“No performance warning, no rectification period—just woke up to a grayed-out store.”

This is a sentence repeatedly mentioned in Amazon seller communities recently. Since the second half of 2026, Amazon has launched a new wave of large-scale account reviews based on Section 3 of the Business Solutions Agreement, which sellers call the “Section 3 sweep.” Unlike the usual “warn first, punish later” process, this round of sweeps is triggered directly by the AI risk-control system. By the time sellers receive the email, listings have already been removed and funds frozen.

Why are veteran sellers more likely to be hit?

Many sellers are confused: they did not fake reviews or sell counterfeits, and their front-end listings are fully compliant, so why were they swept? The answer lies in a fundamental change in review logic.

In the past, Section 3 mainly targeted obvious violations such as fake reviews and counterfeits, and sellers knew what the problem was. But in this round, AI no longer scans only front-end content; it conducts afull-lifecycle retrospective on accounts—category changes, variation merges and splits, image edits, backend keyword stuffing, title and attribute adjustments over several years will all be retrieved and reviewed, and even long-deleted old ASINs will be re-examined.

This means the longer a seller has operated, the more historical operation records there are, and the higher the probability of hitting a landmine.

Many veteran sellers are now having old mistakes dug up: placing products in the wrong category to bypass review, attaching restricted products without qualifications as variations to high-weight old listings to piggyback traffic, modifying image parameters to avoid qualification checks, frequently changing core title attributes after new product launch... These once seemingly “flexible” operating tactics are now all judged as “evading platform detection” under AI retrospective review, directly triggering maximum Section 3 penalties.

“Even if the product is compliant, as long as there are traces of review evasion in the backend, it will trigger a ban.” A cross-border service provider analyzed, “AI only looks at objective operations; unintentional mistakes cannot be used as a defense.”

Video verification: a life-or-death gate with only one chance

After being swept, the seller’s only way out is tocomplete legal-person video interview verification within 14 days. Failure to complete it will result in permanent store closure. And this video verification has only one chance—no second attempt.

From recent hands-on feedback, the depth of platform review far exceeds seller expectations. Many veteran sellers report that the current video verification focus has long gone beyond business licenses, ID cards and other registration documents. “Even if all qualification documents are genuine and complete, it does not mean passing the review. The platform’s core examination is whether the registered legal representative truly understands the store business and participates in operations, rather than merely being a name on the qualifications.”

According to seller community feedback, video verification mainly revolves around three dimensions:

Operational awareness: reviewers focus on the most basic information—what category the store mainly sells, which products are promoted, FBA or FBM, how many team members there are, whether the seller has a clear grasp of daily ad spend and profit/loss.Supply chain: they press for sourcing details—supplier names, cooperation duration, procurement cycles, per-order quantities, whether there are contracts and invoices, and which freight forwarder is used.Capital flow: this has been significantly strengthened this year—which account receives payments, corporate or third-party, how funds are used after returning, whether store deduction items and profitability are clear.

It is worth noting thatoverly mechanical “scripted answers” are more likely to raise suspicion. Some service providers observed that some legal representatives memorized interview answers compiled online perfectly, yet still failed. Reviewers observe response state, including thinking pauses and naturalness of expression, to judge whether the person on camera genuinely handles the business or is a “stand-in” trained in a short-term crash course.

One seller’s experience is quite representative: the store’s legal representative was a friend who did not participate in operations; the credit card and withdrawals were all handled by the seller himself. He admitted in a forum: “If I let my friend attend the video interview, how do we explain the credit card and withdrawals? How can this be solved, and what are the odds?”

Legal representatives who are only nominal and do not participate in operations, mismatches between credit cards and payment account entities, multiple stores sharing network devices—these common operating patterns among small and medium sellers are exactly the high-risk signals flagged by the AI risk-control system.

How should sellers respond after receiving the notice?

First, do not rush to submit an appeal. First read the notification email thoroughly and distinguish which type of review was triggered. If the email contains the keyword “Evade Detection,” it means historical operations on the product side were retrospectively reviewed; if it emphasizes “identity verification,” “fraudulent activity,” etc., it is a verification of account entity authenticity. The appeal logic for the two is completely different—do not apply a generic template.

Second, organize a complete chain of evidence. For identity verification, prepare the latest business license, legal representative ID, bank statements from the past 90 days, and credit card statements. All documents must be clear scans, with names, addresses, and company names exactly matching the backend registration information. For product authenticity issues, prepare purchase invoices from the past 3 months, supplier business licenses, logistics documents, and brand authorization chain files.Invoice amounts must match store sales—a situation where sales are 100,000 but purchase invoices are only 10,000 will be directly judged as false evidence.

Third, train the legal representative to understand business details, but do not memorize templates. In the video interview, the legal representative needs to be able to answer what products the company sells, who the suppliers are, approximate profit margins, and who uses the receiving account. It is okay if the answers are not professional enough, but the logic must be coherent and the details must be accurate. Speaking truthfully in one’s own familiar language is far more effective than reciting standard answers.

For sellers who have not yet received a notice, self-inspection is the best defense. Quickly review listing adjustment records, variation relationships, ASIN creation records, and keyword edit logs from the past 1–2 years, and promptly clean up potentially risky links.

Compliance is no longer a bonus—it is the bottom line for survival

The signal from this round of Section 3 sweeps is very clear: Amazon’s risk-control system has shifted from “post-event punishment” to “full retrospective tracking + preventive strikes.” AI first bans the store, and then the seller must prove innocence. The burden of proof has completely shifted to the seller. A green performance dashboard does not mean safety; historical operation records are what AI is actually scanning.

Looking at the trend from 2026, Section 3 reviews are gradually moving from sudden sweep waves to normalization. For sellers, rather than rushing to appeal after a ban, it is better to keep every operation compliant and every document complete in daily operations. In the AI review era, history does not disappear—it will only be retrieved again.