There was a time when Facebook advertising felt almost unfairly cheap. The targeting was precise enough to reach a very specific type of person. The CPMs — the cost to reach a thousand people — were low enough that even small brands could run meaningful campaigns on modest budgets. The return on ad spend was high enough that performance marketers were, depending on who you asked, either geniuses or just lucky.

That era is firmly in the past.

Meta advertising costs have been rising steadily for two years, and in 2026 the increases are significant enough that brands which built their entire customer acquisition model around cheap Facebook ads are being forced to rethink their unit economics entirely. CPMs that used to cost five dollars now cost twelve. Campaigns that produced a 4x return on ad spend eighteen months ago are now producing 2.5x on the same creative and audience targeting. The budgets are the same. The results are not.

Understanding why this is happening — the specific structural reasons, not just “more competition” — is the first step to actually doing something about it.

The three things pushing Meta ad costs up

1. There are more bidders for the same amount of attention.

The most straightforward explanation is supply and demand. The number of businesses advertising on Meta has grown substantially, but the number of hours people spend on Facebook and Instagram has not grown at the same pace. More money is chasing the same pool of human attention. In an auction-based system, that competition simply produces higher clearing prices.

This is not a problem that Meta can really solve — it is a consequence of the platform’s own success as an advertising destination. And it means that any easy CPM advantage that existed in earlier years is structurally gone. The market has corrected.

2. Meta’s AI system now controls targeting — and sophisticated AI on both sides of the auction is expensive.

Meta has shifted its ad delivery almost entirely to an AI-controlled system called Andromeda. Instead of advertisers manually selecting audiences and Meta serving their ads to those audiences, Andromeda decides in real time which ad to show to which person, based on a vast pool of signals about what that person is likely to respond to.

Here is the counterintuitive part: this system penalises advertisers whose creative is stale or generic. If your ad has been running for several weeks with diminishing engagement, the algorithm interprets that as a signal that the ad is not resonating. It reduces that ad’s reach and raises your effective CPM as a result. The cost of running a tired creative is no longer just lower click-through rates — it is actively higher costs per impression.

At the same time, competitors who have learned to play the game well are now feeding Andromeda with continuously refreshed, diverse creative. The AI on their side is also getting smarter. When sophisticated AI-driven bidding competes against other sophisticated AI-driven bidding, the clearing price goes up for everyone.

3. Privacy changes reduced the data quality that made Meta targeting efficient in the first place.

Before Apple’s iOS privacy changes, Meta could track user behaviour across apps and websites with high accuracy. It could see that someone who was served an ad for a pair of shoes later visited the shoe brand’s website and made a purchase. This data allowed the system to find more people likely to convert, which made campaigns more efficient.

After iOS 14.5, a large portion of that cross-app tracking disappeared behind consent walls that most users do not accept. Meta’s system now has a weaker signal about who converts and who does not. To compensate, it casts a wider net — showing ads to more people to find the same number of converters. A wider net, at auction, costs more. The efficiency loss from privacy changes is real, ongoing, and has not been fully recovered by the various measurement workarounds that Meta has introduced since.

What to actually do about it

Rising costs do not mean Meta advertising stops working. For many brands it remains the most effective paid customer acquisition channel at scale. But the strategies that made it cheap in the 2020 era will not recover those economics. Different approaches are needed.

Treat creative as your primary performance lever, not your targeting.

This is the single most important shift. In the old Meta, you won on audience targeting — the more precisely you found the right person, the better your results. In today’s Meta, the algorithm is doing most of the audience targeting work itself. What you can still control is the quality and variety of your creative.

Successful advertisers in 2026 are producing more creative variations than they ever did before — not just different headlines, but fundamentally different visual approaches, hooks, formats, and tones — and letting Meta’s AI figure out which variations work best for which audiences. The brands spending money on creative production and creative testing are consistently outperforming the brands spending the same money on more precise manual targeting.

Simplify your campaign structure.

Andromeda performs better with more data in one place. An account with three broad campaigns and large budgets gives the AI more signal to work with than an account with thirty narrow ad sets each with small budgets. Fragmented account structures were a best practice in the era of manual targeting. In the AI-delivery era, they are a cost multiplier — the system does not have enough data in each isolated bucket to optimise efficiently, so it hedges by bidding higher.

Get Conversions API working before you do anything else.

Meta’s Conversions API lets you send purchase signals directly from your server to Meta, bypassing the browser entirely. Because iOS privacy restrictions operate at the browser level, CAPI maintains attribution data that browser-based tracking loses. Brands running CAPI alongside their standard pixel typically see more conversions appearing in their Meta reports — not because more conversions are happening, but because the data that was previously invisible is now being captured. More signal means better optimisation. Better optimisation, at scale, produces lower effective cost per acquisition over time.

The friction you need to prepare for: If you implement these changes simultaneously — restructure your campaigns, refresh your creative, add CAPI — expect a four to six week period where performance looks unstable. You have reset the learning period for the algorithm by making significant structural changes. The system needs new data to find its optimal configuration, and during that period your cost per acquisition will likely look worse than before. Do not panic and revert. The instability is temporary. The improvement — if your new creative is genuinely better and your structure is simpler — should be durable.

Meta advertising costs more than it used to. That is not going to change. The question is whether you are paying that higher price for a well-optimised system that still delivers profitable returns, or paying it for a poorly structured account that is working against itself.