1 Meta's AI ad tools are fumbling — and the fine print says it's your problem
Meta is drawing advertiser ire for pushing AI creative tools despite visible blunders: an AI-generated ad for REI featured a bike with two handlebars, and a campaign for a women's networking group was built around a man. Other errors are subtler but worse — AI altering a brand's actual products. When Business Insider asked, a Meta spokesperson pointed to the terms of service: "AI can make mistakes and it is the advertiser's responsibility to review the AI outputs."
The liability for AI creative just moved contractually onto the advertiser — and performance marketers still won't leave, because as one ads consultant put it, "Meta's still the best platform. It has the best data." The practical takeaway: if you run Advantage+ creative, a human QA gate on every AI output is no longer optional. It's your legal exposure now.
2 OpenAI's ad business is pacing 90% below its own forecast
OpenAI has told investors it will hit $2.5 billion in ad revenue this year and $100 billion by 2030. Emarketer's new forecast says US standalone chatbot ads — ChatGPT, Copilot, Google AI Mode, Alexa Shopping combined — will generate under $1 billion in 2026, and the entire market caps out at $5.41 billion by 2030. The forecast assumes OpenAI captures search budgets en masse, dominates a mature chatbot ad market, and outperforms every ad format in history — all at once.
Meanwhile, more than 80% of AI-adjacent ad spend in 2026 sits next to AI content — like search ads beside Google AI Overviews — not inside chatbot conversations. If you're allocating test budget to conversational ad inventory, size it as an experiment, not a channel. The demand-side math hasn't caught up to the pitch decks.
3 Instagram confirmed it will charge for heavy AI use
Instagram head Adam Mosseri confirmed in a July 12 Stories Q&A that a paid subscription tier for heavy AI usage is coming: "these AI models are very expensive to run." Instagram already imposes daily caps on Muse-powered image and video tools, and users hitting the ceiling see a prompt to subscribe. No pricing, feature list, or launch date yet — but the direction is set: basic access stays free, volume becomes a paid benefit.
Free platform AI is ending, and it will land first on the people producing the most content — creators and social teams. Budget owners should expect "AI usage" to appear as a line item in creator tooling costs within the year, the same way scheduling and editing tools did.
4 TikTok Shop banned AI voices and recorded audio from shopping livestreams
TikTok updated its Shop policy to ban non-real-time verbal interaction from shopping livestreams: no AI-generated voices, no audio recordings, no radio-style narration, no looping footage or slideshows, and no avatars covering more than half the screen. Enforcement runs through the Creator Health Rating system, with penalties up to commission restrictions and account bans. AI prep tools like Symphony remain allowed — before the stream, not inside it.
TikTok just drew the line every platform will eventually draw: AI can prepare the performance, but humans must deliver it. When consumers spot AI in brand content, they're four times more likely to trust the brand less. Liveness is becoming a ranked, enforced commodity — plan your live commerce staffing accordingly.
5 Adobe Advertising launched Custom Algorithms — your analytics now train the bidder
Adobe Advertising released its Custom Algorithms product into general availability after a year in beta. The DSP now optimizes campaigns using Adobe Analytics data directly — full customer journey, app and web traffic patterns, bounce-rate signals, and first-party identity data that DSPs typically never see. Adobe's GM calls it "a huge departure from the TubeMogul days."
This is a "performance easy button" with one key inversion: Google's Performance Max and Meta's Advantage+ train on their own data, while Adobe's trains on yours. For brands already deep in the Adobe stack, that's a genuinely differentiated pitch — automated bidding without surrendering the learning loop to a walled garden.
6 DoorDash became a native sales channel inside Shopify
DoorDash announced on July 14 that it's now a native sales channel in Shopify's App Store. Any US Shopify merchant with a brick-and-mortar location can add their catalog to the DoorDash marketplace — 42 million monthly users — with real-time catalog sync, no separate onboarding, and setup cut from weeks to days. International expansion follows.
Commerce distribution keeps collapsing into toggles. For local retailers, on-demand delivery just became a checkbox next to their existing channels — and for marketers, DoorDash's marketplace becomes another surface where product data, not ad creative, determines visibility. Clean catalogs are the new campaign asset.
7 USA Today's AI "shell files" drew 116 million page views — now aimed at the World Cup
USA Today Co. used AI-assisted, pre-built "shell files" — story templates prepared before news breaks — to publish Winter Olympics coverage within minutes of events. The result: 116 million page views across the network from January 1 to February 28, with the flagship title drawing 91 million, up 82% from 2022. The playbook now moves to the World Cup, with five shell files ready each day, explicitly to beat Google's AI Overviews to the reader.
This is the sharpest publisher counter-move to AI search yet: use AI to be faster than the AI summarizing you. For brands, the lesson transfers — in an AI Overview world, being first with structured, publishable content is a distribution strategy, not just a newsroom tactic.
8 Media buying is going agentic — and buyers are becoming managers
Agents will soon execute the full media investment process, from buying to measurement. The human role, per Media by Mother's activation director: become "decision makers and managers" who check the agent's work — a performance spike could be a great buy or a measurement glitch, and "an agent won't be able to recognize these things without the context we have." Ebiquity's CEO adds that building and governing agentic systems are now core buyer skills. The open question: where do junior buyers learn, when the tasks that trained their seniors are handed to agents?
Digiday reports agencies are betting on agents for an edge while fending off brands in-housing the same tech. The org-chart implication is bigger than the tooling one: governance, QA, and test-and-learn design are the new career ladder in media — hire and train for them now.
9 State attorneys general sued to block Paramount Skydance–WBD
Multiple state and regional attorneys general filed suit on July 13 to block Paramount Skydance's merger with Warner Bros. Discovery. Paramount is reportedly considering moving its headquarters out of California to counter the state's impending antitrust action. This stacks onto the EU remedies deadline of July 22 and the UK's "minded to intervene" position on the $110 billion deal.
The consolidation of premium video supply now runs through courtrooms on three continents. If your upfront commitments assume a combined Paramount–WBD footprint, build the scenario where the deal closes late, restructured, or not at all — inventory, measurement currencies, and CTV pricing all move differently in each branch.
10 Zappi shipped Amplify AI — pretest hundreds of video ads before launch
Zappi announced Amplify AI on July 13, a predictive ad testing solution that evaluates hundreds of social video ads before launch, plus Amplify Hub, a connected workspace that pulls advertising insights into one place. The pitch: creative validation at the same speed and volume that AI generation now produces creative.
This is the counterweight to trend #1. If AI tools generate mistakes at scale and platforms make review your responsibility, then pretesting at scale is how you operationalize that responsibility. The brands that pair generative volume with predictive validation get the upside of AI creative without wearing its liability raw.