"We're going to get to a point where you're a business, you come to us, you tell us what your objective is, you connect to your bank account, you don't need any creative, you don't need any targeting demographic, you don't need any measurement, except to be able to read the results that we spit out."
That's Zuckerberg, speaking at Meta's Annual Shareholder Meeting in May 2025 and in an interview shortly after. When the Wall Street Journal reported it on June 2, agency holding company stocks fell immediately. The read was obvious: if anyone can run a Meta campaign by entering a URL and connecting a bank account, what happens to the teams that run Meta campaigns professionally?
The question is the right one. The typical answer is wrong.
In this post:
- What Meta's URL-to-campaign system does at each step
- Why agency stocks fell — and what that reaction misses
- What VaynerMedia's APAC team said that most coverage ignored
- Why the approval layer becomes more critical as automation advances
- How to position now, before Q4 2026 full rollout
What URL-to-Campaign Actually Does
The mechanism starts with your landing page. Meta's AI crawls the URL, extracts brand language, product information, offer structure, and visual assets. It layers in your account's historical performance signals — which audiences converted, which creative formats drove results, which placements produced the best CPM — to generate a complete campaign.
The output: copy variants, imagery (generated or selected from the URL assets), audience parameters, placement mix, bid strategy, and campaign structure. Ready to run.
What AI handles in URL-to-campaign
- Creative copywriting: headlines, primary text, CTAs — multiple variants per format
- Audience configuration: interest, behavioral, and lookalike targeting from account history
- Placement selection: Feed, Stories, Reels, network placements by predicted performance
- Bid strategy: cost cap, lowest cost, or value optimization based on historical CPA
- Creative formatting: automatic resizing and reformatting for each placement
Two things to be precise about. First, current state: Advantage+ campaigns already automate targeting, placement, and bid management and are now the default for all new campaigns. Full creative generation from a URL is in closed beta. Advantage+ campaigns deliver $4.52 return per $1 spent — a 22% ROAS advantage over manual campaigns. Meta isn't automating execution out of ideology; it's following systematic performance data.
Second, timeline: the broad public launch of full URL-to-campaign automation is targeted for Q4 2026. The direction is certain; the exact rollout date is optimistic.
Why Agency Stocks Fell — and What That Misses
When the WSJ reported Meta's vision on June 2, 2025, agency holding companies fell within the same trading session:
| Agency | Single-day drop |
|---|---|
| Publicis Groupe | -3.8% |
| Omnicom Group | -3.2% to -4.3% |
| Interpublic Group | -1.9% to -3.0% |
| WPP | -2.2% to -3.0% |
The investor thesis was direct: campaign production is a significant portion of what agencies bill for. If AI produces the campaign, the production fee disappears. Revenue per managed account compresses.
That thesis is correct for one category of agency — those whose value is primarily building and trafficking ads. If AI generates the deliverable, the production markup goes away.
But markets priced in the worst version of the story. Morningstar analyst Mark Giarelli put the distinction plainly: "The industry is being disrupted, but it's not being disintermediated." The agencies whose value sits in creative strategy, brand direction, and multi-channel planning don't lose revenue when Meta automates campaign construction — they gain capacity. What gets automated is the mechanical layer. What doesn't is judgment.
The APAC View Most Coverage Missed
The Campaign Asia analysis of Meta's automation push surfaced a response that the US-focused coverage almost entirely missed.
Marc Langenfeld, Head of Media at VaynerMedia APAC, was explicit about where the limits are: "AI still has limitations, especially around nuance and accuracy like product descriptions and benefits."
That framing carries weight from a firm that has built its own AI creative infrastructure and isn't arguing against automation from a defensive position. His point is operational, not ideological: Meta's AI trains on global behavioral data, but APAC advertising plays out across 20+ languages, market-specific regulatory constraints, and cultural contexts where a localization error isn't just suboptimal — it's a brand risk. Automated optimization toward conversion can underweight everything it can't measure.
This is the nuanced version of the human-in-the-loop argument. It's not "AI can't write copy." It's "AI can't catch what's contextually wrong with its own copy." That distinction holds in any market — not just APAC.
The Approval Layer Is the Performance Layer
Here's what the fully-automated-ads conversation consistently misses: automation doesn't eliminate human decisions. It concentrates them.
When a media buyer manually builds campaigns, decision-making is distributed across every step — audience selection, creative choices, bid configuration, placement mix. Most of those decisions are mechanical. A handful are consequential.
When AI builds the campaign, the mechanical decisions disappear. One decision remains: approve or modify what the AI proposed.
That approval moment now carries the weight previously spread across dozens of steps. A campaign that runs wrong at scale for two days costs real budget. An AI optimizing toward short-term conversion while quietly eroding brand equity creates a harder-to-fix problem than any manual mis-click. The people reviewing AI-generated campaigns need to understand what good looks like well enough to catch what the AI got subtly wrong — Langenfeld's nuance point, applied to every market.
Approving an AI-generated campaign means holding creative direction, audience signal, bid logic, and placement strategy simultaneously — and spotting where the AI over-optimized or missed context. That requires deep account knowledge, not just interface familiarity.
The approval layer isn't where you save time. It's where performance is either preserved or lost.
Where to Position Now
The teams least disrupted by URL-to-campaign are already operating at the judgment layer — reviewing and directing AI output rather than producing campaign components manually.
bulk is built for that model. It reads your live Meta account, proposes a campaign plan, and executes once you approve. Teams using bulk have already reorganized around strategic approval rather than mechanical execution. The transition that others will need to make when full URL-to-campaign automation arrives, bulk users have already made. bulk handles what used to require manual steps; you own the decisions that actually drive results.
For the full context on where Meta's automation roadmap stands: Meta Full Ad Automation 2026: Threat or Opportunity?. For where Meta's AI assistant sits inside Ads Manager today: Meta AI Business Assistant: What It Does (and Doesn't).
The execution era is ending. The approval layer is where performance is decided — and where the operators who understand that are building leverage right now.
bulk handles the execution layer for Meta ads — reading your account, proposing a plan, and running it once you approve. Try bulk free →