Why Did My Facebook Ads Suddenly Stop Working?

An account runs fine for months, then a Monday arrives where CPA has doubled and nothing obvious explains it. The usual response is to change several things at once: new audiences, a budget increase, some fresh creative, maybe a bid cap. A week later performance is worse, and now nobody can tell which change caused what.

Diagnose before touching anything. Four causes account for nearly every one of these collapses, and three numbers separate them.

Why did my Facebook ads suddenly stop working?

Facebook ads stop working for four reasons, and the first diagnostic question is whether performance degraded or delivery collapsed.

Degradation looks like rising CPA across several days with spend roughly intact. Collapse looks like spend falling off a cliff, often overnight, with impressions near zero. They are different failures and confusing them wastes the first day.

Degradation is normally creative fatigue or auction pressure. Collapse is normally something structural: a learning phase reset from an edit, a bid cap sitting below the market clearing price, an ad rejection, or a policy flag on the account.

A restrictive cost cap is a common and quiet version. If your cap is twenty dollars and the auction is clearing at thirty-five, you have instructed the system to find something that does not exist, and it will politely spend nothing.

The three numbers that separate the cases are CTR, CPM and frequency, read against a 7-day rolling baseline rather than against yesterday.

What you see	Most likely cause	First move
CTR falling, frequency rising past 2.5	Creative fatigue	Launch new concepts, do not edit existing ads
CPM rising, CTR flat	Auction competition	Accept the cost, improve creative, or spend less
CTR flat, conversion rate falling	Site, checkout or pixel	Stop working on the ads entirely
Spend collapsed, impressions near zero	Learning reset, bid cap or rejection	Read the edit history and delivery column

The last row is the one people miss for longest, because Ads Manager presents a broken checkout as though it were an advertising problem.

Before any of that, open the edit history on the campaign. A surprising share of sudden collapses trace to a change somebody made and forgot about, and finding it takes thirty seconds.

Why are my Facebook ads not delivering at all?

Facebook ads that are not delivering at all, meaning zero impressions rather than expensive ones, almost always have an account-level or targeting-level cause you can confirm in Ads Manager in about ten minutes without editing anything. Work this list before you touch the creative, because every fix below is a settings change and none of them is solved by a new video.

Start with the platform change that is actually diagnosable, because there is exactly one. Meta spent 2025 consolidating detailed targeting interests, and the deadline landed quietly.

Deprecated interests became unselectable in new ad sets on 15 December 2025, and on 15 January 2026 any ad set still containing one stopped delivering altogether, per Social Media Today’s reporting of the advertiser notice and Jon Loomer’s write-up of the same announcement.

Meta did not pause those ad sets, reject anything or post an error banner on them. Delivery went to zero while the campaign carried on reporting as active.

Loomer’s instruction is the practical one: open the notice in Ads Manager, click “See affected ad sets”, and edit the ad set or the saved audience to drop the dead interest.

The accounts worst hit are the ones nobody has rebuilt in a year, because an evergreen ad set that has run untouched since 2025 is precisely the one carrying a retired interest.

Meta’s direction of travel explains why this keeps happening. It is removing manual controls because its own testing found the median cost per conversion improved by 22.6% when detailed targeting exclusions were taken away, as reported by Social Media Today.

Agree with that or not, the operational consequence is fixed: controls you built a campaign on can be withdrawn on a published deadline, and the ad set does not tell you when it happens.

What else stops a Facebook ad account from spending?

What else stops the account is almost always one of four things: a limit you set, a payment that failed, a decision made in ad review, or a setting that instructs the system to spend nothing. None of them appear in the performance columns a media buyer spends the day in, which is how they survive a long investigation.

Check your own spending limit first. When an ad account reaches its account spending limit, Meta pauses the ads and stops spending, and that amount does not reset on its own at the end of a month.

It sits there until somebody raises the limit, removes it, or resets the amount spent, and Meta’s help centre puts the delay before ads resume at roughly 15 minutes. A limit set two years ago to stop a runaway test is a common and very quiet cause of a Monday with no spend in it.

Payments come next. An expired card, a declined charge or a payment method hitting Meta’s cap on how many accounts it can be attached to all stop delivery outright, and Meta documents the fix and the prevention steps in its own payments guidance. This is worth checking before the campaign because the symptom is identical and the finance team, not the media buyer, usually holds the answer.

Then the review and restriction layer. A rejected ad, an ad set whose every ad was rejected, or a restriction applied at the account or Page level produces the same zero.

The delivery column names it, so read that column before forming a theory.

Meta also re-reviews ads that are already live, which is why an ad that ran for a week can be pulled back into the queue, and how long ad review takes and what to do when an ad is stuck in it covers the escalation route.

Last, the settings that instruct the system to spend nothing. A cost or bid cap under the auction clearing price, an end date that has passed, a schedule running in the wrong time zone, or an audience so narrow that it cannot be reached at your bid. All four are legitimate instructions and Meta follows them silently.

