Adobe absorbs Semrush, Google retires Dynamic Search Ads for AI Max, Shopify's Agentic Storefronts ship into ChatGPT, Perplexity and Copilot, and the Pentagon picks OpenAI, Microsoft, Google, AWS and Nvidia over Anthropic

Adobe has closed its $1.9bn all-cash acquisition of Semrush. The pitch is no longer "SEO software" — it is a single platform for tracking and influencing brand visibility across Google, ChatGPT, Perplexity and Copilot. Marketers now have a mainstream, enterprise-grade vendor for what Adobe is calling generative engine optimisation and agentic search optimisation.

Google has confirmed Dynamic Search Ads, automatically created assets and campaign-level broad match will be auto-upgraded to AI Max by the end of September. Voluntary upgrade tools have shipped this week. The keyword-only Google Ads account is in run-off.

Shopify's Winter '26 Edition has gone live. Sidekick is now a proactive operator that can build apps, draft Flow automations and generate ShopifyQL reports from a prompt. Agentic Storefronts pushes a merchant's catalogue into ChatGPT, Perplexity and Copilot via a single Shopify Catalog hook.

The Pentagon awarded classified AI contracts to eight firms — OpenAI, Google, Microsoft, AWS, Nvidia, SpaceX, Reflection and Oracle — and conspicuously left Anthropic outside the tent over its refusal to drop warfare and mass-surveillance restrictions.

On the watchlist: Google security research showing indirect prompt injection in the wild rising 32% between November and February, and a Fortune analysis arguing that roughly half of Google's and Amazon's "blowout" AI quarter is paper gains on their Anthropic stake rather than operating revenue.

The pattern across the day is that the AI stack has finished arguing with itself about whether AI is a real channel. It is now repricing the channels we already use — search, ads, ecommerce, procurement — and forcing operators to decide whether they are upgrading on their terms or on the platform's.

