Amazon Ads AI tools, including generated creative, dynamic bidding, automatic targeting and audience modelling, can cut hours of manual work from an account. The common mistakes come from trusting them without boundaries: setting goals that ignore margin, letting automatic campaigns run without negative keywords, publishing generated images without checking the product, and reading Amazon’s attributed numbers as the full story. Each mistake has a simple check that prevents it.
Mistake 1: Optimising for sales instead of profit
Automated bidding pursues the target you set. If the target is a broad account-wide ACoS, the system may spend heavily on low-margin products that hit the number while profitable products starve. A seller with a 15 percent margin product and a 45 percent margin product should not use one target for both.
How to avoid it: group products by margin and set a target ACoS or ROAS per group. Calculate break-even ACoS as margin before ad spend, then set the target below it for products you want to be profitable on ads alone. Products meant to drive launch velocity can carry a higher target, deliberately and for a set period.
Mistake 2: Running automatic targeting without maintenance
Automatic targeting campaigns find search terms and products Amazon thinks are relevant. They are useful for discovery but tend to accumulate waste: irrelevant terms, competitor brand terms that convert poorly, or very broad searches.
How to avoid it: review search term reports weekly. Add irrelevant terms as negative exact or negative phrase. Move proven converting terms into manual exact match campaigns with their own bids, and add them as negatives in the automatic campaign so the two do not compete.
Mistake 3: Using dynamic bidding “up and down” everywhere
The “up and down” option allows Amazon to raise bids by up to 100 percent for top-of-search placements and up to 50 percent for other placements when it predicts a conversion. On top of placement adjustments, this can push costs well above what the seller expected.
How to avoid it: start with “down only” or fixed bids on new campaigns, and use “up and down” selectively on proven campaigns with healthy margins. Review the placement report to see where spend actually goes.
Mistake 4: Publishing generated creative without product checks
Amazon’s generative image and video tools place a product in lifestyle scenes. They can get scale, setting or context wrong: a compact blender shown on a commercial counter, or a pet product in a scene with the wrong animal size. Shoppers notice, and misleading imagery can lead to returns.
How to avoid it: require a reviewer who knows the product to approve every generated asset. Use a short checklist: correct scale, realistic use, no implied features, no unapproved claims in text overlays.
Mistake 5: Ignoring the listing
AI tools optimise ads, but conversion happens on the product detail page. A listing with weak images, missing A+ content, few reviews or unclear bullets will waste ad spend however clever the bidding.
How to avoid it: before scaling spend, check the basics. Main image meets Amazon’s requirements, secondary images show use and scale, bullets cover the main buying questions, and the price is competitive for the category.
Mistake 6: Treating attributed sales as incremental
Amazon reports sales within its attribution window after an ad click. Some of those shoppers would have bought anyway, especially on branded searches. Counting all attributed sales as new revenue overstates the return on ads.
How to avoid it: track TACoS (total ad spend divided by total sales, including organic) over time. Use Amazon Marketing Cloud, where available, to look at new-to-brand customers and paths to purchase. Run a short holdout test on branded campaigns to see how much organic sales pick up when ads pause.
Mistake 7: Changing too many things at once
When a new AI feature launches, teams often switch it on across the account at the same time as changing budgets and creative. When results move, nobody can say why.
How to avoid it: introduce one change at a time in a test campaign, keep a change log and give each test a defined period, usually two to four weeks.
How Alien Road avoids these mistakes in Amazon accounts
In a typical engagement the first step is an account audit. The team reviews campaign structure, search term waste, placement spend and listing quality, and maps margin by ASIN with the client’s finance contact. The audit shows which of the mistakes above are already present and how much spend they affect.
Fixes follow in a set order. Margin-based targets come first, because every automated tool depends on them. Negative keyword lists and the separation of automatic and manual campaigns come next. Dynamic bidding settings are reviewed per campaign. Generated creative is introduced only with the two-step product and claims review.
Reporting runs weekly for operations (search term waste, ACoS by margin group, placement spend) and monthly for business results (TACoS, organic rank on priority keywords, new-to-brand share). Each monthly report includes the change log so the client can see what changed and when.
How do you check an account for these mistakes quickly?
A 30-minute self-audit covers most of them:
- Does each product group have its own target based on margin?
- When were negative keywords last added to automatic campaigns?
- Which campaigns use “up and down” bidding, and what share of spend goes to top of search?
- Has every generated image been reviewed by someone who knows the product?
- Is TACoS rising, falling or flat over the last three months?
- Is there a change log for the last 90 days?
Any “no” or “don’t know” is a place to start.
It also helps to ask who made each of the last ten changes in the account and why. If the answer is “the algorithm” for most of them, the account is running on defaults, and the seller has handed over decisions without setting the boundaries that make automation safe.
Where to start this week
Download the last 60 days of search term reports for your automatic campaigns and sort by spend with zero orders. Add the clearly irrelevant terms as negatives today. Then calculate break-even ACoS for your top five products and compare it with the targets currently set. Those two actions usually recover more budget than any new feature adds.
Keep reading: Amazon Ads · Amazon