Key takeaways
- Meta's only published count is at least 20 diversified ads in Advantage+ shopping campaigns (November 2024), with new ads launched weekly and 20–30% of budget set aside for testing.
- Jon Loomer tested 25 and 50 ads per ad set in 2025 and saw no clear advantage, so genuine difference matters more than raw ad count.
- A new ad counts as diversified when it changes the persona, the angle, the format or the visual style; color swaps and small headline edits likely read as the same ad.
- Motion's 2026 benchmark data shows only about 5–8% of ads become winners, so plan launch volume around how many tries it takes to find one winner.
- For most accounts under $50K a month, 4–10 genuinely different concepts per ad set plus a few new ads every week is a sensible starting point.
Meta's only official number is "at least 20 diversified ads" in Advantage+ shopping campaigns (November 2024), plus new ads every week and 20–30% of budget for testing. Practitioners who tried 25 to 50 ads per ad set saw no clear gain. For most small accounts, 4 to 10 genuinely different concepts per ad set beats 20 near-copies.
The rest of this guide covers what Meta has said, what counts as "different," how to plan concepts with a simple grid, how many to launch at your spend level, and how to read the results.
What Meta has actually said about ad counts
Meta has published one number. In November 2024, Meta for Business recommended three things:
- "Maintaining at least 20 diversified ads in Advantage+ shopping campaigns"
- "Dedicating 20-30% of your overall ad budget towards testing new ads"
- Launching new ads routinely "on a weekly basis," which Meta said is what top performers do
Note the scope. That 20-ad figure is for Advantage+ shopping campaigns (now Advantage+ sales), where all ads sit together. It is not a rule for every manual ad set.
In April 2025, Meta defined creative diversification as "creating a wide range of ad creatives with different themes, messages, and visuals to cater to diverse audience segments." It called diversification "the best lever to find the most relevant audiences." It gave no count.
Meta's old suggestion to keep manual ad sets to about six ads was removed from its help page, as Jon Loomer reported in July 2025. And Meta's October 28, 2026 engineering post on the next generation of Andromeda (the update the industry calls "Andromeda 2.0") added no new setting, account structure or creative count. If you want the background on how retrieval works, read our explainer on Meta Andromeda.
So any "Andromeda needs exactly 15 ads" rule you see on social media comes from a vendor or an agency. Meta hasn't published it.
What happened when practitioners added more ads
Jon Loomer, a long-time Meta ads educator, wrote in November 2025 that he "tested 25 and 50 ads per ad set" and "didn't see any clear advantage to it." His conclusion: "Quality still beats quantity."
In January 2026, Monica Shukla of Mile Marker wrote in AdExchanger that "the belief that uploading 20 or more creatives per ad set is universally beneficial" is a misunderstanding that "often works against the advertiser." Her advice is to match creative volume to budget and funnel stage. Upper-funnel campaigns can carry more concepts if the budget lets each one get traction. Lower-funnel campaigns, where conversion signals are strongest, may do best with fewer.
The reason is simple arithmetic. Meta concentrates spend on a handful of ads no matter how many you upload. Motion's 2026 benchmark of its customers' accounts found about 55% of spend went to winning creatives. On a $100-a-day ad set, the 15th and 25th ads mostly sit idle. They add production work without adding learning.
What counts as "diversified"
Meta's creative differentiation guide warns that when creative is too similar, its system may not recognize it as different and may deliver it to the same audiences, which limits your reach. In October 2026, Meta's engineers described how its multimodal models read each ad's "text and images" and turn it into tokens, where "similar items share tokens." Two ads that look and read alike give the system very little new to work with. (More on that in does Meta read the text in your ads.)
Loomer put it bluntly in October 2025: "Unless they're distinctly different enough, any variation is unlikely to make a meaningful impact on performance." Text changes, he said, should be significant, not subtle.
Use four levers. A new ad should change at least one, and ideally two.
| Lever | Counts as different | Probably reads as the same ad |
|---|---|---|
| Persona | New buyer type with a different pain point (gym regular vs. parent on the school run) | Same buyer, different first name in the quote |
| Angle / message | New reason to buy (price, convenience, proof, identity) | Same claim, reworded |
| Format | New layout archetype (testimonial vs. us vs. them vs. offer banner) | Same layout, new background color |
| Visual style | Lo-fi phone photo vs. clean studio shot vs. text-only | Same photo, cropped or filtered |
A quick test: shrink your ads to thumbnail size and put them side by side. If you can't tell two apart in a second, Meta probably can't either.