Transcript

SPEAKER_00 0:01

Welcome to the Daily Marketing Brief. Your daily AI news and tactics for marketers who move fast. I'm your host, Jen Bryan, and here's today's update. The interesting thing about today's news is what isn't being argued about anymore. Nobody is debating whether AI shows up in search. Nobody is debating whether agents can touch an ad account. Nobody is debating whether shoppers will start a purchase inside a chat window. The argument has moved on. The argument now is who controls the upgrade path, who pays for the migration, and which line on the marketing PL the cost lands on. That is a much more commercial conversation, and it is one this episode is all about. Today we cover four stories that all push in the same direction. Adobe completes its SEMrush deal and builds itself as a brand visibility vendor for the AI search era. Google sets a hard deadline for the end of dynamic search ads and quietly retires the keyword as a unit of control. Shopify ships Winter26 with a genuinely more capable sidekick and a feature that drops merchant catalogues straight into ChatGBT, Perplexity and Copilot. And the Pentagon hands out classified AI contracts to eight tech firms while pointly leaving Anthropic out. We then look at two watch list items, indirect prompt injection going from theoretical to observed, and a counter narrative on whose AI earnings are actually real. So to kick off, Adobe completes its 1.9 billion SEMrush acquisition. So what happened? Adobe announced on 28th of April that it had completed the acquisition of SEMrush holdings for $12 a share, an all-cash deal worth roughly $1.9 billion in equity value. Deal was first agreed in November and has now closed. Adobe is folding SEMrush into Adobe CX Enterprise, the agentic AI customer experience platform it has been assembling around content supply chain, customer engagement, and brand visibility. So what is confirmed? The price, the close date, and the integration plan are all confirmed via Adobe's newsroom and SEMrush's investor disclosures. Adobe explicitly frames the combined offer as covering search engine optimization, generative engine optimization, meaning visibility inside large language model answers, and what it is now branding agentic search optimization, meaning visibility to AI agents browsing on a user's behalf. Adobe also cites third-party data showing AI-driven traffic to US retail sites is up 269% year on year as of March, which is the demand side reason it pays the premium. So why does this matter? Until last week, the tool set for tracking how a brand showed up across Google, ChatGPT, Perplexity, and Copilot was a fragmented mix of legacy SEO suites, scrappy LLM visibility startups, and home-built dashboards. Adobe has now bought the most recognizable mainstream brand in that category and bolted it onto the same procurement contract as Experience Cloud, Workfront, and Gen Studio. That changes the buying conversation in any enterprise that already runs Adobe. So what does that mean for business? Well CMOs and procurement leads will start receiving very persuasive bundle pricing in the next two quarters. Expect Adobe to push GEO and ASO into the CX Enterprise contract at marginal cost, which will squeeze pure play AI visibility vendors like Profound, Athena, and other smaller GEO startups. Some will be required, some will pivot to agencies, a few will survive on debt. So what does that mean for marketers and agencies? Well, two practical shifts. First, the SEO line item is being absorbed into a wider brand visibility budget that includes LLM citations and agent referrals. Agencies that still sell SEO retainers need to relabel and reframe before clients do. Second, reporting standards will start to converge on whatever Adobe ships in the next two release cycles because that is what enterprise clients will be looking at. If you want to influence the format of the conversation, do it now before the dashboards harden. So my take on this, this is sensible, slightly defensible deal from Adobe. SEMrush gives them a credible answer to the most awkward question in any 2026 marketing review. How do we show up in ChatGPT without having to build the crawler stack from scratch? The risk is integration drag. Adobe has a long list of acquisitions being digested and the brand visibility category is moving faster than Adobe's release cycle. Confidence level, high on the deal mechanics, medium on whether the integrated product ships at the speed the category demands. Google has confirmed that three legacy search configurations, dynamic search ads, automatically created assets at the campaign level broad match setting will be auto-upgraded to AI Max by the end of September. Voluntary upgrade tools landed in Google Ads Editor and the API this week. From September, advertisers will no longer be able to create new DSA ads through any interface. So what is confirmed? Google has confirmed officially on their blog post on the AI Max upgrade and the Google Ads announcement feed. Google's headline performance claim is an average of 7% more conversions or conversion value at a similar CPA or ROAS when AI Max is run with the full feature set. Search term matching, text customization, and final URL expansion versus search term matching alone. That is Google's own number on Google's own data, so treat it as a directional ceiling rather than a guarantee. So why does this matter? DSA has been the workhorse for e-commerce lead generation and large catalog advertisers for more than a decade. It is the closest thing remaining in Google ads to dynamic light keyword campaign type that smart operators could still steer. AI Max replaces that with a model-driven matching layer that decides which queries, which landing pages, and ad copy combinations to serve. The control surface narrows, the diagnostic surface narrows further. What that means for business, well, anyone running DSA heavy budget, large retailers, marketplaces, classifieds programmatic SEO sites, lead gen aggregators has a