Facebook Ads Digest, September 2026: Verification, Assign Partner, Costly Rejects

The autumn storm did not surprise anyone with its force. A ban wave across the turn of August into September has repeated for several years running. What was new is that inside a single month Meta moved the point at which it decides whether to trust you three times, and not once was that point the ad itself.

How the Month Actually Went

The first week was the worst: one or two accounts surviving out of ten, domains dying overnight, part of a batch gone before it ever started spending.

On the 9th came a pause. Bans almost stopped and the platform allowed duplicating and reuploading again. The pressure did not disappear, it changed shape: conversion went down, CPM went up, and daily budget limits started arriving.

The middle of the month brought a second wave of much the same force, with farm account survival sitting around two in ten. The final week was noticeably calmer: launches became predictable again, conversion came back, and campaigns started living long enough to be worth reading.

What to plan around is not the moment the storm ends, but whether your infrastructure survives a week in which one account in five makes it.

The New Confirmation Step Before Publishing

Early in the month an extra step appeared before publishing. A window titled "Ad review needed" asks you to check the budget, the URLs and the start and end dates, then tick a box saying "I trust this ad and it's correct". It is not a ban and not an account restriction: confirm it and the ad runs normally.

Later the same mechanism surfaced as a reject in its own right, with the status showing and no stated reason behind it.

This is not tighter moderation. It is a transfer of responsibility: by ticking that box the advertiser signs off that the link goes where it claims to go.

A Reject Now Costs You at the Account Level

Ad accounts started showing a warning banner for a high number of rejects. One observation from the field is worth testing on your own accounts: where people responded to the banner by immediately requesting a review, the accounts kept running, and where the banner was simply dismissed, they died within a couple of days. The sample is small, so treat it as a hypothesis, but requesting the review costs two clicks.

In parallel something else became clear, and it changes how you diagnose. Rejects depend heavily on geo: the same creative that collects refusal after refusal in one market passes quietly in another. Campaigns uploaded minutes apart inside the same account can also land differently.

Together this rewrites an old reflex. A reject used to cost you the time of a reupload. Now it costs the account's standing, and the conclusion "the creative is bad" is wrong about half the time.

Assign Partner, and Why Access Operations Became the Riskiest Thing You Do

Meta rolled out Assign Partner across all ad accounts: access to specific accounts is handed over through a link, and whoever accepts it chooses which Business Manager to place them in. The first days came with bans; after that it settled down and worked.

And in the same month everything else around access became a trigger. Creating an extra invite inside a BM started reliably producing a ban on the BM together with the ad account. An automatic ban arrived on logging into an old BM and setting a password. The normal transfer of BM and ad account access to another account stopped working.

The platform simultaneously shipped a more convenient way to hand out access and made every manual access operation suspicious. So those operations now get planned in batches and kept away from an active run of traffic.

The Selfie Storm Reached the Layer Nobody Talks About

On the 7th a wave of video selfie requirements began. Social accounts dropped out en masse at the verification step and BMs got blocked on interaction, while the ad accounts themselves stayed alive.

The most telling case happened one floor down. One buyer keeps a separate pool of accounts used to seed comments under ads, sorted by geo and gender. When comments were needed for a new geo, every account touched went straight into selfie verification. There is a second observation in that story worth keeping: freshly added accounts in the anti-detect browser drop out noticeably faster than aged ones.

That same week, registering with Meta for Developers began requiring a phone number up front.

Why This Is One Line, Not Five Separate News Items

Across 2025 Meta blocked 3.5 billion accounts, catching suspicious profiles early, before any activity at all. The number matters less than what follows from it: the platform assesses trust not in an individual ad, but in the whole cluster of accounts and behaviour behind it.

That explains all of it at once. The confirmation box, because responsibility shifts onto a person. The selfie, because what is being checked is the human, not the creative. The reject banner, because a history of refusals became a property of the account. And the least comfortable part: a ban arrives for a behavioural pattern even when the specific ad breaks nothing.

And Yet the Month Still Worked

While some were counting losses, others were posting results in the very same weeks.

Spain held up as a strong gambling geo, with a funnel where click-to-install ran around 35%, click-to-registration around 20% and click-to-deposit around 9%. South Africa showed a positive trend and one genuinely useful observation: iOS converted there no worse than Android, against the common assumption. Turkey barely rejected anything at all, though registration-to-deposit around 12% did not satisfy everyone. Egypt started without bans or rejects on a plain recipe: a day of rest for the accounts, an aged fan page, apps, and fresh creatives.

The most telling case is a different one. One buyer put together over a hundred deposits across five funnels in two weeks, and explained it as "I have a plan and I stick to it": five to ten accounts launched daily, funnel tested, best creative scaled, analysis before any decision. In the worst stretch for months, what won was discipline rather than luck.

And one more thing that is easy to walk past. Talk of Facebook shutting the door has been going for about five years. In that time the rules changed, the easy schemes disappeared and the cost of entry went up, but the source did not go anywhere. What changes is not whether the traffic exists, but what is required of whoever comes for it.

What September Rewrote in Daily Work

1. See the reject banner, request the review immediately. Dismissing it costs more than two clicks do.
2. A reject with no explanation means checking the geo before the creative.
3. Do access operations in batches, away from an active run. Invites, transfers and logins into old BMs now trigger bans on their own.
4. The confirmation box is not cause for panic, but it is cause to reread the URL, because you are now signing off on it.
5. Count survival across the batch, not per account. Two in ten is a planning baseline, not an emergency.

Summary

• The month ran in three phases: a hard first week, a pause on the 9th that swapped bans for a drop in conversion, a second wave mid-month, and a calmer finish.
• The new confirmation window is not moderation but a transfer of responsibility for budget, URLs and dates onto the advertiser.
• Rejects acquired a cost at account level, and their cause is often the geo rather than the creative.
• Meta shipped Assign Partner and in the same month made manual access operations a ban trigger.
• 3.5 billion accounts blocked in a year shows the direction: what gets judged is the cluster of accounts and behaviour, not the single ad.
• Despite all of it the month closed with working launches, where the deciding factor was discipline.

If you would rather get through months like this on structure than on stamina, we work in this mode every day and teach how to assemble the process around a specific product. Get in touch and we will go through your case.