What you see	Likely cause	Where to look
Active campaign, zero impressions since January 2026	Ad set carries a deprecated detailed targeting interest	The Ads Manager notice, then “See affected ad sets”
All ads across the account stopped at once	Account spending limit reached, or a failed payment	Payment settings, billing
One ad set stopped, others fine	Rejection, restriction, or an expired schedule	The delivery column on that ad set
Spend falling well short of budget, not zero	Cost cap under the clearing price, or audience too small	Ad set bid strategy and audience size

Work that table top to bottom before you open the creative. Every row on it is a settings answer, and a new video fixes none of them.

How long does a Meta ad last before it fatigues?

A Meta ad typically shows creative fatigue after 3 to 5 days of active delivery, with CTR dropping 20 to 40% from peak by day 7, according to Adsights’ 2026 detection thresholds.

That is faster than most brands plan for. The common shape, according to Atria’s 2026 breakdown of Meta creative fatigue, is a 10-day decay curve: peak performance in days one to three, warning signs in days four to seven, and a 30 to 50% CTR drop by day ten.

Whole accounts hold up longer than individual ads, so most see meaningful fatigue at the account level within one to three weeks depending on audience size and spend rate.

Even the winners decay. Top-performing ads left running unchanged for five weeks lose roughly 38% of their effectiveness, and average campaigns are down 53% by week eight, per Ad Library’s 2026 ad fatigue benchmarks. The ad that carried your account in March is not the same asset in May, even though the file has not changed.

Now the correction that matters, because it means this whole section is the wrong diagnosis for most readers. The Interconnections’ 2026 benchmark, built on 368 creatives across 15 managed DTC accounts, 48,450,354 impressions and $685,791 of spend between April and August 2026, found that 80% of Meta creatives never reach 100,000 impressions and 89% never reach 250,000.

The median creative is live 18 days and delivers 10,665 impressions in its entire life. Most creatives do not fatigue.

They die of never working. If your ad never got past five figures of delivery, refreshing the creative because a dashboard said “fatigue” is treating a condition it never had.

Audience size sets the pace. Per AdsGo’s 2026 guidance, audiences under 200,000 are the ones at risk, and a pool that size saturates in days at moderate spend while a broad one takes weeks. This is why two brands spending identical budgets see completely different fatigue timelines, and why copying another brand’s refresh cadence rarely works.

The practical implication is a production question rather than a media question. If assets fatigue in one to three weeks, an account needs new concepts arriving continuously, and the right testing volume for your spend tier is what determines whether you have a replacement ready or a gap.

What frequency is too high on Meta?

On prospecting, a weekly frequency of 2.5 is the early warning and 3.5 is the point to refresh, per The Interconnections’ 2026 Meta creative fatigue benchmark. Those are 7-day figures, 2.5 for top of funnel and 3.5 for mid-funnel, from the same 368-creative dataset.

Retargeting behaves differently and should be judged separately. A retargeting pool sustains roughly 3 to 8 weekly exposures, and CTR falls 28 to 40% once frequency passes 7, per 2026 Meta frequency benchmarks, because the audience already knows you and repetition is doing a different job.

Applying one threshold across the whole account produces two wrong answers at once: it panics you about healthy retargeting and lets prospecting run well past the point of damage.

Frequency is the most useful confirming signal because it moves before revenue does. By the time CPA has visibly worsened, the ad has been fatiguing for days. A frequency chart pointed upward while CTR points down is a diagnosis, not a hypothesis.

Be careful about reading CTR alone, though. According to The Interconnections’ dataset, CTR on creatives that did reach volume fell 5.5% by the 100,000 to 250,000 impression mark and then held roughly flat, while CPA rose 19.6% by the 500,000 to 1 million mark and stayed there.

Cost is the metric that should trigger the refresh. A CTR that has stopped falling is not an ad that has stopped getting more expensive.

Frequency also has a self-inflicted cause worth ruling out. Audience overlap above 30% between ad sets means you are bidding against yourself in the same auction, per Niblin’s 2026 performance drop diagnosis, which raises your own CPM and inflates frequency for the people caught in the overlap. Meta’s audience overlap tool answers this in about a minute, and consolidating the overlapping sets usually fixes both problems together.

Did you break the learning phase without noticing?

An ad set needs roughly 50 conversion events in 7 days to exit the learning phase, and any significant edit restarts the count.

Significant means the creative, the optimisation event, the audience definition, or a large budget change. During learning, delivery is unstable and costs are typically worse, so the week after a well-intentioned overhaul often looks worse than the week that prompted it. Then the overhaul gets blamed on the wrong thing, and someone changes more.

Learning Limited status, where an ad set never accumulates the 50 events, is the chronic version. The usual cause is fragmentation: running more ad sets than the budget can feed.