Build a diversification grid
The easiest way to plan genuinely different ads is a grid. Put formats down the side and personas across the top. Each cell is one concept.
Here is an example for a hypothetical insulated water bottle brand, using six formats from our library.
| Format | Gym regular | Office commuter | Parent on school runs |
|---|---|---|---|
| Testimonial | Real review about the bottle surviving gym bags, over a gym-floor photo | Real review quote on a desk shot | Real parent review over a lunchbox flat lay |
| Us vs Them | Bottle vs. vending-machine plastic bottles | Bottle vs. paper cups from the coffee machine | Bottle vs. leaky kids' bottles |
| Before & After | Gym bag full of empties vs. one bottle | Desk covered in cups vs. one bottle | Car floor with juice boxes vs. one bottle each |
| Offer | Bottle plus straw-lid bundle | "Desk set" bundle | Family 3-pack discount |
| Statistics | One substantiated number (hours cold) | Refills per workday | Dishwasher cycles tested |
| Native | Notes-app style "gym bag essentials" list | UGC-style phone photo on a desk | Lo-fi photo from the school-run car |
You won't run all 18 cells. Pick 4 to 6 to start. Moving across a row changes the persona. Moving down a column changes the format. Moving diagonally changes both, which gives you the most distinct ads.
Static formats make this cheap. Motion's 2026 benchmark ranked "text only" and "product image with text" as the top two of 15 asset types by hit rate, ahead of high-production creative. A static concept takes an hour, not a week, so you can fill more of the grid. Our guide to static ads covers how to build each format, and static vs. video ads covers when video earns its higher production cost.
If you don't have a designer, a template library helps here. In Static Ad Templates you can filter by format (Testimonial, Us vs Them, Before & After, Offer/Sale, Statistics, Native and 11 more) and open any layout in Canva, so each grid row starts from a different proven structure.
How many concepts to launch at your spend level
Motion published weekly launch volume by spend tier in its 2026 Creative Benchmarks. This is vendor data from Motion's own customer accounts (578,750 creatives across 6,015 Meta accounts, September 2025 to January 2026). Motion counts a "winner" as a creative that spends at least 10x the account's median creative and at least $500. It measures where Meta put the money. It says nothing about ROAS.
| Monthly Meta spend | New creatives per week, average (Motion) | New creatives per week, top 25% (Motion) | Winner share (Motion) | Live concepts per ad set (our starting point) |
|---|---|---|---|---|
| Under $10K | 2.8 | 4.83 | 3.7% | 4–6 |
| $10K–$50K | 4.1 | 8.09 | 6.2% | 6–10 |
| $50K–$200K | 6.67 | 15.95 | 7.3% | 10–15 |
| $200K–$1M | 11.24 | 31.11 | 8.1% | 15–25, split across ad sets |
| $1M+ | 18.85 | 54.64 | 8.2% | 20+, split across ad sets |
The last column is our recommendation, not Motion's data. It is sized so each ad can get meaningful spend at that budget.
Run the numbers for a small account. At 2.8 launches a week and a 3.7% winner share, you'd find roughly one winner every two months. Moving to the top-quarter pace of about 5 a week nearly doubles that. Volume matters, but only if every new ad is genuinely different. Twenty variations of one idea give you one idea's worth of learning.
A weekly refresh cadence
Meta's guidance is weekly launches and 20–30% of budget on testing. Here is a simple routine for an account spending under $50K a month.
- Monday: review last week. Note which concepts got spend and which hit your CPA or ROAS target. Turn off clear losers.
- Pick 3 to 5 new cells from your grid. Make about half of them net-new concepts (a new persona or angle). Make the rest iterations on last week's winner (same angle, new format, or same format, new hook).
- Build and name them consistently. Use a pattern like
Persona_Angle_Format_v1, so you can later see which formats and personas win across many tests. - Launch into your testing setup. Many buyers run a separate testing campaign with equal budget per concept, then move winners into their main Advantage+ campaign. Meta's own Creative Testing tool can also reserve part of an ad set's budget to test up to 5 ads, though it can't test existing ads directly (it works with duplicates).
