Marketing in 2026 means operating a system whose distribution layer someone else owns. Twenty years ago the marketer chose the medium, the placement, the audience and the message. Today, in their default automated campaign types (Advantage+ on Meta, Smart+ on TikTok), the platforms choose the placement and the audience in real time. That is why "creative is the new targeting" has become the shorthand of 2026: the ads you supply are now the main signal deciding who sees them.
What the marketer still controls has narrowed to three inputs: the creatives going in, the conversion data coming back, and the budget that connects them. The job is no longer to craft the one right message for the one right segment. It is to run a production and testing system:
- supply a steady flow of varied, realistic creative;
- feed the platform clean purchase data, so it optimises toward money rather than clicks;
- keep the structure simple enough that each ad set learns;
- read results on the metrics that reflect profit;
- retire creative the moment it fatigues.
Creative is also the biggest lever. NCSolutions analysed nearly 450 consumer packaged goods (CPG) campaigns across digital and TV. It attributed 49 percent of incremental sales lift to creative, more than reach, targeting and recency combined (14, 11 and 5 percent), a share unchanged since its 2017 study.
Generative AI has just made that lever cheap. The question is how to use it at a volume the platforms reward and a quality consumers accept, within the EU disclosure rules that apply from 2 August 2026.
I lecture at Business Academy Copenhagen and research how marketing teams, especially in Danish SMEs, adopt AI. This article tests Meta's and TikTok's claims against independent evidence and ends with an operating model for your own account.
Why is creative the new targeting on Meta and TikTok?
Meta's Andromeda-based delivery and TikTok's Smart+ campaigns now pick which ad each person sees, and they use the ads themselves to find the audience. Meta's own guidance names creative diversification, not niche targeting, as the best lever for finding audiences.
Meta defines creative diversification as "an advertising strategy where marketers create unique campaign assets tailored to different potential customer personas or use cases" (Meta for Business, December 2025). In practice that means distinct concepts for distinct buyers, not small tweaks of one idea. See how to plan distinct creative concepts for Meta broad targeting.
What Meta has published
In December 2024 Meta's engineering team described Andromeda, its new retrieval engine for ads (Meta Engineering). The system "is designed to maximize ads performance by utilizing the exponential growth in volume of eligible ads available to the retrieval stage".
Meta says Andromeda, rolled out in the second half of 2024, made its ad retrieval models 10,000 times more complex. In April 2025 its business blog drew the conclusion: the focus "has shifted from niche targeting to creative diversification as the best lever to find the most relevant audiences".
On the July 2026 earnings call, CFO Susan Li credited advances in Meta's user understanding models, plus the GEM ranking model and sequence learning. Together they "generated an 8.3% increase in ad clicks and a 15.7% uplift in conversions on Facebook". These are Meta's own, self-reported numbers.
What TikTok has published
TikTok's Creative Jumpstart handbook (August 2023) states that "creative refresh frequency is the #1 factor that determines the lifespan of an ad group" (TikTok Creative Center). Its Smart+ suite, announced at Advertising Week on 7 October 2024, automates targeting, bidding and creative selection across web, catalog and app campaigns (TikTok Newsroom).
The phrase "creative is the new targeting" itself is industry shorthand. It appears in neither platform's help documentation. The underlying claim, however, is now written into both platforms' own published guidance.
Meta and TikTok: the key numbers at a glance
Meta ad sets need about 50 results a week to exit learning; TikTok cites about 25 results or 7 days. TikTok also asks for a daily budget of at least 10 times target CPA and 3 to 5 ads per ad group.
| Rule | Meta | TikTok |
|---|---|---|
| Learning stabilises | About 50 results in the week after the last significant edit | About 25 results or 7 days; an older guide says 50 in 7 days |
| Ads per ad set | Fewer ads, but diverse assets in each ad set; 6 or fewer until at least February 2025 | 3 to 5 per ad group; at least 6 assets for Smart+ Web |
| Daily budget | No published rule; 50 purchases a week is about 7 times CPA a day (my arithmetic) | At least 10 times target CPA; Smart+ Web: ideally 30 times |
| Restarts learning | A new ad; edits to targeting, creative, optimization event or bid strategy; a pause of 7 days or more | Pausing campaigns or ad groups, and edits that retrigger learning; change one setting at a time, 2 days apart |
| Refresh creative | "Creative fatigue": cost per result at least double that of past ads | Results decline steadily, or daily new users are low |
| Sources | learning phase, significant edits, ad volume (2025 archive), creative fatigue | learning phase, ad delivery, bidding, Smart+ Web, Ace the Auction, creative best practices |
Why is my Meta ad set stuck in learning, and what resets it?
An ad set stays in learning until it reaches about 50 optimization events in the week after its last significant edit. Adding an ad, editing targeting, creative, optimization event or bid strategy, or pausing for 7 days or more restarts that count.
