Your Creative Is Your Targeting Now. What Meta’s Andromeda Means for How You Run Ads

Meta rebuilt the system that decides who sees your ads. It happened gradually through 2025 and it’s increasingly becoming the default. If you’re running campaigns the same way you were two years ago, the gap between your approach and the platform’s expectations is widening. But, if you’ve already been investing in creative quality and simpler structures, you’re closer to where you need to be than most of the panic online suggests.

Meta’s Andromeda algorithm replaced the legacy ad retrieval engine in late 2024 and is now the default system powering every ad on Facebook and Instagram. Instead of starting with the audience you define and finding ads to show them, Andromeda starts with your creative and finds the right audience for it. Your social media advertising performance now depends more on what your ads look like and say than on which targeting boxes you tick.

How you build campaigns, how you think about creative, and whether your brand assets are helping or hurting you all look different under this system. If you’re a business spending money on Facebook or Instagram ads in Australia it’s worth understanding what Meta is trying to build, even if your current campaigns are performing well, because the platform is going to keep moving in this direction.

What Actually Changed

Andromeda is Meta’s personalised ads retrieval engine. It evaluates historical engagement, ad copy, creative elements and format to predict which users are most likely to engage with a given ad. Rather than matching ads to predefined audiences, it builds its own understanding of who should see what based on the creative itself. Meta reports a 10,000x increase in model capacity for personalisation over the previous system, with a 22% ROAS increase for advertisers using Advantage+ creative.

Working alongside Andromeda is GEM (Generative Ads Recommendation Model), which processes behaviour sequences across organic interactions, ad engagement and messaging to continuously refine predictions. Meta’s 2026 performance announcement confirmed that GEM delivered a 3.5% lift in ad clicks on Facebook and a 1%+ gain in conversions on Instagram. A new runtime model increased conversion rates by a further 3% in Q4 2025, and Meta Lattice consolidation drove a 12% increase in overall ads quality.

Together, they formalise a shift that’s been building for years: the platform increasingly works outward from your creative assets rather than your audience settings. The difference now is that Meta has made it official. Detailed targeting is advisory. Creative signals drive delivery.

The migration isn’t optional:

• Late 2024: Andromeda shipped as Meta’s default ad retrieval engine
• January 2026: Legacy detailed targeting options deprecated entirely
• February 2026: Detailed targeting inputs officially became “suggestions not constraints”
• June 2026: Full migration deadline. Legacy targeting completely gone

If your campaigns still rely heavily on narrow interest targeting and you haven’t tested broader structures, it’s worth starting now. The system has already changed. The June deadline formalises it, but the algorithm isn’t waiting.

What “Creative Is Targeting” Actually Means in Practice

Under the old system, you built an audience (interests, demographics, lookalikes), attached creative to it, and Meta delivered ads to that group. Your targeting choices did most of the heavy lifting. Creative mattered, but the audience definition was the primary lever. You could get away with average creative if your targeting was sharp enough.

Good agencies were already moving away from that model before Andromeda had a name. Creative testing, broader audiences, simpler structures. What’s different now is that the platform has caught up to the practice and made it the default. If you’re already running consolidated campaigns with diverse creative, Andromeda hasn’t broken anything it has simply rewarded the approach.

Where it goes further: Search Engine Land’s analysis of the Andromeda and GEM systems describes how the platform now evaluates creative elements like visuals, themes, hooks and language to determine relevance. Your audience inputs are guidance signals. The algorithm can and will reach users outside those groups when it predicts better performance elsewhere.

So “testing creative” no longer means running the same message in three different formats. A testimonial video and a testimonial static image aren’t “different” to Andromeda because the message is identical. The system needs genuinely varied angles: different pain points, different value propositions, different stages of the buying decision.

Meta also tracks a Creative Similarity score across your ad account. As Search Engine Land reports, high Creative Similarity tells the algorithm your ads are repetitive, which raises your CPMs. You’re paying a premium for sameness.

Your creative library needs to say different things to different people, not just look different. That’s a Facebook ads management problem and a messaging problem in equal measure.

Your Brand Has Always Mattered. Now It’s Measurable.

If the algorithm reads your creative to decide who sees it, then the quality and consistency of your brand assets directly affect your media efficiency.

Andromeda is reading your visuals, your copy, your hooks, your colour palette and your messaging to build a profile of who your ads resonate with. If those elements are inconsistent across your ad account, it’s getting conflicting data. It can’t build a coherent picture of who responds to what because the creative itself doesn’t have a coherent identity. Scattered targeting, higher costs.

