If your Facebook and Instagram ad results have felt unpredictable lately, Meta Andromeda is very likely the reason. The AI that selects which ads have the chance of being displayed on a user’s page is called Meta Andromeda. This AI was unveiled by Meta in December 2024 and then gradually introduced into its applications during the year 2025. It basically shifts the responsibility of audience selection from your targeting options to the ad itself.
This guide explains the Meta Andromeda ad retrieval system in plain language and shows you what to do about it. At Tiju’s Digital Marketing Institute, this is now one of the most asked-about topics in our Meta ads classes, so we’ve answered the common questions here.
What Is Meta Andromeda, and How Does It Change Ad Targeting?
What is Meta Andromeda?
Meta Andromeda is Meta’s next-generation ad retrieval engine. For every person who opens Facebook or Instagram, it quickly picks a shortlist of relevant ads from millions of live ads. It uses a large deep neural network running on powerful GPU hardware, including NVIDIA’s Grace Hopper Superchip and Meta’s own MTIA chips.
According to Meta’s engineering team, Andromeda:
- Supports a huge jump in model complexity over the older system
- Improved retrieval recall by about 6%
- Improved ad quality by around 8% on selected segments
How does Meta Andromeda work?
Meta Andromeda is the first step in a three-step process:
- Retrieval: Andromeda scans the ads index and does candidate ad selection, narrowing tens of millions of ads to a few thousand.
- Ranking: Meta’s ad ranking system, powered by models like Meta Lattice and the Meta GEM model, predicts how likely the person is to click, watch, or buy.
- Auction: The top ads enter the ad auction, where bid, predicted action rate, and ad quality decide the winner.
If your ad doesn’t pass retrieval, it never reaches ranking. It simply doesn’t exist for that user.
What is the difference between ad retrieval and ad ranking?
Meta ad retrieval decides which ads can compete. Ranking decides which of them wins. Think of retrieval as the job shortlist and ranking as the final interview. The retrieval stage vs. ranking stage is really about recall and precision: retrieval tries not to miss any good ad, and ranking picks the very best one.
How does this change ad targeting?
Andromeda reads your ad itself as a targeting signal. It looks at visuals, text, format, landing page, and how people react, then matches this with audience signals from user behaviour.
A yoga mat video showing a busy mom stretching at home will be retrieved for different people than a video of an advanced athlete. This is why marketers now say creative is the new targeting. It’s the biggest of the recent Facebook ads targeting changes, and every student at the best digital marketing institute in Kerala should understand it before running real budgets.
Why Did Meta’s Traditional Ranking-First Ad System Stop Working?
The old system stopped working well because it couldn’t handle the huge growth in ads and the loss of tracking data. It depended on advertisers narrowing the audience first, then asked a limited ranking system to do the heavy lifting.
Here’s what broke it:
- Too many ads: AI creative tools made it easy to produce far more variations than the old system could handle.
- Signal loss: After Apple’s iOS 14.5 changes, the data behind manual interest targeting became less reliable.
- Narrow audiences: Tight, detailed targeting gave the system fewer people to learn from, pushing costs up.
- Hardware limits: Older CPU-based setups couldn’t run complex models fast enough for every user.
GPU-based retrieval fixed this, making real machine learning ad delivery and AI-powered ad personalisation possible at scale. That’s why the Meta ads algorithm 2026 looks so different from a few years ago. In Tiju’s digital marketing course, we teach this shift before opening Ads Manager, because it explains why old tactics feel broken.
Does Interest Targeting Still Work on Facebook Ads?
Yes, but it now works more like a suggestion than a rule. With Advantage+ audience, Meta can go beyond your chosen interests whenever it finds better-performing people.
Is broad targeting better than interest targeting on Meta?
For most advertisers, broad targeting now performs as well as or better than interest targeting. Broad targeting Facebook ads give Andromeda room to find the right people using your creative and conversion data.
A quick look at Meta interest targeting vs. broad targeting:
- Broad works best when you have good conversion data, enough budget, and varied creatives.
- Interests still help for new accounts with little data or very niche products.
- Lookalike audiences now work best as signals, not strict walls.
What matters more now is data quality. This signal-based targeting, also called algorithmic targeting, depends on a well-set-up Meta Pixel, the Conversions API (CAPI), and clean first-party data like customer lists. When these are strong, audience expansion works in your favour.
Why Creative Similarity Is Quietly Capping Your Reach?
Creative similarity caps your reach because Andromeda treats very similar ads as one ad. Ten near-identical ads may only get the reach of one.
What is creative similarity in Meta ads?
Creative similarity refers to how alike your ads are in terms of visual appearance, messaging, and form. Meta and their partners have provided information on how near duplicate ads are grouped under one Entity ID for retrieval purposes.
