The week AI became a direct media seller,
a direct search surface, and a direct creative studio

Google used the first anniversary of AI Max to announce its most consequential structural change to paid search in years: Dynamic Search Ads will be automatically upgraded to AI Max beginning in September 2026, and the creation of new DSA campaigns will be blocked from the same date. The migration is mandatory, with no formal opt-out. Campaigns that are not voluntarily migrated before September will have AI Max features, including search term matching, text customization, and final URL expansion, enabled automatically using Google's default settings rather than advertiser-configured preferences. For brands whose ads currently run on DSA, voluntary early migration is now the only way to retain meaningful control over how the transition is configured.

The new anniversary features extend AI Max's utility across verticals. AI Brief, now rolling out for search campaigns in English, allows advertisers to write plain-language creative briefs that guide what the AI does and does not include in ad copy. Matching Guidelines let teams specify which queries to prioritize and which to avoid, addressing the perennial complaint that AI-driven targeting surfaces irrelevant queries alongside relevant ones. Travel-specific campaigns now unify text ads, travel promotion ads, and booking links under a single AI Max structure with consolidated reporting. For regulated industries, text disclaimers can now guarantee required disclosure language appears in responsive search ads even when AI Max's final URL expansion is active, a genuinely important compliance update for financial services and healthcare advertisers.

Google's internal data claims that campaigns using the full AI Max feature suite achieve an average of 7% more conversions or conversion value at similar CPA or ROAS compared to using search term matching alone. For campaigns that previously relied heavily on exact and phrase match keywords, the cited uplift is substantially higher. L'Oreal, referenced in Google's case study materials, reported a 2x higher conversion rate and 31% lower cost per conversion using AI Max to find net-new search queries. The performance data is directionally encouraging, but advertisers should note that AI Max amplifies whatever signals already exist in an account. Poorly structured accounts with weak creative assets and thin conversion data are unlikely to see the headline numbers.

The broader GEO context remains relevant for search teams alongside the paid search developments. Google AI Mode, which launched in May 2025 and now processes over one billion queries monthly, continues to draw user engagement that differs meaningfully from standard search behavior. Users spend an average of 49 seconds per session in AI Mode compared to 21 seconds in AI Overviews, and brand comparison queries average 77 seconds of engagement. Semrush research shows that AI Mode draws only 54% domain overlap with Google's organic top 10, meaning brands that rank well in traditional search are not automatically visible in AI Mode results. This gap between organic ranking and AI Mode visibility is the core challenge that GEO strategy needs to address alongside, not instead of, the technical AI Max account transition.

On May 5, OpenAI launched a beta self-serve Ads Manager for all US advertisers, opening ChatGPT advertising beyond the select group of large brand partners who entered the pilot in February. The platform is accessible at ads.openai.com after verification. The launch simultaneously introduced CPC bidding alongside the existing CPM model, with OpenAI suggesting starting bids between $3 and $5 per click. A Conversions API and pixel-based measurement tools confirmed their availability, giving performance marketers the attribution infrastructure they needed before allocating meaningful budget. The $50,000 minimum spend threshold that gated the original pilot has been dropped entirely. OpenAI has confirmed plans to expand testing into the UK, Brazil, Japan, South Korea, and Mexico in the coming weeks.

The creative format remains constrained: a small favicon with text, reflecting OpenAI's explicit commitment that ads cannot compromise user experience. The context hint system, which replaces keyword targeting, requires advertisers to write natural language descriptions of the conversations in which their ads should appear. That is a conceptually different skill from keyword research, closer to audience intent targeting than search term matching. Until measurement infrastructure matures, creative quality is the primary lever available to improve performance. Brands that enter this week will spend the next several months building the attribution and learning data that later entrants will have to pay a higher CPM to access.

