Amazon announced roughly ten new AI-powered advertising capabilities in July 2026. Two are worth understanding properly because they change the shape of how a campaign is managed rather than just adding a feature: Brand+ and Performance+.
Both are, at bottom, the same offer that every major ad platform has made over the last decade. Hand over more of the decision-making, get access to signals you cannot otherwise reach. It is a real offer with a real cost, and it is worth being clear-eyed about both halves.
What they are
Performance+ optimizes toward conversion outcomes using Amazon's first-party shopping signals, with substantially less manual targeting and bidding input from you. You set the objective and the constraints. The system decides who to reach and what to pay.
Brand+ points the same capability at upper-funnel objectives, reaching shoppers earlier in consideration rather than optimizing purely for the immediate conversion.
The underlying pitch for both is that Amazon can see purchase intent signals that are not exposed through keyword targeting, and that a model with access to those signals will allocate your budget better than you will.
That pitch is not wrong. Amazon does have signals you cannot see. The question is what you are trading for access to them.
The actual trade
What you gain is signal access and a meaningful reduction in management time. Manual keyword and bid management at scale is genuinely expensive in hours, and some of that work genuinely belongs to automation, and for many accounts an automated system will beat a busy operator who is managing forty campaigns and can only really pay attention to six of them.
What you give up is legibility. When performance moves, you have less ability to say why. You cannot inspect the bid decisions. You cannot audit the targeting the way you can audit a search term report. If the system decides your budget is better spent on a segment you would not have chosen, you often find out from the outcome rather than from the settings.
The left column is the important one, and it is easy to forget it exists. Handing over the inside of the campaign does not mean handing over the outside of it.
The incrementality problem
Here is the thing to watch for when you evaluate these tools, and it applies to every automated optimization product on every platform.
Systems optimizing toward attributed conversions are very good at finding shoppers who were going to convert anyway. Retargeting a customer who already has the item in their cart produces a fantastic attributed ROAS and close to zero incremental revenue. The reported number looks excellent. The business impact is nil.
This is not an accusation about Amazon specifically. It is a structural property of optimizing toward attribution. It means that comparing Performance+ reported ROAS against your manual campaign reported ROAS is close to meaningless as an evaluation, because the two are not measuring the same thing.
If you can run a proper holdout, run one. If you cannot, at minimum look at total account sales and total account spend across the test window rather than at campaign-level attributed performance.
How to test without betting the account
Carve out, do not convert. Take a defined slice of products and run it through the new campaign types alongside your existing structure. Do not migrate a working account wholesale to find out.
Set the outer envelope first. Whatever the model does inside the campaign, the budget ceiling and the hours the campaign is allowed to spend in are still yours to set. This is the part advertisers give away without meaning to. An automated system with an unbounded schedule and a generous budget can discover an expensive lesson at 3am on a Sunday, and you will read about it on Monday.
Setting a hard ceiling and a defined run window before you start does not fight the optimization. It bounds the blast radius while you find out whether it works for you. The model still optimizes freely inside the envelope you drew.
Give it a real window. These systems have a learning period, and judging them in week one measures the learning period rather than the steady state. Four weeks minimum.
Decide your kill criteria in advance. Write down what result would make you stop, before you start. It is remarkably hard to shut off something that is producing a good-looking number, even when you suspect the number.
The honest summary
These tools are likely to be a real improvement for accounts that are under-managed, which is most accounts. An operator with forty campaigns and six hours a week is not going to out-optimize a model with access to Amazon's signals.
They are a more genuine trade-off for accounts that are actively and competently managed, where the operator's category knowledge and margin awareness are doing work that a conversion-optimizing model does not know about. A model does not know that one of your SKUs is a loss leader and another has a 60% margin unless the objective you set encodes that.
Either way, the thing to hold onto is the envelope. Budget ceilings and run windows are cheap to set, they do not interfere with the optimization, and they are the difference between an experiment and an exposure.
Frequently asked questions
What is Amazon Performance+? Performance+ is an Amazon advertising product that uses Amazon's machine learning and first-party shopping signals to optimize campaigns toward conversion outcomes, with less manual targeting and bidding input from the advertiser. It was announced as part of a set of AI-powered capabilities in July 2026.
What is Amazon Brand+? Brand+ is oriented toward upper-funnel objectives, using Amazon audience signals to reach shoppers earlier in their consideration process rather than optimizing purely for immediate conversion.
Should I move all my campaigns to Performance+? No. Test it in a carve-out alongside campaigns you control directly, and compare over a meaningful window. Automated optimization products tend to perform well on attributed metrics partly because they are good at capturing demand that would have converted anyway, so incrementality matters more than reported ROAS when you evaluate them.
What control do I keep with Amazon's AI campaign types? You keep the outer envelope: which products are in scope, your budget ceilings, and when campaigns are allowed to run. What you give up is granular visibility into targeting and bid decisions inside the campaign. That trade is the thing to evaluate.
Off Hours is dayparting and budget automation for Amazon Ads. Set your rules once in plain English, and they run every day across every account you manage. Start a free 14-day trial.