These would likely count as the same creative:
- One video with three different headline colours
- One image with five slightly different captions
- The same product shot in different crops
Changing a button colour isn’t creative diversity. To Andromeda, it’s the same ad in a different shirt.
Why are my Facebook ads not reaching new audiences?
Usually because your creatives are too similar, so Andromeda keeps showing them to the same small group. Warning signs include:
- Frequency climbing every week
- CPM rising while results stay flat
- Reach saturation, where new reach stops growing
- Falling CTR and conversion rate
We see this often when students bring real accounts to class at our digital marketing academy in Kerala. The fix is rarely a new interest. It’s a genuinely new idea.
How to Build Creative Variety Without Losing Your Brand Identity?
Change the concept, but keep the brand system fixed. The story, angle, and format should change, while your logo, colours, fonts, and tone stay recognisable.
Vary these:
- Creative angles: problem-solution, testimonial, founder story, comparison, offer-led
- Creative formats: static vs video ads, carousels, Reels ads, and phone-shot UGC ads
- Personas: a student, a working parent, a business owner
Keep these consistent:
- Logo placement and colours from your brand guidelines
- Font style, brand voice, and call-to-action style
A modular creative approach helps. Build a kit of fixed brand pieces, then swap in new hooks, scripts, and visuals. This protects brand consistency while giving Andromeda real creative variety. That’s the core of smart creative diversification Meta ads strategy, and a hands-on module in Tiju’s AI digital marketing course, where students use AI tools to create variations faster.
How many creatives should I run per ad set on Meta?
Most advertisers do well with around 8 to 15 genuinely different concepts per ad set. Meta allows up to 50 ads per ad set, but ten distinct ideas will beat fifty near-copies. Dynamic creative can help test combinations, as long as the raw ingredients are already different.
Why Creative Fatigue on Meta Is Now an Organizational Problem?
Meta creative fatigue is now an organizational problem because fixing it needs a steady flow of new ideas, not a quick tweak in Ads Manager. Each distinct creative finds its audience fast under Andromeda, so it also wears out fast.
That makes it a team issue:
- Creative volume must stay high every month
- The creative production workflow must replace tired ads quickly
- Media buyer and creative team alignment turns data into better ads
- A clear creative team structure and content calendar for ads keep output steady
Brands that win treat creative operations like a production line, with weekly reviews feeding the next batch. That’s what performance creative really means. Our trainers at the best digital marketing academy in Kerala often say the ad account is only as strong as the team behind it.
How often should I refresh Facebook ad creatives?
Most brands should add new concepts every two to four weeks, sooner with heavy spend. Your ideal creative refresh rate depends on budget and audience size. Watch for these signs of ad fatigue:
- Rising frequency on prospecting campaigns
- CTR dropping over 7 to 10 days
- CPA increasing with no change in offer
- ROAS slipping on former winners
How do I avoid creative fatigue on Meta?
Introduce new ideas before the old ones become irrelevant. Maintain a reserve of proven hooks, introduce new ideas at regular intervals, recycle winning angles using new mediums, and leverage the power of AI to generate drafts.
Why Your Hook Rate Matters?
Hook rate shows whether your ad stops the scroll in the first three seconds. If people don’t stop, nothing else matters, and Andromeda gets a weak signal about who your ad is for.
How do you calculate hook rate?
Hook rate = (3-second video views ÷ impressions) × 100
If the video receives 10,000 impressions but 2,800 3-second video views, the hook rate will be 28%. This is sometimes referred to as the thumb-stopping ratio by
What is a good hook rate for Facebook ads?
Around 25% to 30% is a common benchmark, and above 30% is usually strong. Pair it with hold rate, which shows how many kept watching after the hook. A high hook rate with a low hold rate means your opening works, but your story loses people.
Quick ways to improve it:
- Show the result or problem in the first frame
- Add bold on-screen text for silent viewers
- Open with a face, a question, or a surprising visual
- Test three to five openings on the same video
Tracking hook rate Facebook ads data closely is a habit we build early at Tiju’s Digital Marketing Institute, because it’s the quickest way to spot a strong idea.
How to Adapt Your Creative Strategy for Meta Andromeda?
Simplify your campaigns, go broad, fix your data, and focus most of your energy on diverse creative.
How to structure Meta ad campaigns after Andromeda?
- Consolidate campaigns and ad sets so each gets more data.
- Go broad, keeping only essential limits like location and age.
- Fix signals by checking Pixel, CAPI, and event setup.
- Feed diverse creative with distinct angles and formats.
- Test concepts, not tweaks. Good ad creative testing compares big ideas, like testimonial vs. demo.
- Review weekly using hook rate, CTR, CPA, and ROAS.
Do Advantage+ campaigns work better with Andromeda?