forced migration on the calendar. The risk is not the upgrade itself. The risk is that historical query level data, negative keyword libraries, and landing page mappings get lost or misapplied in the move. Plan the migration like a replatforming, not a settings change. So what does it mean for marketers and agencies? Well, three practical things. One, audit which campaigns will auto-upgrade and run them through the voluntary tool now while you still have the chance to review that mapping. Two, refresh your negative keyword strategy on the assumption that AI Max will pick up adjacent intents you previously excluded. Three, accept that traditional search query reports become a far less useful diagnostic and start budgeting for incrementality testing or geo holdouts as your real read on performance. If you sell PPC retainers, this is the moment to reframe the offer around testing design and creative not keyword management. My take, the 7% lift number is plausible, but it is also the floor of Google's argument, not the ceiling of yours. AI Max will probably win modestly on the aggregate. It will lose for advertisers with weak feeds, weak landing pages, or sensitive brand queries. The bigger story is that Google has finished the keyword ERA. From September, the unit of control is the model and the eligibility signals. Confidence level. High on the timeline, medium on the performance claim, low on the assumption that any individual account will see the average. Next up, Shopify Winter 26. Sidekick goes proactive and Agentix Storefront ships into ChatGPT Perplexity and Copilot. So what happened? Shopify's Winter 26 edition is live with more than 150 product updates. The headline items are an upgraded sidekick that moves from chat assistant to a more autonomous operator, a new feature called Agentix Storefronts that pushes a merchant's product catalog into AI assistants and SimGim, an AI Shopper sandbox that lets merchants test the journey before real customers see it. So what is confirmed? Sidekick can now build custom Shopify admin apps from a prompt, generate flow workflows in plain English, surface proactive recommendations through Sidekick Pulse, and produce Shopify QL reports without manual querying. Agentic Storefronts uses the Shopify catalog as the syndication layer to expose products to ChatGPT, Perplexity, and Microsoft Copilot with more partners flagged as coming. SimGym is in an AI research preview. So why this matters? Two distinct shifts. The Sidekick upgrade pushes more of the merchant operations layer into a natural language prompts. Junior StoreOps time was the bottleneck in most small to mid Shopify operations. That bottleneck is now visibly shorter. Agentix Storefronts is more strategic. It is Shopify's bet that buy eventually moves into the conversation surface and that the merchants who appear in those conversations will compound an early mover advantage in product discovery. So what does that mean for business? Well, for D2C brands and Shopify Plus retailers, this is a low-effort feed side change with potentially material upside if conversational commerce develops. The work needed is largely catalog hygiene, structured data quality, and review-rich product detail pages. The same hygiene that compounds across SEO and paid social anyway. The cost of opting in is low. The cost of being absent if conversational commerce scales is harder to recover. So what it means for marketers and agencies. Well, e-commerce agencies should treat agentix storefronts as a diagnostic before treating it as a channel. Run the audit. Are product speeds clean? Are descriptions accurate enough for an LLM to summarize faithfully? Are reviews structured? Are returns and shipping policies machine readable? Then move to monitoring. There is no clean attribution layer for AI assistant referrals yet, and you will need a method, even if imperfect before you start optimizing. Sidekick changes the agency conversation too. A lot of we'll set up your flow automations line item is now achievable in merchant by a non-technical operator with a prompt. But my take on this, well, the sidekick step up is real but evolutionary. Shopify has been telegraphing this trajectory for two release cycles. Agentic Storefronts is the most interesting bet, and it is genuinely too early to call. Conversational commerce as a meaningful share of GMV is unproven. What is true is that Shopify has made the cost of being present in those surfaces close to zero for any merchant on the platform, which is the right product decision regardless of how big the demand in the channel actually becomes. Confidence level, high on sidekick utility, medium on agentix storefronts as a near-term revenue channel, but high that being absent will look careless within six months. Last but not least, Pentagon awards classified AI contracts to eight firms and leaves Anthropic on the outside. So what happened? The Department of Defense in the US announced on the 1st of May that it has assigned agreements with seven major technology companies OpenAI, Google, Microsoft, AWS, Nvidia, SpaceX, and Reflection to deploy AI on its most sensitive classified network, with Oracle added hours later for a total of eight. The contracts cover impact level 6 and 7 environments, which is mission planning, intelligence analysis, and weapons targeting territory. Anthropic was not included. So what is confirmed? Confirmed via the DOD announcement and reporting from CNN, the Washington Post, TechTrunch, and Defense News. The reporting is consistent that Anthropic was excluded over its refusal to drop usage restrictions that prevent its models from being used for warfare and mass surveillance. Anthropic had a $200 million contract with the DoD last year and sued the administration after the initial freeze, and a federal judge in California blocked the government's earlier attempt to end the relationship. Reporting also notes that the White House has reopened discussions following Anthropic's release of its Mythos cybersecurity tool. Why