Amazon Ads Agent and Auto-Optimization: What Sellers Lose and Gain

At unBoxed on September 29, 2026, Amazon merged the sponsored ads console and Amazon DSP into a single platform: Ads Agent. Alongside it came campaign auto-optimization, a campaign intelligence dashboard, and a set of AI features that collectively hand Amazon's algorithm decisions that used to be yours.

Amazon's ad business did $19.8 billion in Q2 2026, up 26%. That context matters when reading the performance claims below — these tools are built to increase spend as well as results, and those two things are not the same.

Here is what is actually launching, when, what the numbers do and do not prove, and where keeping manual control is still the right call.

What Ads Agent is

One platform, three campaign types:

• Sponsored Ads
• Full Funnel Campaigns
• Display, Video and Audio

The structural change is that sponsored ads and DSP no longer live in separate consoles. For anyone running both, that is a genuine workflow improvement.

Full Funnel Campaigns — the big trade

You set one budget and choose products. Amazon's AI then handles targeting, creative and pacing across both sponsored and DSP formats. Currently US open beta.

The stated trade-off is blunt: less segment-level control. You are exchanging the ability to see and steer which audiences, placements and formats are working for the convenience of one budget line.

For a seller with one brand and a simple catalogue, that may be a fair trade. For anyone managing distinct audiences, flagship versus clearance ASINs, or defending specific placements, it is not — because when performance drops you cannot see which segment caused it.

The launch calendar

Auto-optimization: what it actually does

Arriving in the US in January 2027, auto-optimization works inside your existing campaign settings rather than as a new campaign type. Switched on, it:

• Expands product targeting to similar and complementary products and categories
• Adds relevant keywords
• Pauses weak keywords
• Spots seasonal demand shifts

Amazon's example: a planter brand had a search term added automatically when searches for it spiked.

The controls: one click to enable, every action shown with its spend and conversions, and any action editable or reversible at any time. Amazon has not stated whether it will be on or off by default — which is the single most important unanswered question about it. Check that setting the week it launches.

Read the performance claim carefully

Amazon's beta numbers: first sale 40% faster, sales up 13%, from internal US data, July–September 2026.

Note what is measured. Speed to first sale is a launch metric, not a profitability metric. Sales up 13% is revenue, not margin. Neither figure says anything about ACoS, TACoS or contribution profit — and a system whose levers are "expand targeting" and "add keywords" will almost always increase sales, because it increases spend.

If you enable it, judge it on TACoS and contribution margin, not on the metrics Amazon chose to report.

The other numbers, with the caveats attached

The second row deserves attention because it is the most quoted and the most misleading: shoppers who click prompts convert better and spend more than those who do not. That is a statement about who clicks prompts, not about what prompts do to them. Self-selected high-intent shoppers behave like high-intent shoppers.

The features worth switching on now

Not everything here is a trade-off. Three are straightforwardly useful:

• Campaign intelligence dashboard (October 2026, all US advertisers) — surfaces targeting, bidding and budget recommendations with a visual of likely impact, and you apply, edit or skip each one. Nothing applies automatically. That is recommendations without loss of control.
• Ads Data Manager — available now, globally, all campaign types, no extra budget. Upload first-party data once and reuse it across ad products. Confirm your privacy policy covers the use and that you hold consent.
• Review Requests (open beta through October, ASINs under 1,000 reviews) — you pay only when a shopper clicks to submit a review. For a product under the review threshold going into Q4, that pricing model is hard to argue with.

Sponsored Prompts: you are already enrolled

Worth flagging because it requires no decision from you and most sellers have not noticed. Sponsored Prompts in Alexa for Shopping reached US general availability on March 25, 2026, and eligible campaigns are enrolled automatically with no extra setup.

Clicks are charged at your existing CPC — there is no separate bid. Amazon writes the question and answer itself, from your detail page, your reviews and your campaign data.

There is no described opt-out switch. The control that exists is pausing prompts — which matters for low-margin or low-stock products, where you do not want an AI-written answer driving traffic you cannot profitably serve. Compare prompt spend and conversion against the parent campaign in the Prompts report.

Where manual control still wins

Automation is good at breadth and bad at judgement. The decisions worth keeping are the ones that depend on information Amazon's algorithm does not have:

1. Anything margin-dependent. Amazon optimises to sales and ROAS. It does not know your COGS, your peak fulfillment fees or your true contribution margin — so it cannot know your real break-even.
2. Inventory-constrained ASINs. Auto-expansion driving traffic to a product you will stock out of before Black Friday is actively harmful.
3. Brand defence. Branded terms are a strategic hold, not a performance line — and an algorithm pruning "weak" keywords may cut them.
4. Launch campaigns with deliberate ACoS tolerance. Buying rank at a planned loss looks like failure to an optimiser.
5. Segment-level reads. If you need to know why performance moved, do not run Full Funnel.

The practical position

Turn on the dashboard. Use Ads Data Manager. Take Review Requests if you qualify. Audit which prompts you are already enrolled in and pause them on thin-margin or low-stock ASINs.

On auto-optimization in January: test it on a contained set of campaigns, not the account. Check the default state on day one, review every automated action against spend and conversions, and measure it on TACoS and contribution margin rather than on speed to first sale.

Amazon's automation is genuinely good at finding demand you would have missed. It is structurally incapable of knowing what a sale is worth to you. That gap is the job.

Related reading

• Amazon Rufus Ads 2026: What Sponsored Product Prompts Mean for PPC — the placement you are already enrolled in
• Amazon TACoS Explained: How to Measure True Profitability — the metric to judge automation on
• Amazon PPC Bidding Strategy: How to Maximise Returns
• Amazon PPC Optimization Checklist (Weekly & Monthly)

Deciding what to automate and what to keep manual across your Amazon account? I manage PPC for sellers in US, UK and UAE markets, with a focus on TACoS and contribution margin rather than vanity ROAS. Get in touch.