Facebook ads stopped working? Find the cause, then fix it

Why have my Facebook ads stopped working?

When Facebook ads stopped working last week but worked last month, something changed, and you can usually find it. The usual suspects are tracking, your offer or landing page, the audience and the ad itself. After those come frequency, your budget and a reset of Meta's learning phase. Check them in that order, because the early checks are quick and rule out the causes that would make every later fix pointless.

The worst move is to change five things at once, because even if results recover, you won't know which change worked. Each big edit can also push Meta's delivery system back into learning, which tends to make results shakier for a while. Find the cause first, then make one change and give it time.

This guide gives you the seven checks and where to look in Ads Manager for each one. You also get what to fix first, two prompts for your AI, a worked example and a diagnosis log you can reuse next time.

The seven checks, in order

Before you start, export two periods from Ads Manager: the last 7 days, and a 7-day stretch from when things were going well. Use the same columns for both: amount spent, results, cost per result, impressions, reach, frequency and clicks. Most of the checks below compare the two periods side by side.

1. Is it really down, or is tracking broken?

Ads Manager only knows about a sale if your website tells Meta about it. If a site update removed or broke that tracking code, the ads can look dead while sales carry on. Compare Ads Manager with a record that can't break in the same way, such as your shop's own order list, your booking system or your bank statements.

• If your real orders held steady while Meta's reported results fell, suspect tracking before anything else.
• In Events Manager, Meta's test events tool lets you open your site and watch events arrive as you click through. Go through a test purchase or sign-up and check that each step shows up.
• Meta's help page notes that new event data can take 30 minutes or more to appear in reporting, so a quiet last hour isn't a reason to panic.

2. Did your offer or page change?

Ads send people to a page, and if that page got slower, the price went up, a discount ended or stock ran out, the ad gets the blame anyway. Click your own ad on your phone and go all the way to the payment step. Then ask what changed on your side during the week results dropped, including anything a colleague or a website developer did.

Your numbers hold a useful clue here. If clicks held steady while sales fell, the problem usually sits after the click, on the page or in the checkout. If clicks fell as well, look at the ad and the audience instead.

3. Did the audience change?

Narrow audiences eventually run out of new people to reach, and changed targeting, a new excluded list or a shorter country list can all shrink who sees the ad. Compare the ad set's audience settings with how they looked in the good period. Also check whether two of your own ad sets now target the same people, which makes your ads compete with each other.

4. Is the ad itself worn out?

People stop responding to an ad they've seen many times, which Meta calls creative fatigue. For eligible ad sets with one creative, Meta shows a status in the Delivery column. It reads "Creative limited" when the cost per result is above your past ads but less than double. It reads "Creative fatigue" at double or more (as of November 2026).

You can spot fatigue yourself before any status appears. Look at your best ad's click-through rate, the share of people who saw the ad and clicked, over several weeks. If it slid steadily while nothing else changed, the ad is probably tired.

5. Is frequency climbing?

Frequency is the average number of times each person saw your ad, and Meta's definition of frequency works it out as impressions divided by reach. Meta also says that if performance drops while frequency rises, your audience may be experiencing ad fatigue.

There's no single "too high" number that fits every business, so compare your own good period with your bad one. If frequency rose sharply while results fell over the same days, you've found a strong suspect.

6. Did you change the budget or make big edits?

Look at what you changed, and exactly when. Meta's page on significant edits and the learning phase lists the edits that send an ad set back into learning. They include any change to targeting, the ad creative or the optimisation event. Adding a new ad, changing the bid strategy and pausing for 7 days or more also count. A budget change may count too, depending on its size: Meta's example says $100 to $101 isn't likely to matter, while $100 to $1,000 may.

7. Is the ad set stuck in learning?

Meta's guide to the learning phase describes it as the period when the delivery system is still working out how to deliver an ad set. During it, results are less stable and the cost per result is usually higher. Ad sets usually leave learning after about 50 results in the week after the last significant edit. An ad set that can't get enough results to get there shows "Learning limited" in the Delivery column.

If your drop started right after an edit and the status says "Learning", the honest answer may be to wait. Meta's own advice is to avoid editing during learning unless you have a clear reason to think the edit will help.

