Meta’s Current Recommendations on Ad Targeting and Creatives: What the Official Sources Say

Meta’s advice to advertisers has quietly flipped. The old playbook said: define your audience as precisely as you can, then hand-craft one strong ad for it. Meta’s current documentation pushes the opposite direction – let its AI decide who sees your ads, and feed the system many different creatives so it can match the right ad to the right person.

This post is a researched article, not a personal account. Everything below comes from Meta’s own documentation and training pages, linked in the Sources section at the end. Where performance numbers appear, they are Meta’s own figures from Meta’s own pages, with dates attached. Where Meta does not publish something, that is said plainly rather than filled in.

1. Targeting: Meta now recommends letting its AI find the audience

The clearest statement of Meta’s current position is its Advantage+ audience feature. In Meta’s own words, it “lets you use Meta’s advanced AI to find the audience for your Meta ad campaign, instead of defining that audience yourself.” The system looks at signals such as past conversions, Meta Pixel data and interactions with your previous ads, then decides who should see the ad.

Meta attaches performance figures to this recommendation on its help page. According to Meta, Advantage+ audience campaigns can achieve a 14.8% lower cost per result with the Awareness objective, 9.7% lower with Traffic, Engagement and Leads objectives, and 7.2% lower with Sales and App promotion objectives. These are Meta’s numbers from Meta’s documentation – the page does not describe the test conditions behind them, so treat them as the vendor’s own claims, not independent measurements.

Two qualifications Meta itself states. First, it recommends A/B testing Advantage+ audience “for almost all campaign types” – which is Meta telling you to test its own recommendation rather than trust it. Second, it does not recommend Advantage+ audience for retargeting campaigns. For sales, leads and app promotion objectives, Meta steers advertisers toward the more fully automated Advantage+ campaign setup.

2. What you still control

Handing targeting to the AI does not mean giving up all control, and Meta’s documentation draws a sharp line between two things. Audience controls are hard limits: locations, minimum age, custom audiences to exclude, and languages. Meta’s page says that if you set locations, “we won’t target beyond your locations” unless you explicitly opt in to wider reach. Audience suggestions, by contrast, are hints. Meta’s own example: if you suggest the gender Women, “it’s possible that your ads could also deliver to men if Meta’s AI finds them likely to respond.” If a constraint matters to you, it has to be a control, not a suggestion.

One more boundary worth knowing: advertisers in Meta’s Special Ad Categories – financial products and services, employment, housing, and social issues, elections or politics – do not get access to Advantage+ audience at all, because those categories carry stricter targeting rules.

3. Creatives: Meta wants volume and variety, and it may modify what you upload

On the creative side, the matching feature is Advantage+ creative. When it is on, Meta’s systems automatically generate optimized versions of your images and videos for different viewers. The list of enhancements on Meta’s help page is long and many are marked as using generative AI: background generation, image generation, text generation and improvements, adding music, video touch-ups, translations and more. Meta’s page is direct about what this means: “The media and text you upload may be adjusted to help improve ad performance while maintaining the core message of your campaign.” Some enhancements are on by default; each can be turned off individually.

Two governance details from the same page. Ads created or significantly edited with Meta’s generative AI features may carry an “AI info” label, and Meta’s supplemental terms for generative AI ad features took effect on May 6, 2024. Advertisers in regulated verticals such as financial services and pharma get slower access to these features.

4. The words Meta uses in its own training: creative diversification and Andromeda

Beyond the help pages, Meta’s official training platform, Meta Blueprint, published a session in May 2025 titled “Creative Diversification, Andromeda & The Auction.” It is the clearest official statement of how Meta thinks about creatives now. The session describes how Meta’s AI-powered delivery system uses creative signals – “format, messaging, persona, hook type” – to identify and reach new audience segments, and presents creative diversification as a performance lever: a lack of diversity in your creatives “may limit both reach and efficiency.”

Andromeda, in the same session, is the name of Meta’s ad retrieval system – the stage that shortlists which ads are even considered for a given impression. Meta originally described Andromeda in an engineering blog post, but that post is no longer available at its old address; the Blueprint session is Meta’s current official public reference. The practical implication for advertisers: if the retrieval system matches creatives to audience segments, then running several genuinely different ads – different hooks, different personas, different formats – gives the system more ways to find buyers. Running five near-identical variations of one ad does not.

5. The machinery behind the push (background)

None of this is sudden. In May 2023, Meta’s AI blog described Meta Lattice, a model architecture that replaced many smaller, siloed ad-ranking models with one large system learning across surfaces and objectives. Meta reported that early deployment on Instagram produced “an ~8 percent improvement in ads quality” – again, Meta’s own figure for its own early results, from a dated blog post. Lattice, Andromeda and the Advantage+ features are steps on the same road: more of the ad delivery decision-making moves from the advertiser’s settings into Meta’s models.

6. What this means for a small advertiser

Read together, Meta’s current recommendations add up to a clear division of labor. You decide the goal, the budget, the boundaries and the raw material. Meta’s AI decides who sees what, in which version.

Test the automation instead of arguing with it. Meta itself recommends A/B testing Advantage+ audience rather than assuming it wins. Run it against your manual audience and keep your own records – the percentages on help pages are the vendor’s numbers.

Put real boundaries in controls, not suggestions. Locations, minimum age and exclusions belong in Audience controls. Anything set as a suggestion can be overridden by the system.

Diversify creatives on purpose. Meta’s own training says its system reads format, messaging, persona and hook type. Give it genuinely different ads to work with, not minor edits of one winner.

Check the automatic enhancements. Advantage+ creative can change how your ads look and read. Review which enhancements are on before you publish, especially anything that rewrites your text.

A companion piece on this site looks at the same shift from the operator’s side: AI Can Build Your Meta Ads. What Should You Still Control?

FAQ

Is Advantage+ audience the same as boosting a post? No. Boosting is a simplified promotion flow. Advantage+ audience is an AI targeting option inside full campaigns in Ads Manager, and it can be A/B tested against manual targeting.

Can I still exclude people with Advantage+ audience on? Yes. Locations, minimum age, excluded custom audiences and languages can be set as Audience controls, which Meta’s documentation says act as hard limits.

Will Meta change my ad images and text? It can, if Advantage+ creative enhancements are on. Meta’s help page lists the enhancements and states that uploaded media and text may be adjusted while keeping the core message. Each enhancement can be turned off individually.

Sources

All sources are Meta’s own pages, accessed October 29, 2026:

1. About Advantage+ audience – Meta Business Help Center (official). Source of the cost-per-result figures and the A/B testing recommendation.

2. About Audience controls and Audience suggestions in Advantage+ audience – Meta Business Help Center (official). Source of the controls-vs-suggestions distinction and the Special Ad Category exclusion.

3. About Advantage+ creative – Meta Business Help Center (official). Source of the enhancement list, AI labeling and the May 6, 2024 generative AI terms.

4. Ask the Experts Session: Creative Diversification, Andromeda & The Auction – Meta Blueprint (official Meta training), published May 15, 2025. Source of the creative signals list and Andromeda as Meta’s ad retrieval system.

5. New AI advancements drive Meta’s ads system performance and efficiency – Meta AI blog (official), May 11, 2023. Background on Meta Lattice; the ~8% ads quality figure is Meta’s early result on Instagram.