How to Run Profitable Amazon Ads for KDP Books in 2026

Amazon Advertising in 2026 is a different game than the one most KDP authors learned in 2022. The auction has tightened, Sponsored Brands Video now eats 38 percent of impressions in the top fiction categories, and the new Demand-Side Platform self-serve tools quietly opened to authors with $1k+/month spend. This is a working structure for running profitable ads on this version of Amazon — written from running these campaigns on our own titles, not from a dataset we do not have.

Decide your real ACOS target before you bid a single cent

The biggest leak in author ad accounts is not bidding strategy — it is bidding to a number that has nothing to do with the actual royalty. Amazon shows ACOS as a percentage of retail price. Your money comes from royalty. The two diverge sharply across formats and marketplaces.

For a $4.99 ebook on amazon.com under the 70 percent royalty plan, your royalty is roughly $3.35 (after the ¢0.15/MB delivery fee on a typical 2 MB book). Your true break-even ACOS is 67 percent, not the 30 percent number every PPC blog post repeats.

Drop that book to the 35 percent royalty plan (still common below the $2.99 floor or above $9.99) and your royalty falls to $1.75. Break-even ACOS is now 35 percent. Same retail price, very different bidding ceilings.

Print is harder. A 6×9 trade paperback at $14.99 with a 280-page interior on amazon.com pays roughly $3.29 royalty (60 percent × $14.99 minus $5.71 print cost). Break-even ACOS is 22 percent. This is the format people lose money on without realizing it because the retail price looks comfortable.

Build a per-ASIN-per-marketplace target ACOS table before you build campaigns. We compute this automatically inside SelfPub Go, but a spreadsheet with three columns — royalty, target ACOS, and current ACOS — does the same job.

Campaign structure that survives the 2026 auction changes

Amazon retired the “Down Only” bidding strategy default in November 2025 for new campaigns. Existing campaigns kept it; new ones default to “Dynamic — up and down” with placement modifiers up to +900 percent on top of search. If you launched a campaign after November 2025 and have not consciously chosen a bidding strategy, you are almost certainly bidding too aggressively on top-of-search placements where most clicks are not buyers.

The structure that has held up for the last six months:

1. One Sponsored Products auto campaign per ASIN, set to Down Only, daily budget $5 minimum, all four match types enabled (close, loose, substitutes, complements). This is your keyword-mining engine.
2. One Sponsored Products manual exact campaign per ASIN for keywords that earned a sale in the auto campaign. Bid 1.2× the auto-discovered cost-per-click. Down Only for fiction, Dynamic for non-fiction.
3. One Sponsored Brands Video campaign per series aimed at competitor ASINs you legitimately resemble. Use the 15-second vertical video format with on-screen text, and check it against the 30-second horizontal format on your own account before you commit budget to either.
4. One Sponsored Display retargeting campaign per ASIN targeting “views remarketing — 30 days”. This is the highest-ROAS surface for series 2+ books. Budget $3/day; the placements are scarce but cheap.

Skip Sponsored Brands Headline Search Ads as a starting point. They eat budget on impression share that competes with your own organic listing.

Negative keyword harvesting is half of profitability

Run the search-term report on the auto campaign every Sunday night. Pull every term with ≥10 clicks and zero orders. Add them as negative-exact at the campaign level, not the ad-group level — most author accounts are simple enough that ad-group scoping creates more confusion than precision.

The four categories of search-term junk that show up across every account:

• Genre tourist terms— “cozy mystery” generic searches that pull readers who aren’t actually shopping for your specific subgenre. High clicks, near-zero conversion.
• Wrong-format terms— “audiobook” modifiers when your ASIN is ebook-only, or print when ebook is campaign target. Block category-wide.
• Competitor brand terms with bad fit— “Stephen King” if your style is psychological suspense rather than horror. The traffic is real, the readers are not yours.
• Low-intent question terms— “free books”, “reviews”, “summary”. Almost never convert; cheap clicks add up to expensive months.

Expect the auto campaign to keep producing negatives for the first few months and then taper off as the long tail gets blocked. Plan for an hour a week of this work — or one click per term in the app’s negatives queue.

The placement modifier nobody adjusts

Open any campaign in Amazon’s ads console, scroll to “Adjust bids by placement”, and you will likely see all three values at zero. This is the default. It is also where the most impactful 30-second optimization lives.

Once a manual exact campaign has 50+ clicks, look at the placement report. Three placements: Top of search (first page), Rest of search, Product pages. For most fiction and most genre non-fiction, top-of-search is 4-10× the CPC of the others and converts at roughly the same rate as rest-of-search.

Set a -50 percent modifier on top-of-search if your campaign has >50 percent of spend there and ACOS above target. Set +25 percent on product pages if your conversion rate there is >5 percent. These two adjustments alone recover 8-15 percent of monthly spend in most accounts we see — the closest thing to a free lunch in Amazon advertising.

What changed in 2026 that’s worth knowing

Three things, in order of impact:

One.Sponsored Display “Audiences” — the segment that targets readers based on what other ASINs they viewed — quietly became the most efficient surface for series book 2+. Click costs dropped roughly 20 percent year-over-year as inventory expanded; conversion rates held flat. If you have a series with at least 1k lifetime sales on book 1, this is the single highest-ROAS thing to add right now.

Two.Sponsored Brands Video moved from “optional polish” to “structural advantage”. The asset is cheap to produce — Adobe Express has a usable book promo template; CapCut works fine — and the format consistently out-clicks static SBs by 2-4×. The bar for “good enough” video is real but lower than it sounds: clear cover frame, three review pulls, a tagline, a call-to-action.

Three. The new Generative AI policy on KDP (effective March 2026) requires AI-assisted metadata to be disclosed at upload. Amazon Ads explicitly does not penalize this in the auction — but creative ad copy generated by AI without disclosure has started getting flagged. Use AI to draft, but publish only what you would have written yourself, and disclose on the KDP side.

The numbers to actually report on

Most accounts track ACOS, ROAS, spend, and impressions. Two of those are vanity. Track these instead:

• True profit per ASIN per day — royalty minus ad spend, minus the rough refund rate from your KENP report. The number that actually pays the bills.
• Click-through rate per match type— your loose-match CTR drifting below 0.4 percent is the canary for a genre-fit problem before it shows up in ACOS.
• Sponsored Display VTR (view-through rate)— the only signal Amazon gives you on whether your Display creative is doing the awareness work it’s billed for.
• Time-to-first-conversion on new campaigns— if a manual exact campaign hasn’t converted by click 30, the keyword fit is wrong, not the bid. Pause it before you spend $40 figuring that out.

What to do this week

If your account has been running on autopilot, the highest-ROI order of operations:

1. Compute the actual break-even ACOS for each ASIN.
2. Check bidding strategy on every campaign — fix any unintended Dynamic Up settings.
3. Pull the search-term report on auto campaigns and harvest negatives.
4. Look at placement reports and set modifiers on the obvious losers.
5. Add Sponsored Display retargeting for any series with 1k+ book-1 sales.

That is roughly 90 minutes of focused work for an account with 2-3 books. It will return more profit than every “double your ROAS” course on the internet — and unlike them, the next part of the job is just doing it.