it matters. So the marketing relevance is indirect but real. First, it is a hard data point on AI vendor concentration risk. When eight of the firms you might depend on for an AI enterprise workload also share a primary government customer, single vendor strategies become harder to defend in a procurement review. Second, brand association. Anthropic's positioning around safety is now an enforced commercial constraint, not just the tagline, and clients with views on AI ethics will start to notice which vendor is on which list. So what does that mean for business? Well, if you operate in regulated sectors, finance, healthcare, public sector, and certain defense adjacent supply chains, you should expect more procurement questions about which underlying models are in your stack. If you are in customer brand territory, a few clients will start asking explicit questions about AI vendor positioning. You're ready to answer in one paragraph rather than improvising. What does it mean for marketers and agencies? Well, two practical implications. One, multi-model is now hygiene position, not a hedge. If your agency's AI workflows are single vendor, name a second model in your client documentations quarter. Two brand teams working on values led positioning will get pulled into vendor selection conversations. Treat that as an inevitable and useful, an agency that cannot speak credibly to model selection is more defensible than one that cannot. So my take. The bigger signal is that the US government has decided AI vendor selection is a national interest decision, and that posture will spread to large enterprise procurement playbooks within 12 months. My confidence level, high on the contract awards and the exclusion, medium on whether the freeze holds, high on the spillover into enterprise procurement language. Two things on the watch list today. Indirect prompt injection moves from research paper to observed behavior. Google's online security blog reported a 32% relative increase in malicious prompt injection content on the open web between November and February, drawn from a base of 2 to 3 billion crawled pages a month. Sophistication is still low. Most attempts are pranks, scraping defenses, or low-grade data exfiltration. But the trend is up and the cost of automating attacks is falling. If you are deploying agents that browse the web on a user's behalf, this is the moment to write down your isolation, allow list and human in the loop policy before you have to write it under pressure. And secondly, a fortune analysis of last week's big tech earnings argues that roughly half of Google and Amazon's reported AI profit upside came from mark-to-market gains on their respective anthropic stakes, not from operating revenue. The number is contested and the methodology is not bulletproof, but the framing is useful. Not every AI earnings beat is operational performance, worth holding in mind the next time a board paper cites hyperscaler earnings as proof that AI is paying back today. So what matters the most? The signal across these four stories is repricing that novelty. Adobe is repricing the marketing technology bundle to absorb AI search visibility. Google is repricing search ads by removing the keyword as the unit of control. Shopify is repricing e-commerce ops by handing sidekick a far longer to-do list and pushing the catalog into chat. The Pentagon is repricing AI vendor risk in the most explicit way available by writing some companies in and one out. The noise this week is the league table chatter about model benchmarks and valuations, so ignore it. The operator question is which line on your own PL moves as a result of these four shifts and on whose timeline? Here are the four things I would think about if I were running this account, brand, or agency this week. Run a one-day GEO and ASO audit on your top three clients on your own brand. Pick five high-intent questions, ask them in ChatGPT perplexity and copilot and screenshot the citations. That is your baseline before Adobe Semrush integrates and the category professionalizes. Five hours of work useful for a year. Map every campaign in your Google Ads account that will be auto-upgraded in September and run the voluntary upgrade tool on non-critical campaign first. Document what AI Max changed in the URL expansion and search term matching that the previous setup did not allow. This document is the basis of your future client briefing. For any Shopify Plus or D2C client, switch on agentic storefronts and audit the catalog feed. Treat it as a catalog hygiene work, not as a channel launch. Budget the time as part of a routine merchandising, not a project. Write a half-page note on AI vendor diversity for your top client retainers. Name the models you use, why, and what the fallback is. File it, you will get asked. Add an indirect prompt injection clause to any agentic workflow your team is running. Define what the agent is allowed to read, what it is not, and who reviews any external write action. This is 15 minutes of policy that prevents an embarrassing incident. So what I'm going to tell my clients about today, the keyword only Google Ads ERA ends in September. Use the next four months to design the test, not the migration. Your AI search visibility is now a measurable line item with a credible mainstream vendor behind it. Read it like SEO in 2010, early scrappy but worth getting in on the agenda. Single model AI workflows and are your procurement exposure, not just a technical preference. Name your second model in writing. The thread running through today is that platforms have stopped asking permission. They are upgrading the path, retiring the legacy product, and shipping the agent into surface where the customer already is. The operators who move fast will treat this as a four small migrations done deliberately. The ones who wait will discover that the migration has already happened to them, and the only remaining question is how much was lost in the move. That's been today's episode. See you tomorrow.