Simultaneously, Omnicom's Q1 2026 earnings call contained the most significant agency statement on programmatic buying in years. CEO John Wren described ad tech intermediaries as extracting a "toll" from client budgets and confirmed the company is actively investing in shortening the media supply chain. Chief Technology Officer Paolo Yuvienco disclosed that Omnicom has already executed real client media buys using an agent-to-agent framework built around AdCP, the Ad Context Protocol, an open standard for advertising automation that allows AI agents to discover inventory, negotiate directly with publishers, and transact without requiring manual human steps at every stage. The holding company's Q1 revenue reached $6.2 billion, with organic growth of 3.9% across core operations.

The Trade Desk's Q1 2026 results, reported this week at $689 million in revenue representing 12% year-on-year growth, came with a set of partnership announcements that matter for media buyers. LinkedIn selected The Trade Desk as its first DSP partner for activation of B2B data for CTV advertising, enabling brands to target professional audiences in streaming environments using LinkedIn's firmographic data outside LinkedIn's own walled garden. DramaBox programmatic inventory went live globally through The Trade Desk, opening short-form drama content to CTV-style buying. AccuWeather, The Guardian, Hearst Magazines, Hearst TV, Newsweek, People Inc., and Ziff Davis joined as the first publishers to adopt OpenAds, The Trade Desk's transparent auction standard.

HubSpot's Spring 2026 Spotlight has now fully deployed, with AEO available in Marketing Hub Pro and Enterprise. The tool allows teams to track how often their brand appears in answers from ChatGPT, Perplexity, and Gemini, compare visibility against named competitors, and act on content recommendations without leaving HubSpot. That last point matters: GEO monitoring has until now required specialist third-party tools, which creates workflow friction and makes it harder for marketing teams to connect AI search visibility data to the content operations already managed in their CRM. HubSpot's decision to build AEO natively signals that AI search visibility is now a marketing operations problem, not a specialist SEO problem.

The accompanying Prospecting Agent has moved to outcome-based pricing: $1 per qualified lead recommended for outreach, replacing the previous monthly-per-contact model. HubSpot's Customer Agent, which now resolves 65% of conversations and has cut resolution time by 39% across 8,000 customers, is priced at $0.50 per resolved conversation, down from $1.00. The shift from seat-based to outcome-based pricing is a meaningful structural change in how martech platforms extract value from AI. Brands evaluating HubSpot renewals should model the per-resolution and per-lead costs against current seat pricing to understand total cost of ownership as agent usage scales with automation adoption.

Salesforce's Headless 360 announcement, which opens Salesforce data and workflows to AI agents through structured APIs rather than requiring user interface navigation, is the platform's clearest signal yet that the long-term architecture of enterprise martech is moving from screen-based interaction toward AI-orchestrated execution. In this model, marketing automation is not about campaign builders clicking through dashboards. It is about agents accessing data, triggering workflows, and executing tasks on behalf of marketing teams operating at a scale and speed that manual processes cannot match. The Agentforce positioning has matured from a product launch into an architectural commitment that will reshape how marketing operations teams are structured and what skills they require.

Braze this week added AI agent features that predict the optimal send time for emails and push notifications based on individual user behaviour and generate content variations for A/B testing automatically. Sprinklr's Spring 2026 platform update introduced automated governance tools and conversational bots for enterprise service teams. Clinch and OpenGlass partnered to deliver personalized ads to television screens when a viewer pauses content, creating a new CTV ad format that merges pause-state inventory with dynamic creative personalisation. Each of these updates represents a distinct workflow being partially automated, compressing the time between brief and execution while raising the baseline skill requirement for teams managing the inputs those agents require.

Google launched a meaningful but limited gesture toward publishers this week: AI Overviews and AI Mode responses will now display a "Subscribed" label on citations from publications the user subscribes to, and early testing showed users are significantly more likely to click through to labelled sources. This is a positive signal for subscription-based publishers whose content Google is summarising, but it does not address the fundamental referral traffic problem for ad-supported publishers whose entire business model depends on page views. For the many publishers without subscriber relationships to surface, the feature is largely irrelevant.

