If Amazon's AI proposes a content change to one of your detail pages and you don't respond within 14 days, it publishes. Not "it gets flagged." Not "it waits for you." It publishes, on its own, to a live listing.
That's Amazon's own language in the Seller Central forums, and it's the part of yesterday's title chaos nobody is talking about. Everyone spent Monday watching their titles get compressed to 75 characters. The mechanism doing the compressing has been running since February 16, 2026, it isn't limited to titles, and its default answer is yes.
If you run $200K/mo on Amazon, this is the most important sentence in this post: your catalog review cadence is slower than Amazon's auto-publish clock. Most brands look at their listings monthly, quarterly, or when something breaks. The clock is 14 days. That gap is where rank goes to die quietly.
What happened
Amazon's Review Listing Changes (RLC) tool launched February 16, 2026 — a Brand-Registry-gated dashboard that shows Amazon-initiated changes to your detail pages over a 60-day lookback, plus what Amazon describes as "AI-powered content suggestions when we detect potentially incorrect or policy-violating content." Per Amazon's forum documentation, if a brand doesn't address a suggestion within 14 days, "it will be published automatically."
Access requires Brand Registry enrollment plus the Brand Representative selling role, granted under User Permissions → Brand Benefits. Non-registered sellers don't get the dashboard at all.
One more thing, hedged honestly because I can't confirm the current state: sellers in the forums reported roughly three weeks ago that the RLC dashboard URL redirects to the Manage Products page instead of opening, with no Amazon moderator response in the thread. It may be fixed by the time you read this. Go click it and find out — that's a 30-second test and it's step one below.
Why most brand owners will read this wrong
Dumb take #1: "This is a title thing that started Monday." No. The 75-character rule is one input into a system that's been proposing edits to your catalog for five months. The title deadline didn't open this door. It just turned the lights on in a room you'd been walking past since February.
Dumb take #2: "Amazon's suggestions are probably fine — they know the catalog better than my VA does." Sometimes they are. That's not the issue. The issue is architectural, and quality is beside the point.
Here's the real signal, and it's the same one I wrote about Friday with the Opus 5 model swap: this is a write path with a human gate whose default is approve.
Every catalog automation I've built has one non-negotiable rule — the human approves the write. Not reviews it, not gets notified about it. Approves it, affirmatively, before anything touches a live listing. I built that constraint into the attribute fill-rate tool because a tool that writes on silence isn't a tool, it's a liability with a UI.
Amazon built the inverse. The gate exists. It fails open.
And a gate that fails open isn't a gate. It's a notification you didn't read.
What actually changes for someone running $200K/mo
Your top 20 ASINs carry 70-80% of your revenue. That's where this lands, and here's the shape of it.
Rank moves before you see anything. An edit that removes an exact-match phrase you rank for doesn't announce itself. It surfaces as an organic slide 7-14 days later. CTR reads first — call it 2-3 weeks. CVR takes 30-60 days to stabilize into a number you'd trust. By the time the trend is legible in your reports, the change is 6 weeks old and you're diffing against a memory.
Then it gets misdiagnosed as a PPC problem. This is the sequence I've watched in takeover audits more times than any other single pattern: a catalog change degrades organic rank → organic traffic softens → the same ad spend now buys a bigger share of a smaller pie → ACOS climbs → the brand restructures campaigns. Three weeks of PPC work aimed at a problem that isn't in the campaigns. Sometimes they restructure twice.
The 14-day window is shorter than your process. This is the whole thing in one line. If your catalog gets a real look once a month, you are structurally guaranteed to auto-approve some percentage of changes without ever seeing them. Not because you're careless — because your cadence and Amazon's cadence are mismatched by design.
The role gate is a single point of failure. RLC needs Brand Registry plus the Brand Representative role. If the person holding that role left the company, or the notifications land in an info@ inbox nobody opens, your coverage is zero and the clock still runs. I've seen the registry admin sitting under a departed employee's email at brands doing eight figures. This is the same organizational gap, with a 14-day timer attached.
Non-registered sellers have no gate at all. If you're selling without Brand Registry, there's no dashboard, no 14 days, and no review. Enrollment stopped being a someday project.
What I'd do this week
1. Open the dashboard today and confirm it opens for you. Seller Central → Manage All Inventory → the listing changes view. If it redirects or 404s, open a support case now. The clock does not pause because the UI is broken, and "I couldn't access it" restores exactly zero rank.
2. Fix the role assignment before you fix anything else. Check who holds Brand Representative. Add a second person. Route the notifications to a monitored inbox with a human name on it, not a shared alias. Ten minutes, and it's the difference between having a gate and having a gate nobody is standing at.
3. Baseline before you review a single suggestion. Export the category listing report today — titles and key attributes, verbatim — and screenshot organic rank on your top 10 keywords for your top 20 ASINs. Do this first. Once edits start landing you'll want to prove what changed, and you can't diff against a file you never pulled.
4. Diff on a schedule shorter than 14 days. Weekly pull of the listing report for the top 20, compared against the baseline. It's a 10-minute job and it's the only thing that converts "silence is consent" back into an actual decision. Anything slower than weekly and the auto-publish clock beats you on some percentage of changes, permanently.
5. Write your rejection standard down in advance. Decide now, cold, what you'd reject: anything that removes an exact-match phrase you hold rank on, anything that alters a compatibility or fit attribute, anything touching a regulated field. Deciding in the moment on day 13 is how rubber-stamping happens — you're busy, it looks reasonable, you click through. A written standard means the decision was made by the version of you that had time to think.
What I'd ignore
The broken-URL story as the headline. It's a bug. It may already be fixed. The durable fact — the one that's true in September regardless — is the 14-day auto-publish default. Build your process around the mechanic, not the outage.
Panic-rewriting your catalog. A rushed catalog-wide edit carries more risk than the thing you're rushing to avoid. Start with the top 20 and let the rest wait for a real process.
Agencies pitching "AI listing change monitoring" as a new retainer line. It's a weekly report export and a file comparison. If someone quotes you a monthly number for it, ask them to describe the deliverable in one sentence and watch what happens.
Reverse-engineering the suggestion algorithm. You can't see the weights, Amazon isn't publishing them, and the thread-writers claiming to have cracked it are guessing. Your leverage is the diff and the rejection standard, both of which work regardless of what the model is doing.
The "Amazon is stealing your listings" discourse. It generates engagement and zero decisions. Amazon has been editing catalog data for years. What's new is a documented review window with a default, and defaults are an operations problem, not a betrayal.
The part that should bother you
The brands most exposed here are the ones with clean catalogs.
If your listings are a mess, you're already in there weekly and you'd notice an edit. If you did the work — filled the attributes, wrote the titles deliberately, got the fields right — you stopped looking. That's rational. Clean things don't need checking.
Except the assumption underneath is that clean stays clean unless you change it. That assumption stopped being true on February 16, and yesterday's title deadline just started a 14-day clock on a lot of catalogs at once.
Go click the dashboard. Then go look at what's already been published while you weren't looking — you get 60 days of history, and that window is the only free lookback you're going to get.