agents, meanwhile, automation savings may be reduced by the cost of checking outputs and handling exceptions. (digiday.com)

When planning investments, I therefore suggest calculating each project’s benefits after subtracting integration, oversight, and correction costs, while accounting for the risk of error. The conditions under which an agent will receive greater autonomy also need to be established upfront. The most urgent task is not to deploy another agent, but to determine which decisions are actually worth entrusting to one.


Topic Groups


Branded Demand and the Accuracy of AI Answers



Independent Sources and Brand Reputation


  • How AI Answer Engines Cite Brands in 2026This agency article links a brand’s AI visibility to consistency in how it is described and its presence in independent sources. The correlations cited do not establish causation. The article is best treated as a basis for hypotheses and an audit, not a guaranteed formula for earning recommendations. (actualagency.ai)

  • Why Should CMOs Treat Earned Media as a Data Play?The author presents earned media as a source of information that influences AI answers before contact with sales. The cited figure of 84% of citations coming from “earned sources” should not automatically be equated with the impact of a specific brand’s PR activity. (cmswire.com)


Audience Acquisition, First-Party Data, and Monetization



AI Agent Profitability, Validation, and Security


hello@1000i.pl

+48 503 04 1000

bottom of page