What to look at in Ads Manager

Ads Manager can show you most of the evidence for the seven checks, as long as you add the right columns through "Columns" and then "Customize columns". Menus move from time to time, so if a name differs from the one below, search the column list.

If every ad set dropped on the same day, look for a cause that affects the whole account, like tracking, the website, a price change or a seasonal shift. If only one ad set or one ad dropped, focus your checks there.

Let your AI read the numbers with you

Your AI is good at comparing two exports and noticing which numbers moved together, but it has no way of knowing what you changed on your website. Give it both, the numbers and a plain list of changes on your side, so it can match the timing of each change against the results.

You are a careful Meta ads analyst. My Facebook ads stopped working and I want the cause before I change anything.

I'll give you:
- Export A, a good period: [paste or attach, with dates]
- Export B, the bad period: [paste or attach, with dates]
- My real orders or leads from my own system for both periods: [numbers]
- What changed on my side (site, prices, stock, offer, audience, budget, new ads, pauses): [list with dates]
- Delivery column statuses I can see now: [e.g. Learning limited on ad set 2]

Work through these checks in this order, and stop to ask me if you need something:
1. Tracking: do Meta's results and my real orders move together?
2. Offer or page: did clicks hold while results fell?
3. Audience: did reach or targeting change?
4. Creative fatigue: did click-through rate fall on the main ads?
5. Frequency: did it rise as results fell?
6. Edits and budget: which changes could have restarted learning, and when?
7. Learning phase: how many results has each ad set had since its last big edit?

Output:
- A table: check, evidence from my numbers, verdict (likely cause, possible, ruled out).
- The single most likely cause, and how sure you are, in plain words.
- ONE first change to make, and what to leave alone while it runs.
- What to check again, and when.

Rules: base every verdict on my numbers or my change list. If the data can't tell, say so. Don't suggest more than one change at a time.

What should you fix first?

Fix things in the same order you checked them, because each fix depends on the ones before it. A fresh ad can't help while tracking is broken or the checkout is down, since you won't be able to see whether it works.

1. Broken tracking or a broken page: fix these today, because until you do, your numbers can't tell you anything else.
2. An offer that changed: if a discount ended or the price rose, the ad now promises something different. Update the ad to match the page, or change the page back.
3. A tired ad: add a new ad with a different idea rather than a recoloured copy of the old one. Meta's creative fatigue page suggests a new ad with a materially different image or video, and notes that keeping the original running may get more results than pausing it.
4. A shrinking audience: widen the audience, or combine ad sets that target similar people so they stop competing.
5. Too many edits: stop editing, make the one change your diagnosis points to, and then give it a week before you judge it.

If your diagnosis says "tired ad", this prompt writes the brief for a replacement while keeping the reason to buy that made the old ad work.

My best Facebook ad is tired: its click-through rate fell and frequency rose while nothing else changed. Help me brief a replacement that keeps what worked.

The tired ad: [paste the primary text, headline and describe the image or video]
Its results by week: [paste]
My product facts: [paste fact sheet]
Customer words, if I have them: [paste reviews or comments]

Do this:
1. Say in one line what this ad's reason to buy was.
2. Suggest 3 new ads that keep the same reason but change the idea a viewer would notice first: a new hook, a new picture or a new format. Make each one clearly different from the old ad and from each other.
3. For each, write the hook, a one-line picture or video idea, and the primary text.
4. Say which one to launch first and why, in one line.

Rules: use only my facts. Don't repeat the old hook with new words. No claims the fact sheet can't prove.

For a deeper look at refreshing a fading ad, see our guide to ad creative testing, which shows how to test the new ideas fairly against each other.

Worked example: an imagined drop for Liquid Death

Liquid Death sells canned water and other drinks in aluminium cans, which it pitches as an alternative to plastic bottles, and we'll use it to walk through the method. Everything below is an imagined scenario with made-up example figures. It is not Liquid Death's data, and Liquid Death has no link to this article.

The situation (example figures)

Say a sparkling water ad set cost about $20 per purchase for a month, and then climbed to about $34 over the last two weeks. Nobody touched the ad itself, but the marketer now wants to rewrite everything from the ad to the landing page.

