2026 has been the year Meta stopped asking advertisers to configure campaigns and started asking them to supply creative. Across the year, targeting controls were removed, a new LLM-based ranking system replaced the old one, and a metric appeared in Ads Manager that scores how varied your creative is. If your account is still structured the way it was in 2024 — tight ad sets, manual placements, three near-identical creatives per set — it is structured against the way the system now works. Here is the changelog that actually matters, and what to do about each item.
The Generative Recommender Changed How Ads Are Ranked
On 29 July 2026, Meta rolled out the Generative Recommender, an LLM-based ad ranking system that evaluates creative and user preference together rather than treating them as separate inputs. Meta has reported a 15.7% conversion lift on Facebook and roughly 1% on Instagram from the change. The strategic implication is bigger than the headline number: when the ranking system reads creative semantically, the creative itself becomes a targeting input. What your ad says and shows now helps determine who sees it — which is why creative volume and variety started mattering more than audience configuration.
Creative Diversity Score: Meta Is Now Grading Your Variety
Since 26 August 2026, Ads Manager carries a Creative Diversity Score rating ad sets Low, Medium or High based on visual and thematic variety. This is not a vanity metric — it reflects how the Generative Recommender can segment audiences against your ad set. Five colour variants of the same static image is Low diversity: the system has one idea to work with and one audience pocket to find. A static, a UGC video, a founder talking to camera, a product demo and a customer testimonial is High diversity: five genuinely different hooks that can each find a different pocket of demand.
- →Vary the format — static, video, carousel, UGC, motion
- →Vary the message angle — price, speed, proof, risk reversal, identity
- →Vary the hook in the first two seconds, which is where video is won or lost
- →Vary the social proof — different customers, different industries, different outcomes
- →Stop shipping colour and crop variants and calling it creative testing
Placement Exclusions Are Gone — Plan Creative Accordingly
On 21 August 2026, Meta removed the ability to exclude individual placements at the ad-set level. Account-level restrictions and value rules remain, but the familiar workflow of unticking Audience Network or Reels inside a single ad set is finished. Practically, this means every ad you run should be built to survive every placement. Design for 9:16 and 1:1 as first-class formats rather than cropping a 16:9 asset and hoping. Keep text away from the edges where Reels and Stories overlay UI. Assume the ad will be seen with sound off and make the first frame carry the message on its own.
Exclusion-Only Custom Audiences Filled a Real Gap
On 10 August 2026, Meta introduced exclusion-only custom audiences — a dedicated audience type for permanent suppression across the network. This partially closes the gap left when detailed targeting exclusions were removed. Use it for the audiences you genuinely never want to pay for again: existing customers on a product they already own, refunded or chargeback customers, job applicants, your own staff, and competitor employees where you can identify them. For lead generation in particular, suppressing converted leads stops you paying to re-acquire people already in your pipeline.
Advantage+, Brand Memory and AI-Generated Creative
Meta spent 2026 making AI the default workflow rather than an option. In May, expanded Advantage+ capabilities, automated audience discovery and AI creative generation were integrated directly into Ads Manager. In June, Brand Memory launched, letting the system learn and apply consistent brand visual identity and messaging patterns across AI-generated variations. In July, the Muse image generation tool was integrated into Advantage+ Creative, letting the system generate original image variants inside campaigns. Used well, this solves the volume problem that Creative Diversity Score creates. Used badly, it produces a feed of plausible, generic, forgettable ads. Our rule: AI generates variations on a human-decided angle, never the angle itself.
What Changed for European Advertising
Two 2026 changes specifically affect advertisers targeting the EU. In June, certain targeting capabilities were removed in European markets under Digital Markets Act requirements. In July, Meta began charging location-based fees on top of standard CPMs for European advertising, which raises effective cost per acquisition in EU markets. If your media plan was built on 2025 EU CPMs, rebuild the model — and if you serve European clients from India, quote on current numbers rather than last year's.
The Account Structure That Fits 2026 Meta
Put the changes together and the structure that works follows logically. Fewer, broader ad sets — the system segments better than your ad-set splits do. More creative per ad set, genuinely different from each other, aimed at a High Creative Diversity Score. Exclusions handled with exclusion-only custom audiences rather than targeting settings. Every asset built for every placement, because you can no longer opt out. And conversion data fed back accurately, because a semantic ranking system optimising against bad conversion signal will scale the wrong thing faster than the old system did.
Key Takeaway
The through-line across every 2026 Meta update is the same: Meta has taken the configuration levers away and handed advertisers a creative brief instead. That is genuinely bad news for accounts whose edge was clever targeting, and genuinely good news for businesses with real customers, real proof and something specific to say. If your Meta account is still built on 2024 assumptions, the fix is not a settings change — it is a creative production plan. That is the work we do for clients: a tested angle library, enough format variety to satisfy the Recommender, and conversion tracking honest enough that the system optimises toward revenue rather than noise.