Splitting a modest daily budget across four or five ad sets means none of them gets to the volume needed to optimise, and all of them perform badly, per AdsGo’s 2026 learning limited guide. Very narrow audiences under 200,000 hit the same wall even at reasonable budgets.

The fix is consolidation rather than addition. Fewer ad sets, broader audiences, and where possible an optimisation event that occurs often enough to reach 50 in a week. If purchases are too rare at your volume, optimising for a higher-frequency event upstream sometimes stabilises delivery, at the cost of a looser signal.

There is a discipline point buried in here. Batch your changes, make them at a consistent time, and give each set a full week before judging it.

Which changes actually cost you the week is a short and partly surprising list, including adding a new ad to an ad set that is already running, and it is set out in what counts as a significant edit and what it does to the learning phase.

Whether a particular edit actually cost you the week is also checkable rather than arguable, since an ad set’s learning status and the timestamp of its last significant edit are both fields Meta documents. Accounts edited daily never leave learning, and their owners conclude the platform is broken.

Before concluding anything, check the age of the thing you are judging: how long Facebook ads take to start working sets out when each number becomes readable, and a brand new account carries a setup and trust clock in front of that one.

What should you do in the first 48 hours?

In the first 48 hours you should read rather than edit: work the diagnosis in a fixed order and change one thing at a time, because a collapse investigated by simultaneous edits becomes permanently unexplainable.

Hour one is reading, not editing. Pull the edit history. Check for rejections and account-level flags. Look at CTR, CPM and frequency against a 7-day baseline, and check conversion rate on the site separately, ideally against a source outside Meta such as your ecommerce platform or analytics.

If the site conversion rate moved, stop working on the ads. A broken checkout, an expired discount code, a failed pixel deployment or an out-of-stock hero product presents in Ads Manager as an advertising problem, and no amount of creative fixes it.

If frequency is up and CTR is down, launch new concepts rather than editing existing ads, since editing resets learning while launching adds to the pool. If CPM is up and CTR is flat, the auction got more expensive and the honest options are to accept a higher cost, improve creative to win impressions more cheaply, or reduce spend until it settles. Nothing inside the campaign settings fixes external competition.

Then wait a week before judging. This is the hardest instruction to follow with a CPA chart pointed the wrong way, and it is the one that separates accounts that recover from accounts that thrash.

What mistakes make it worse?

Five common mistakes make a bad week worse, and the largest is turning everything off and rebuilding the account. Rebuilding is the most common response to a bad week and it discards every piece of learning history, which means starting from zero delivery stability while the underlying cause remains undiagnosed.

Do not raise budgets significantly as a first move. Increases above 30% often reset learning, per AdsGo’s 2026 guidance, and if the cause is fatigue, more budget buys more impressions of an ad people have already tuned out. Frequency rises, CPM follows, and the situation gets worse in a way that feels like bad luck.

Do not duplicate a winning ad set to scale it. Duplication creates audience overlap, and overlap above 30%, per Niblin’s diagnostic above, means the two sets bid against each other and raise your own costs.

Do not attribute the week to a platform-wide algorithm change without evidence. Sometimes true, rarely diagnosable, and never actionable. The available response is the same as for any auction shift: improve creative or spend less.

And do not swap a creative that is genuinely still working because a dashboard suggested a refresh. Fatigue is measurable, and the measurements above tell you when an asset is done. Retiring an ad early is a real cost, because the replacement is statistically unlikely to be as good.

So what should you check first?

Check three numbers before you change anything: CTR, CPM and frequency, each against a 7-day rolling baseline rather than against yesterday. Falling CTR with frequency past the 2.5 early warning on prospecting is creative fatigue, and the fix is new concepts rather than edits to existing ads.

Rising CPM with flat CTR is auction competition, which is external and not solvable inside the campaign settings. Collapsed delivery is structural: an edit that reset learning, a bid cap under the clearing price, or a rejection.

And if the ads look fine while conversions do not, the problem is on your site, where the ad account cannot see it.

The habit worth breaking is reaching for “creative fatigue” as the default explanation.

It is the most flattering diagnosis available, because it implies the ad worked until the audience wore out, and according to The Interconnections’ 2026 benchmark it is usually false: 80% of Meta creatives never reach 100,000 impressions and the median one is dead in 18 days on 10,665 impressions.

An ad that never accumulated delivery did not fatigue. It lost the auction, which is a different problem with a different fix, and swapping in a fresh version of the same idea will lose it again.

The metric that separates the two is hook rate, since an ad nobody watched past three seconds was never in a position to wear anybody out: what a good hook rate looks like, and how to fix one below the benchmark.

Where this stops applying: on a small, narrow audience under 200,000 people, genuine saturation happens in days at moderate spend, per AdsGo’s 2026 guidance, and the fatigue reading is usually correct. The broader your targeting, the more likely the real answer is that the creative was never strong enough to buy impressions cheaply.