A business with a strong, consistent brand identity gives the algorithm clear signals to work with. Distinct visual language, consistent messaging angles, a recognisable tone. Andromeda can learn faster which audiences respond to which creative variations because the underlying brand provides a stable foundation.

Most coverage of Andromeda frames it as a paid social tactics story: test more creatives, simplify your campaign structure, refresh your assets more often. That’s all true, but it stops short. The upstream issue is branding. If your brand foundations are inconsistent, no amount of creative volume will give the algorithm what it needs.

We see this as the point where brand strategy and paid social performance stop being separate conversations. Businesses still treating them as different budget lines are going to find that their ad costs keep climbing regardless of how many new assets they produce.

If your Meta ads performance has shifted and you’re not sure why, we can take a look and tell you what we’d actually change.

What This Means If You’re Spending Under $5k a Month

If you’re an Australian SME spending between $1,000 and $5,000 a month on Meta ads, the Andromeda shift actually works in your favour in some ways. Complex audience segmentation never worked brilliantly at lower budgets anyway. There wasn’t enough data for the algorithm to optimise across multiple narrow audiences.

Under Andromeda, simpler campaign structures perform better. Consolidating into fewer campaigns with broader targeting gives the system a much larger pool to learn from. That’s good news if you were always uneasy about building elaborate funnel architectures.

But “simple structure” doesn’t mean “lazy creative.” The opposite. With fewer campaigns, your creative is doing all the differentiation work. You need genuinely varied assets that speak to different customer segments and stages of the buying journey. A broad campaign with five creatives that all say “we’re the best, call us today” is just as wasteful as the old approach. The algorithm needs distinct messages to test against distinct audience segments.

On refresh cycles: smaller accounts spending a few thousand dollars a month should aim for a monthly creative refresh. Larger accounts need weekly refreshes. If you’re running the same three ads you launched in January, the algorithm has already squeezed the value out of them. You’ll see it in your frequency metrics first. When frequency climbs and CPAs follow, that’s Andromeda telling you it’s exhausted the audience that responds to those specific creative signals.

Australian Facebook CPC benchmarks sit around $0.50 to $2.00, which means creative quality has an outsized impact on ROI at lower budgets. A 20% improvement in click-through rate from better creative saves more per dollar at $2,000/month than it does at $20,000/month, proportionally speaking.

Investing in strong brand assets up front reduces your ongoing ad costs because the algorithm learns faster with consistent, distinct creative. It’s cheaper to get your social strategy and brand foundations right than to keep feeding the system mediocre creative and paying inflated CPMs.

What to Do About It

Most of this isn’t new advice. If you’re working with a good agency, some of it is probably already happening. But Andromeda has raised the stakes on each of these, and it’s worth checking where you actually stand.

Audit your creative library. Not the number of assets. The number of genuinely different messages. If you have 15 ads but they all say the same thing with different images, that’s one message. Andromeda needs variety in what you’re saying, not just how it looks.

Simplify your campaign structure. Stop building complex audience segments with separate campaigns for cold, warm and retargeting. Consolidate. The algorithm is smart enough to figure out who needs retargeting and who is seeing you for the first time. Your job is to give it creative that speaks to both.

Fix your brand foundations before scaling spend. If your visual identity is inconsistent across channels, if your messaging changes depending on who made the ad that week, if there’s no clear brand system behind your creative, fix that first. Scaling ad spend on top of a weak brand just amplifies the inefficiency.

Build creative around angles, not formats. One angle exploring a customer pain point. Another showcasing a result. Another addressing a common objection. Another featuring a customer’s words. That’s four meaningfully different signals for Andromeda to test, regardless of whether they’re video, static or carousel.

Leave campaigns alone during learning. Once you launch or make significant changes, wait at least a week or 50 to 75 conversions before adjusting. The system needs time to calibrate. Tweaking bids or budgets mid-learning resets the process and wastes the data it’s already collected. We see this mistake constantly: a campaign launches on Monday, CPAs look high on Wednesday, someone panics and changes the budget. Now the learning phase resets and you’ve lost three days of data for nothing.

Look at who’s making your brand and who’s running your ads. If those are two different agencies, or worse, a freelance designer and a separate media buyer, you have a structural problem. Andromeda has made brand quality and ad performance interdependent. The people building your creative assets and the people optimising your campaigns need to be in the same room, working from the same strategy.

We’ve always structured Distl this way because it produces better work, not because an algorithm told us to. Andromeda just made the business case easier to explain. Branding, creative and paid social sit under one roof, working from the same briefs and the same brand systems. If you’re finding that your current setup isn’t delivering what it used to, let’s have a conversation about what’s changed and what to do next.

Becca Humphrey

Paid Media Service Lead