Often, yes. Meta Advantage+ campaigns give the algorithm maximum freedom, which suits how Andromeda works. They perform best with solid conversion data and plenty of creative variety.
How does Andromeda affect Facebook ad performance?
Andromeda rewards accounts with diverse creative and clean data, and it punishes narrow targeting and repetitive ads. A strong creative strategy for Meta ads is now a repeatable system of research, briefs, production, and testing. Learning this with live campaigns is why picking the best digital marketing course in Kerala matters.
Conclusion: Your Ads Are Now Your Targeting
Meta Andromeda has changed Facebook and Instagram advertising at its core. The system finds audiences mainly by understanding your ads, so creative quality and variety now matter more than any interest you select.
To sum up:
- Andromeda decides which ads get a chance to compete
- Broad targeting with strong data usually beats narrow interests
- Similar creatives get grouped and limit your reach
- Creative fatigue is a team and workflow issue
- Hook rate is your earliest signal of a winning idea
Ready to master Meta ads the modern way? Tiju’s Digital Marketing Institute is recognized as the best digital marketing institute in Kerala for practical training in Advantage+ campaigns, creative testing, CAPI setup, and AI-powered ad production. Tiju’s AI digital marketing course helps beginners and professionals build the skills brands are hiring for now. Visit tijusacademy.com to explore the course and book your seat.
Frequently Asked Questions:
A: Meta Andromeda is Meta's AI-powered ad retrieval engine. It picks a shortlist of relevant ads for each Facebook and Instagram user from millions of live ads. Meta announced it in December 2024 and rolled it out across its apps during 2025.
A: Meta Andromeda works in three steps. First, it retrieves a few thousand relevant ads from the ads index. Next, Meta's ad ranking system predicts which ads a person is most likely to act on. Finally, the top ads enter the ad auction, where the winner is shown.
A: Ad retrieval decides which ads are allowed to compete, while ad ranking decides which ad wins. Retrieval focuses on recall, so no good ad is missed. Ranking focuses on precision, so the best ad is chosen. If an ad fails retrieval, it never reaches ranking.
A: Yes, but interest targeting now works more like a suggestion than a strict rule. With Advantage+ audience, Meta can show your ads beyond your chosen interests whenever it finds people who are more likely to respond.
A: For most advertisers, broad targeting performs as well as or better than interest targeting. It gives Andromeda more room to find the right people using your creative and conversion data. Interest targeting can still help new accounts or very niche products.
A: Because Andromeda reads your ad itself to decide who should see it. It studies visuals, text, format, landing page, and user reactions. So your creative now tells Meta who your ad is for, more than your targeting settings do.
A: Creative similarity is how alike your ads are in visuals, message, and format. Near-duplicate ads can be grouped under one Entity ID, so Andromeda treats them as a single ad. This limits reach, since ten similar ads may perform like one.
A: Your ads are likely too similar, so Andromeda keeps showing them to the same small group. Common signs include rising frequency, higher CPM with flat results, reach saturation, and falling CTR and conversion rate. The fix is usually a genuinely new creative idea.
A: Most advertisers do well with around 8 to 15 genuinely different concepts per ad set. Meta allows up to 50 ads per ad set, but ten distinct ideas will perform better than fifty near-identical copies.
A: Keep your logo placement, colours, fonts, brand voice, and call-to-action style consistent. Then change the story, angle, format, and people in each ad. A modular creative approach makes this easier by combining fixed brand pieces with new hooks and visuals.
A: Most brands should add new creative concepts every two to four weeks, and sooner if they spend heavily. Refresh when you see rising frequency, CTR dropping over 7 to 10 days, increasing CPA, or falling ROAS on former winners.
A: Meta creative fatigue needs a steady flow of new ideas, not a quick fix in Ads Manager. Under Andromeda, creatives find their audience fast and wear out fast. This makes creative volume, production workflow, and team alignment essential.
A: Hook rate = (3-second video views ÷ impressions) × 100. For example, 2,800 three-second views from 10,000 impressions gives a hook rate of 28%. It shows whether your ad stops people from scrolling.
A: A hook rate of around 25% to 30% is a common benchmark, and above 30% is usually strong. Pair it with hold rate to see whether viewers keep watching after the first three seconds.
A: Consolidate campaigns and ad sets, use broad targeting with only essential limits, and fix your Meta Pixel and CAPI setup. Then add diverse creative, test big concepts instead of small tweaks, and review performance weekly.
A: Often, yes. Meta Advantage+ campaigns give the algorithm maximum freedom, which suits how Andromeda works. They perform best when you have solid conversion data and plenty of creative variety.
A: Tiju's Digital Marketing Institute offers practical training in Advantage+ campaigns, creative testing, CAPI setup, and AI-powered ad production. Tiju's AI digital marketing course is designed for both beginners and working professionals. Visit tijusacademy.com to learn more.
