Google & Meta Just Opened Ads to AI Agents

Welcome back to the front row.

Something shifted in 48 hours at the end of April. Most practitioners missed it. This week, we're connecting the dots.

This week: Google and Meta both opened their ad infrastructure to external AI agents in the same 48-hour window. Not a coincidence. Not a test. Infrastructure. Here's what it means for every practitioner running campaigns right now.

MAIN STORY

The door is open. Are you walking through it?

Let me be direct about what happened last week.

On April 30, Google announced AI Brief — a natural language layer inside AI Max that lets you describe your campaign in plain English. Not keywords. Not match types. Not a list of excluded search terms. You write what your campaign should and shouldn't say, which searches to prioritize, which audiences to reach and how to message them. Google's AI reads your brief, generates sample assets, and shows you previews before anything goes live. You review. You iterate. You approve.

The same week, Meta launched its official Ads MCP Server and CLI. Full read and write access. Twenty-nine tools. Any MCP-compatible agent — Claude, ChatGPT, or any agent you deploy yourself — can now query your campaigns, create new ones, update budgets, manage audiences, and run diagnostics. No developer approval wait. No custom API build. One authentication flow and your AI agent is inside your Meta account.

Two platforms. Two doors opened. Same week.

This is not a product update. This is a structural change in what the interface to advertising looks like.

From dashboard to conversation

For twenty years, managing ad campaigns meant navigating a UI. Click into the campaign. Set the match type. Choose the audience. Approve the creative. Export the report. The dashboard was the interface — and whoever knew it best had an edge. That era is ending.

AI Brief doesn't ask you to navigate Google Ads. It asks you to brief it. The same way you'd brief a human account manager before a campaign launch. "We sell premium skincare to women 35–55 in the UAE. Focus on product benefits, not discounts. Avoid searches with 'cheap' or 'free'. Prioritise users showing purchase intent." That's a campaign brief. That's now also a Google Ads input.

Meta's MCP goes further. It doesn't just change the input layer — it opens the execution layer. Your own AI agent can now sit inside your Meta account and operate it. Not read it. Operate it. Create campaigns, adjust budgets, manage ad sets — all from a natural language instruction in whatever AI tool you use. Acadia's head of paid media called it a bigger deal than people realise, given how manual the platform has always been at scale.

And within the same week, Meta also launched "Describe Your Audience" — a natural language targeting option that replaces manual interest selection. You describe who you want to reach. The system figures out the rest.

The pattern is the same across both platforms. The prompt is replacing the click.

What this means for your campaigns right now

Two things are happening simultaneously and most practitioners are only seeing one of them.

The first is obvious: these tools make campaign management faster. AI Brief saves hours of manual setup. Meta MCP removes the friction of navigating Ads Manager. If you adopt them, you move faster than competitors who don't.

The second is less obvious — and it's the one that matters more.

The quality of your input is now the quality of your output. Always was, to some degree. But the gap is about to widen dramatically.

When you brief AI Max in plain language, a vague brief produces vague targeting. When your AI agent manages Meta campaigns based on your instructions, a weak instruction produces weak campaigns. The AI doesn't know your product positioning. It doesn't know your customer's language. It doesn't know that the audience who converts for you sounds nothing like the audience you'd describe in a demographic brief. You do.

The practitioners who will compound their advantage from here are the ones who learn to write better briefs than everyone else. Clear product positioning. Specific audience language. Honest signal about what works and what doesn't. As Digiday noted, what was once a labor-intensive process of building keyword lists and exclusions is now a prompt-based system where marketers describe intent and set guardrails. That starts to look less like media buying and more like prompt engineering.

There's also a third layer forming in real time. Within the same week Google and Meta opened their infrastructure, AdKit — a Singapore-based platform — launched an MCP service connecting AI agents to both Google and Meta ad accounts with a draft-first safety layer. Every agent action is staged for human review before it touches a live account. Early users report campaign setup dropping from 45 minutes to a single prompt. The founder built it after a decade of hating Ads Manager. The tool exists because the infrastructure just opened, and the ecosystem is already building on top of it.

This is what a platform shift looks like in real time. The infrastructure opens. The tools follow. The practitioners who move early set the standard everyone else eventually follows.