Going through the checks

1. Tracking: the shop's own order count fell by about the same share as Meta's results, so the drop is real and tracking is probably fine.
2. Offer and page: the price is the same, the product is in stock and the checkout works on a phone. Clicks fell along with sales, which points to something before the click.
3. Audience: the targeting is the same as it was in the good month.
4. Creative fatigue: the main video ad's click-through rate slid week by week, and the Delivery column now shows "Creative limited" on it.
5. Frequency: weekly frequency went from about 1.8 to about 4.1, while reach barely grew.
6. Edits and budget: the daily budget went from $200 to $600 on day three of the bad stretch. A jump that size may have pushed the ad set back into learning.
7. Learning phase: the ad set shows "Learning" and has about 30 results since that edit, short of the roughly 50 it needs.

What the diagnosis said

The main cause is a tired ad shown too often to the same people. The budget jump made it worse by adding more showings to the same audience and restarting learning. The first change is to add one clearly different new ad while keeping the old one running. The budget stays where it is for a week, and then the results get read again.

Notice what the method prevented. The marketer was about to rewrite the page and change the audience as well, yet neither was the cause, and both changes would have reset learning again.

A diagnosis log you can reuse

Keep one row per check in a shared sheet. It takes about ten minutes, and the next time results drop, you can see what you found last time and which fix worked.

Common mistakes, and what to do instead

• Changing everything at once: you can't learn what fixed it, and each big edit can restart learning, so make one change and then wait a week.
• Trusting Ads Manager over your till: Meta can only count what your site reports to it, so always check your real orders first.
• Judging on one or two days: daily numbers jump about for no lasting reason, so compare full 7-day stretches instead.
• Pausing the old winner straight away: Meta notes that keeping the original ad running alongside a new one may get more results, so let the new ad prove itself first.
• Calling a recolour a new ad: the same hook on a new background is still the same ad to the viewer, so change the idea people notice first.
• Raising the budget to "push through": a big jump can restart learning and show the tired ad to the same people even more often.

Where AI Blocks helps

Each item below is free: download the skill or copy the prompt, then use it in your own AI with the numbers you exported.

• Ad performance troubleshooter: paste your campaign results to get the weak link, the root cause and the fix to try first.
• Ad performance change investigator: give it current and past numbers to get a ranked cause, a confidence score and what to check next.
• Meta delivery troubleshooter: share at least seven days of ad data to get a cautious diagnosis, one change to test and a comparison plan.
• Fading ad refresh planner: give it a fading winner's numbers to get the cause, ranked revival moves and 10 hook briefs.
• Conversion tracking setup and repair planner: define what counts as a sale or lead, and get a plan to set up or repair your event tracking.
• Meta ads daily checker: check spend pacing, winners, money wasters and tired ads from your own numbers in a few minutes a day.
• Meta Ads results reviewer: review results against your targets to get ranked findings, tracking gaps and proposed fixes.

If the cause turns out to be the copy, our guide to Facebook ad copy shows how to write the replacement from real facts. If the cause is the page, work through the landing page checklist instead.

Questions people ask

Why did my Facebook ads stop converting suddenly?

A sudden drop on a single day usually points to something that broke, such as tracking, the website, the checkout or a big edit that restarted learning. A slow slide over several weeks points more towards creative fatigue and rising frequency. Check tracking and your page first, since those checks are quick and rule out the most.

What is Facebook ad fatigue?

Ad fatigue is what happens when people have seen the same ad so often that they stop responding, and Meta calls it creative fatigue. The signs are a falling click-through rate and a rising cost per result while frequency climbs. The fix is a new ad with a clearly different idea, because more budget only shows the tired ad more often.

How long does the Facebook learning phase last?

Meta says ad sets usually exit learning after about 50 results in the week after the last significant edit. How long that takes depends on your budget and on how often your chosen result happens, and if the ad set can't get there, Meta shows "Learning limited".

Should I turn off ads that stopped working?

Not straight away, because you should diagnose the cause first. If the problem is fatigue, add a fresh ad alongside the old one, since Meta says keeping the original running may get more results than pausing it. Turn off an ad once a better one has clearly taken over, or when it costs more than a sale is worth to you.

Does increasing budget fix Facebook ads that stopped working?

Rarely. If the ad is tired, more budget shows it to the same people even more often, and a big jump can also restart learning. Fix the cause first, then raise the budget later in gradual steps once results are steady.