- Keep the testing budget at 20–30% of spend. Below that, new ads rarely get enough spend to tell you anything.
Expect most new ads to fail. That's how the process works.
How to read the results
Expect a low hit rate. Motion's rule of thumb from its 2026 data is that 5–8% of ads win, so 20 launches produce roughly 1 to 1.6 winners. If 1 in 15 new ads becomes a winner, you're on track.
Uneven spend is normal. Meta will put most of the budget behind one or two ads. Low spend on an ad is a signal, but check its CPA before you call it a loser. Sometimes a low-spend ad converts well and deserves its own test.
Judge winners on business results. Spend share tells you what Meta likes. CPA, ROAS and conversion rate tell you what makes money. Give each test enough spend to reach a decision; a common rule of thumb is 1 to 2 times your target CPA before cutting an ad.
Watch the Delivery column. Ads Manager can label ads "Creative limited" or "Creative fatigue" when cost per result rises against that ad's own history, per Jon Loomer's write-up. Meta's help content on fatigue, as quoted in third-party write-ups such as soku.ai's, calls for new creative that is "materially different from the original." A recolored version of a tired ad is still a tired ad.
Look for patterns, not single winners. After a month of consistent naming, sort by format and persona. If "Native × parent" keeps winning, build more ads in neighboring cells (Testimonial × parent, Native × office commuter).
Policy checks before you scale variety
More concepts means more chances to trip a policy. Four common ones:
- Persona copy. Meta's personal attributes policy bars copy that implies you know the viewer's health, finances or other personal traits. "Busy mornings?" is fine. "Struggling with your weight?" is not.
- Before & After. Weight-loss before/after imagery is not allowed. Cosmetic before/after needs 18+ targeting and real results.
- Testimonials. The FTC's rule on fake reviews took effect October 21, 2024 and covers AI-generated reviews. Use real customer words only. Our guide to testimonial ads covers how to source and format them.
- Mockup formats. AirDrop, notification and fake-button designs risk rejection under Meta's non-existent functionality policy. Keep them clearly non-clickable.
What to do next
- Write down your 3 main personas and the single pain point each one cares about most.
- Build a grid with those personas across the top and 6 formats down the side.
- Pick 4 to 6 cells that differ in both persona and format, and launch them this week.
- Set aside 20–30% of your budget for testing, and add 3 to 5 new ads every Monday.
- After four weeks, sort results by format and persona, and fill the cells next to your winners.
Sources
- Meta for Business, "Three steps to optimize your performance with creative diversification" (Nov 19, 2024), https://www.facebook.com/business/news/three-steps-to-optimize-your-performance-with-creative-diversification (accessed Nov 2026)
- Meta for Business, "The Creative Advantage: unlocking the power of diversification with Meta Andromeda" (Apr 22, 2025), https://www.facebook.com/business/news/the-creative-advantage-unlocking-the-power-of-diversification-with-meta-andromeda (accessed Nov 2026)
- Meta for Business, "Creative differentiation" guide, https://www.facebook.com/business/m/small-business/creative-differentiation (accessed Nov 2026)
- Engineering at Meta, "Meta Andromeda: Powering the Next Generation of Ads Retrieval Through Model-System Co-Design" (Oct 28, 2026), https://engineering.fb.com/2026/10/28/ml-applications/meta-andromeda-next-gen-ads-retrieval-gpu-co-design/ (accessed Nov 2026)
- Engineering at Meta, post on GenRecs tokenization in ads recommendation (Oct 19, 2026), https://engineering.fb.com/2026/10/19/ml-applications/genrecs-tokenization-transforming-ads-recommendation-at-meta-with-stable-model-vocabulary/ (accessed Nov 2026)
- Jon Loomer, post on Meta Andromeda and ad retrieval (Nov 10, 2025), https://www.jonloomer.com/meta-andromeda-ad-retrieval/ (accessed Nov 2026)
- Jon Loomer, post on Meta Andromeda and creative diversification (Oct 20, 2025), https://www.jonloomer.com/meta-andromeda-creative-diversification/ (accessed Nov 2026)
