1 Google's new Ads terms took effect — automation is the default, the liability is yours
On July 1, Google Ads' first Terms of Service update in eight years quietly took effect. The new terms authorize Google to use your inputs — URLs, text typed into conversational campaign features — across its AI-powered tools to generate, select, and optimize ads, and they make clear that advertisers remain responsible for reviewing and approving whatever the machine produces. No opt-in, no action required: acceptance was automatic.
Read that fine print again: automation just moved from a feature you enable to the baseline you operate inside. Build an asset-review cadence now — weekly audits of AI-generated headlines, images, and landing-page matches — because "Google generated it" is not a defense your legal team can use.
2 Meta put Muse inside Advantage+ — creative generation moved inside the buy
On July 8, Meta announced it is bringing Muse, the image-generation model from Meta Superintelligence Labs, into Advantage+ creative. The model powers smarter ad-creative generation plus new room-restyling capabilities designed to make shopping ads more immersive.
Meta doesn't just want the auction anymore — it wants the asset. When the platform generates creative on its own model, your differentiation shifts upstream to inputs: product data, brand systems, and offers only you can make. Feed it well, and audit what it makes — the liability logic from item 1 applies here too.
3 Microsoft started swapping OpenAI and Anthropic out of Office
Bloomberg reported July 7 that Microsoft has begun replacing OpenAI and Anthropic models with its in-house MAI models inside Excel and Outlook, with tens of thousands of prompts a week already running on Microsoft's own stack. The stated driver is straightforward: AI costs.
The model layer is becoming invisible infrastructure — vendors will swap engines under your workflows without a changelog you'll ever read. If a tool's AI output matters to your brand, ask which model runs it and what re-testing happens when it changes. Prompt libraries tuned to one model quietly degrade on another.
4 Threads ads went global with carousels and video — and Reels got post-view
Meta expanded static carousel and video ads on Threads to all markets, and is rolling out post-view ads in Instagram Reels to more advertisers — a format that plays a promoted video after organic Reels longer than 60 seconds. Everything runs from the same Ads Manager you already use.
Threads remains an under-priced reach pocket relative to Instagram, and post-view is Meta manufacturing new inventory inside existing attention. Test both while competition is thin — but watch frequency and fatigue metrics, because new ad slots always feel additive until users notice the load.
5 Dentsu wired Meta's creator APIs into its operating system
Dentsu announced a partnership with Meta on July 7 to integrate the Creator Marketplace and Partnership Ads APIs into dentsu.connect, starting with its UK&I Creator & Trends Studio. The platform pairs Meta's creator data with dentsu's Velocity Score, which ranks creators on acceleration rather than static popularity — and early deployments cut creator discovery from hours to minutes while speeding planning and activation up to 4x.
Influencer marketing is becoming infrastructure: queryable, scored, and activated at hundreds-of-creators scale. If a holdco can activate a thousand creators as one media line, brands buying creators one DM at a time are competing against a different machine. Watch which platform opens its APIs next — that's where the integration race goes.
6 IAS named Lidiane Jones CEO — verification gets a product-era operator
Integral Ad Science appointed Lidiane Jones as CEO, effective July 7. The former Bumble CEO and Slack CEO — who also spent more than a decade at Microsoft — succeeds Lisa Utzschneider, who led IAS for over seven years and stays on as a board advisor through 2026.
A verification company choosing a product-and-AI operator over a sales leader — days after private equity circled Criteo at a 50% premium — says the measurement layer knows its next fight is technical. As more of the stack runs on autopilot, the vendors that grade the machines become more valuable, not less.
7 Talp raised $20M to simulate your customers before you spend
Talp raised a $20 million pre-seed led by Formus Capital, with Sunshine Lake Ventures, Aito Capital, and the a16z Scout Fund participating. The startup builds AI personas that simulate customer behavior so companies can test products, pricing, and campaigns before launch.
That's an enormous pre-seed for synthetic research, and it signals where budgets are heading: pre-testing everything cheaply against simulated audiences before real dollars move. Use it as a filter, not an oracle — synthetic personas inherit the biases of their training data, so validate winners against real panels before you bet the launch.
8 Chinese AI models are winning US budgets on price
CNBC reported July 7 that Chinese AI models are gaining ground with US companies as OpenAI and Anthropic costs surge. Cost-conscious buyers are routing workloads to models in the DeepSeek, Qwen, and GLM class, which price at a fraction of US frontier rates.
Your martech vendors feel the same pressure — expect more of them to quietly route your workloads to cheaper models. That's fine for volume tasks, but ask about model provenance, data residency, and brand-safety testing before your product copy runs through a stack you've never evaluated.
9 YouTube Shorts killed the dislike — feedback became routing, not judgment
YouTube is reworking the Shorts player: the thumbs-up becomes a heart, and the dislike button disappears in favor of "Not interested" and "Don't recommend this channel" responses that feed the recommendation system directly.
Public sentiment signals keep getting replaced by private routing signals. Engagement benchmarks reset again — hearts won't map cleanly to historical likes — and the practical read on a Short's performance shifts even further toward watch time and retention. Update your dashboards before the comparisons break.
10 India widened its ad-claims crackdown — substantiation is now table stakes
On July 8, India's food regulator FSSAI issued show-cause notices to Lotte India and Ferns N Petals over misleading claims, giving both companies seven days to respond under the Food Safety and Standards Act — an expansion of the crackdown that put six energy-drink brands on notice days earlier.
Claims enforcement is compounding across markets, and AI-generated copy raises the odds an unsubstantiated claim ships. If you sell into India — or anywhere with an awake regulator — keep a substantiation file per claim, per market, and make legal review part of the creative pipeline rather than a post-mortem.