The usual causes are too few conversions for the budget, too many ad sets splitting them, or edits that keep resetting the clock. Meta says ad sets exit learning "as soon as they can deliver stably" (About the learning phase). That "usually occurs after about 50 results in the week after the ad set's last significant edit".
On TikTok, the troubleshooting guide (updated August 2026) still says to "give your ads a full 7 days to reach 50 conversions" (TikTok Ads Manager). The learning phase page (updated June 2026) says "volatility starts to decline after about 25 campaign results or 7 days" (About learning phase). Both are live, which suggests TikTok is lowering the bar as its models improve. Treat 50 as the conservative planning figure and 25 as the optimistic one.
How much daily budget an ad set needs
The threshold dictates budget. TikTok's bidding guidance says to "set the budget to be at least 10 times the CPA bid" (best practices for bidding strategies). For Smart+ Web campaigns on Target CPA or Minimum ROAS, its Smart+ guide asks for a daily budget of ideally 30 times historical CPA and "no less than 10 times".
With a EUR 30 target CPA:
- TikTok's rule puts the minimum daily ad group budget at EUR 300.
- Meta publishes no multiplier, but 50 purchases a week is EUR 1,500, or about EUR 214 a day: roughly 7 times CPA.
- A EUR 75 daily budget buys about 17.5 purchases a week, below both the 25 and the 50 threshold. It is very unlikely to exit learning.
What resets learning, and what does not
Meta's own list of significant edits is short:
- a change to targeting, creative, optimization event or bid strategy;
- adding a new ad;
- pausing the ad set for seven days or longer.
Budget changes are less clear-cut. In Meta's example, raising a budget from $100 to $101 "isn't likely" to restart learning, while a jump to $1,000 may. TikTok's learning phase page says to avoid pausing campaigns or ad groups, and edits that retrigger learning, until the phase ends. Its Ace the Auction guide adds: adjust one setting at a time, then "wait at least two days before making additional adjustments".
Why fewer, fatter ad sets learn faster
Meta's Performance 5 framework puts "simplify your account structure" first, and its learning phase page notes that "by combining similar ad sets, you also combine learnings". TikTok's troubleshooting guide says to "consolidate active ad groups with similar targeting settings into one ad group". Its creative best practices suggest "3-5 diversified ad groups per campaign". Fifteen ad sets with three purchases a week each starve all of them of data.
How to get an ad set out of learning
All but the last of these remedies come from Meta's guidance on ad sets marked Learning Limited:
- Merge ad sets until each one can plausibly reach 50 purchases a week.
- Broaden the audience rather than narrowing it.
- Raise the budget, or the bid or cost cap.
- Use a more frequent event as a bridge. If purchases are too few, optimising for add to cart is legitimate until purchase volume allows a switch back.
- Leave the ad set alone for seven days after the change. TikTok's Smart+ guide says the same: "run for seven days without any significant edits".
How many creatives should you put in one ad set?
Fewer than most advice suggests: 3 to 5 creatives per TikTok ad group (at least six assets for Smart+ Web), and fewer ads with diverse assets per Meta ad set. Meta's Advantage+ sales campaigns (formerly Advantage+ shopping) are the exception, testing up to 150 creative combinations.
Meta's help page on managing ad volume now says "decrease ads per ad set, but maintain diverse creative assets per ad set". One ad can hold up to 10 creative assets. Until at least February 2025 the page said "Use 6 or fewer creatives per ad set" (archived version). Past six ads, "there is little marginal benefit".
TikTok is more specific. Its creative best practices suggest "between 3-5 different creatives per ad group", and its Smart+ Web guide says to "add at least six creative assets during campaign creation".
Advantage+ shopping campaigns, since renamed Advantage+ sales campaigns, work differently. Meta's product page asks you to import all eligible existing ads so it can test "up to 150 creative combinations". That is a ceiling, not a target.
The case against piling everything into one ad set
Monica Shukla, a vice president at the agency Mile Marker, made that case in a January 2026 AdExchanger opinion column. Uploading 20 or more creatives per ad set, she argued, often works against the advertiser. When creative supply dramatically outweighs the available budget, she wrote, "learning slows, delivery fragments, and the system struggles to distinguish signal from noise".
The reconciliation: breadth at the concept level, depth at the variant level. That means several distinct angles across the account, a handful of ads per ad set, and a weekly cadence of retiring losers and adding new concepts. A creative testing matrix that separates concepts from variants makes that split explicit.
How many ad creatives do you need each week for Meta ads?
Plan volume per week, not per ad set, and scale it with spend. In Motion's 2026 benchmark, top-quartile Meta accounts launched 4.8 new creatives a week below $10,000 of monthly spend and 15.9 at $50,000 to $200,000.
Motion's Creative Benchmarks 2026 cover $1.29 billion of Meta spend across 578,750 creatives and 6,015 advertiser accounts, from September 2025 to January 2026 (methodology). The window includes the holiday season. "Top quartile" means the accounts with the most winning creatives within each spend tier (Motion, top 25 percent).