The structural context makes that gap more acute. People Inc., the publisher of People, Entertainment Weekly, and related titles, reported this week that Google referral traffic to its brands has fallen by 63% over the past two years. Despite that collapse, the company reported $103 million in digital revenue for Q1 2026, with non-website income representing 41% of total digital revenue and growing 24% year on year. The People Inc. numbers are a case study in publisher survival strategy: the brands that adapt fastest are not the ones that recover lost search traffic. They are the ones that build revenue streams, including licensing, events, direct commerce, and video, that do not depend on web visits to monetise.

Chartbeat data from more than 2,500 publisher sites continues to show AI chatbot referrals accounting for less than 1% of total pageviews, even as ChatGPT traffic has grown more than 200% year on year in absolute terms. The traffic that does arrive from AI platforms converts at meaningfully higher rates: reporting from The Washington Post found AI-referred visitors converted to subscriptions at 4 to 5 times the rate of traditional search visitors. That conversion quality gap is the most important metric for brands running content marketing programmes alongside paid media, because it changes how to value traffic from AI search surfaces. Volume is low but commercial intent is high, a profile that looks more like branded search than informational SEO.

Reuters Institute data published in January and circulated widely this week continues to define the planning context for publisher-adjacent media strategies. Chartbeat data from 2,500 sites shows Google search traffic globally down 33%, and 38% in the United States specifically, over the year to November. Media executives surveyed by Reuters expect a further 43% traffic decline over the next three years. Publishers are pivoting investment toward YouTube, where the platform's net priority score among news executives rose from 52 in 2025 to 74 in 2026, and away from traditional Google search, Facebook, and X. Brands that advertise within or alongside publisher content, particularly in ad-supported environments built on search-referred traffic, face a deteriorating inventory quality and volume picture over that same three-year horizon.

Meta opened its AI Business Assistant to all advertisers globally this week, simultaneously launching a zero-code Conversions API setup that removes the developer requirement from the attribution infrastructure that powers Advantage+ campaigns. The significance of the zero-code Conversions API is practical: the technical barrier that prevented smaller advertisers and brands without engineering resource from implementing server-side conversion tracking is now gone. Meta's AI handles the configuration. Any advertiser not running the Conversions API has lost the excuse of complexity. Without it, Advantage+ and other AI-optimised campaign types operate on signal-poor data that limits their ability to optimise toward outcomes rather than engagement proxies.

TikTok added ByteDance's Seedance 2.0 AI model to its Symphony ad creative suite, giving advertisers access to one of the most capable AI video generation models currently available embedded directly within TikTok's ad creation workflow. Symphony already offered AI-powered script generation, spokesperson avatars, and asset adaptation. Seedance 2.0's integration adds significantly more sophisticated video generation capability to that stack, with improved motion quality, prompt adherence, and temporal consistency compared to earlier models. For brands producing short-form video ads at scale, this reduces the per-unit creative production cost substantially without requiring brands to operate external AI video tools and then import assets.

LinkedIn expanded AI conversational search to all users worldwide, allowing members to use natural language queries to surface content, connections, and professional information across the platform. For B2B advertisers, this is relevant for two reasons. First, it changes how professionals discover content and vendors on LinkedIn, which affects what content formats and topics attract organic visibility. Second, LinkedIn is building the AI discovery infrastructure that will eventually inform how its advertising products surface brands in AI-generated professional context. Separately, LinkedIn announced that from June 22 onwards, going live on the platform will require a pre-scheduled Event attachment, changing how brands use live video for community building and product launches.

X Communities is confirmed to shut down on May 30, removing the platform's primary community-building tool. For any brand that invested in an X Communities presence for audience engagement, migration planning is now urgent. The practical destination for most brand communities from X is Discord, LinkedIn groups, branded newsletters, or owned platforms. YouTube's likeness detection tool, previously limited to entertainment industry figures, expanded this week to all creators and public figures. This has compliance implications for any brand running AI-generated video content that features real voices or faces: existing published content should be audited against the new detection scope to assess removal risk before the next campaign brief lands.