The urgency layer

None of this is theoretical. AI Mode ads are live — Shopping and Travel formats active inside Google's conversational search layer, with a Direct Offers pilot running for brands showing purchase intent. The Shopping tab grid now carries sponsored placements for the first time ever. Google and Shopify reported 11x growth in AI-mediated orders between January 2025 and March 2026.

AI assistants now account for up to 56% of global search volume according to Graphite.io. The majority of search is happening in a layer that wasn't built around keywords. And the platforms just handed you the tools to operate in that layer — not as a future capability, but right now.

The door is open.

The brief is the strategy.

Walk through it.

PROMPT OF THE WEEK

Run this before your next campaign brief goes live:

"Read this campaign brief: [paste your brief]. Tell me where it is vague, what intent it could accidentally match, and the three things you'd rewrite to make it sharper. Be ruthless."

If the AI finds more than two things to fix — your brief isn't ready. Sharpen the input before you spend the budget.

THIS WEEK

Google AI Brief and AI Max expansion — April 30. Google launched AI Brief, a natural language input layer for AI Max, alongside expanding AI Max to Shopping and Travel campaigns. Text disclaimers for regulated industries also launched — guaranteeing required legal text appears in ads even when final URL expansion is active. Healthcare and finance advertisers can now adopt without compliance risk.

Meta officially opens ad accounts to external AI agents — April 29. Meta launched its official MCP Server and CLI, giving any MCP-compatible AI agent full read and write access to Meta ad accounts. Same week: "Describe Your Audience" launched, replacing manual interest targeting with plain language input. Two changes that together redefine what managing a Meta account looks like.

AI assistants cross 56% of global search volume. New Graphite.io data shows AI assistants including ChatGPT and Gemini now account for more than half of global search volume. The tipping point is not ahead of us. It's behind us. If your visibility strategy is still built entirely around the classic ten blue links, you're already operating in the minority surface.

Meta overtakes Google in digital ad revenue — first time ever. Emarketer forecasts Meta at $243 billion in 2026 ad revenue versus Google's $239 billion — the first time in history Meta leads. Meta growing at 24% year-over-year versus Google's 12%. The company that moved faster on AI infrastructure is now the largest advertising platform on earth.

TOOL OF THE WEEK

GEO Engine — by NVM (Launching soon)

The issue this week is about the new layer of advertising — AI search surfaces, conversational interfaces, agent-managed campaigns. But here's the question nobody is asking loudly enough yet: are you even visible in that layer?

Most practitioners don't know. Their SEO tells them they rank. Their paid campaigns are running. But whether ChatGPT, Gemini, or Claude would cite them when a qualified buyer asks a relevant question — they have no idea.

GEO Engine is the tool I built to answer exactly that. Drop in a URL. Get a full AI visibility audit — how citable you are across the major AI systems, where your real competitors stand, and the exact content gaps where AI citations are up for grabs. One URL in. Full picture out. No agency needed.

It's not live yet. But if this issue made you realise you need to know where you stand in the AI layer — drop your email at nvm.marketing and you'll be first through the door when it launches.

Built by a practitioner. For practitioners.

MY TAKE

The platforms opening their infrastructure to AI agents is the most significant structural shift in performance marketing since Smart Bidding replaced manual CPC. Most practitioners will underestimate it for six months and then scramble to catch up.

Here's what I keep coming back to. The platforms are not replacing strategists. They are replacing operators. The person who knew every dropdown in Ads Manager, every targeting toggle in Meta, every bid strategy nuance in PMax — that operational knowledge is compressing fast. The AI handles execution better than most humans do at that layer now.

What doesn't compress is judgment. The ability to articulate what a qualified buyer actually looks like. The instinct to know when a campaign is matching the wrong intent even when the numbers look fine. The clarity to write a brief that gives the AI something real to work with instead of something generic to amplify.

The brief is your strategy now. Everything else is execution.

Own the brief.

That's Issue #3 of NVM. If this gave you one thing to act on this week, share it with someone who needs it. Reply to this if you have a question — I read every one. See you next week. — Mostafa

NVM · Front row seats to AI and digital marketing