- Jon Loomer, note on the removed six-ads-per-ad-set guidance (Jul 2025), https://www.jonloomer.com/qvt/no-more-six-ad-limit-per-ad-set/ (accessed Nov 2026)
- Jon Loomer, post on creative fatigue in Meta ads, https://www.jonloomer.com/creative-fatigue-meta-ads/ (accessed Nov 2026)
- Jon Loomer, post on Meta's Creative Testing tool, https://www.jonloomer.com/meta-creative-testing/ (accessed Nov 2026)
- AdExchanger, Monica Shukla (Mile Marker), Data-Driven Thinking column on what Meta's Andromeda update changes (Jan 28, 2026), https://www.adexchanger.com/data-driven-thinking/what-metas-andromeda-update-actually-changes-and-what-it-doesnt/ (accessed Nov 2026)
- Motion (Runneth), "Meta Ads in 2026: How Many Creatives Do You Actually Need to Launch?" (recorded Mar 2026), https://runneth.com/library/talk/meta-ads-in-2026-how-many-creatives-do-you-actually-need-to-launch/ (accessed Nov 2026)
- Motion (Runneth), Creative Benchmarks 2026 methodology page, https://runneth.com/library/research/creative-benchmarks-2026/methodology (accessed Nov 2026)
- Motion (Runneth), Creative Benchmarks 2026 top asset types page, https://runneth.com/library/research/creative-benchmarks-2026/top-asset-types (accessed Nov 2026)
- Motion (Runneth), Creative Benchmarks 2026 key benchmarks and insights, https://runneth.com/thumbstop-pulse/cb2026-key-benchmarks-and-insights (accessed Nov 2026)
- Federal Trade Commission, "Federal Trade Commission Announces Final Rule Banning Fake Reviews and Testimonials" (Aug 14, 2024), https://www.ftc.gov/news-events/news/press-releases/2024/08/federal-trade-commission-announces-final-rule-banning-fake-reviews-testimonials (accessed Nov 2026)
- Meta Transparency Center, Advertising Standards: privacy violations and personal attributes, https://transparency.meta.com/policies/ad-standards/objectionable-content/privacy-violations-personal-attributes/ (accessed Nov 2026)
- Meta Transparency Center, Advertising Standards: non-existent functionality, https://transparency.meta.com/policies/ad-standards/deceptive-content/non-existent-functionality/ (accessed Nov 2026)
- Meta Transparency Center, Advertising Standards: health and wellness, https://transparency.meta.com/policies/ad-standards/restricted-goods-services/health-wellness (accessed Nov 2026)
- soku.ai, guide to Meta Andromeda (quotes Meta Help Center fatigue guidance), https://soku.ai/blog/meta-andromeda (accessed Nov 2026)
Frequently asked questions
How many ads should I put in one Meta ad set?
+
There is no official per-ad-set number for manual campaigns. Meta's only figure is at least 20 diversified ads in Advantage+ shopping campaigns. For accounts spending under $50K a month, 4–10 clearly different concepts per ad set is a practical range. Add more only when your budget lets each ad get enough spend to show whether it works.
What is creative diversification on Meta?
+
Meta defines creative diversification as creating a wide range of ads with different themes, messages and visuals for different audience segments. In practice it means each new ad targets a different persona, angle or format, so Meta's delivery system has distinct options to match to different people instead of near-copies of one idea.
How many creatives should I test per week?
+
Motion's 2026 benchmarks show accounts under $10K a month launch about 2.8 new creatives a week on average and 4.8 in the top quarter, rising to about 4.1 and 8.1 for $10K–$50K accounts. Meta recommends launching new ads weekly and spending 20–30% of budget on testing. Three to five new ads a week suits most small accounts.
Does changing the headline or background color count as a new creative?
+
Probably not in any meaningful way. Meta's creative differentiation guidance says creative that is too similar may not be recognized as different and may reach the same audiences. Jon Loomer advises that text differences should be significant, not subtle. Change the persona, the message and the visual together if you want Meta to treat the ad as new.
Why does Meta only spend on one or two ads in my ad set?
+
That is normal. Meta concentrates delivery on the ads it predicts will perform best. In Motion's 2026 benchmark data, about 55% of spend went to winning creatives. Uneven spend is expected. Check CPA or ROAS on the low-spend ads before turning them off, and replace losers with genuinely different concepts.