Adobe launched Firefly AI Assistant in public beta, introducing a conversational interface that orchestrates complex, multi-step creative workflows across Photoshop, Premiere, Lightroom, Express, Illustrator, and Firefly. The assistant allows creators to describe the outcome they want in plain language, with the AI handling the execution across multiple Creative Cloud applications. This is a conceptually different tool from the generative image and video features Firefly has offered previously. Those tools produce a single output from a prompt. Firefly AI Assistant coordinates a sequence of actions across multiple applications to reach a finished state, collapsing what was previously a multi-software, multi-hour production process into a directed conversation.

Firefly's model ecosystem now includes more than 30 industry AI models under a single subscription, including Google's Veo 3.1, Runway's Gen-4.5, Kling 3.0 and Kling 3.0 Omni, ElevenLabs' Multilingual v2, and Adobe's own commercially safe Firefly Image Model 5, which reached general availability this week. New Firefly Video Editor capabilities, including studio-quality sound generation, advanced colour adjustments with Color Mode, and Adobe Stock integration, extend the platform toward a complete short-form content production pipeline. Custom model training, now in public beta, allows brands to upload their visual assets and train a private model that preserves brand-specific colour palettes, stroke weights, character consistency, and photographic style across all subsequent generations.

The custom model feature is the most commercially significant update for brand marketing teams this week. Until now, brand consistency in AI-generated creative required significant prompt engineering and manual quality review. A brand-trained custom model produces outputs that reflect the brand's established visual identity without constant human intervention, making AI-generated creative viable for localisation at scale, campaign variant production, and always-on content pipelines that would previously require significant agency or in-house creative resource. Adobe's "do not train" policy means brand model data is not used in Adobe's own model development, addressing the IP concern that has made enterprise brands cautious about training proprietary creative systems on third-party platforms.

Adobe's AI Translation and Generative Dubbing feature, released in May 2026, allows video content to be adapted for multiple language markets with AI-generated voiceovers that match the original speaker's tone and pacing. For brands running regional or global campaigns, this reduces the production cost and timeline for localised video content substantially. Dubbing a 30-second brand video for five language markets previously required five recording sessions with native voice talent, review cycles, and post-production editing. Generative Dubbing compresses that to a fraction of the time and cost, with quality that is now commercially deployable for branded content if not yet for premium broadcast production.

The Trade Desk reported Q1 2026 revenue of $689 million, 12% higher than the same period last year, in results released Thursday. CEO Jeff Green cited strong performance from CTV, audio, and retail media alongside the strategic upgrades the company has been making, including the Ventura CTV ecosystem, the OpenAds transparency standard, and the LinkedIn B2B data partnership for CTV activation. The earnings were The Trade Desk's first major results release in a quarter shaped by the company's Trading Modes product announcement and the broader industry conversation about whether TTD's transparency model holds up as AI makes outcome-focused buying simpler for clients to demand.

TVision's acquisition by Viant was confirmed this week, combining Viant's programmatic platform with TVision's household-level CTV viewership data, attention signals, and engagement metrics. The deal gives Viant's advertiser clients AI-powered signals about which CTV inventory delivers genuine audience attention rather than passive exposure, a meaningful differentiation in a CTV market where measurement quality varies widely across publishers and platforms. For brands allocating significant CTV budget, Viant's enhanced attention data layer is now worth evaluating as a complement to the impression-based buying that currently dominates most CTV planning.

Google previewed Meridian Studio ahead of Google Marketing Live on May 20, describing an enterprise version of its Meridian media mix modeling platform alongside an updated open-source GeoX tool for measuring incrementality across geographic regions. The pre-announcement signals that measurement and attribution are likely to be central themes at GML 2026, not just product launches. For brands that have relied on platform-reported attribution from Google Ads, the introduction of first-party media mix modeling tools raises a practical question about whether in-house MMM capability should now be on the 2026 technology roadmap. Google Marketing Live 2026 is confirmed for May 20 and is expected to cover AI Mode advertising rollout, AI Max expansion, Performance Max reporting enhancements, and further YouTube commerce developments. Media teams should block the day.