How We Set Up AI-Powered Amazon Ads for a KDP Author in 2026

Actualización: 6 de abril de 2026

Why Automate Amazon Ads at All?

Nuestro cliente es un autor de no ficción con un libro en Amazon. Copia impresa a $29.95, Kindle a $16.99. El libro tuvo un buen lanzamiento en 2023, luego se quedó quieto. Las ventas orgánicas empezaron a recuperarse — cerca de $50/mes para principios de 2026, impulsadas principalmente por redes sociales y marketing de contenido manejado por el mismo agente de IA.

El contexto del mercado es relevante: Amazon representa más del 70% de todas las ventas de libros electrónicos en Estados Unidos (Statista, 2025), y los autores de KDP que utilizan Amazon Ads reportan un incremento promedio de ventas de 30-50% dentro de los primeros 90 días de activar campañas optimizadas (Written Word Media, 2026). La pregunta era: ¿puede un agente de IA manejar esa optimización sin que el autor aprenda PPC?

The question was: can we pour fuel on this with Amazon Ads, without the author having to learn PPC?

The answer is yes — but the setup process has more steps than you'd expect.

Por qué Amazon Ads con IA tiene sentido para autores de KDP

• Amazon representa más del 70% de todas las ventas de libros electrónicos en EE.UU. (Statista, 2025) — no hay alternativa de distribución equivalente
• Los autores de KDP que usan Amazon Ads reportan incrementos de ventas de 30-50% en los primeros 90 días con campañas optimizadas (Written Word Media, 2026)
• El ACOS promedio óptimo para libros de no ficción está entre 25-40% — por encima de eso, las campañas consumen más de lo que generan
• Los sistemas de monitoreo con IA pueden revisar métricas y proponer optimizaciones diariamente — vs. la revisión semanal/mensual típica de un autor manual
• $3.70 de retorno por dólar invertido es el promedio de empresas que implementan automatización con IA para sus operaciones de marketing (Graft Growth Partners, 2025)

El caso de uso de este artículo demuestra que la automatización de Amazon Ads no requiere un equipo de marketing. Requiere el setup correcto — que describimos paso a paso abajo.

Step 1: Confirm You Already Have an Amazon Ads Account

If you've published through KDP, you already have an Amazon Advertising account. Go to advertising.amazon.com and sign in with your KDP credentials. You'll land on a Campaign Manager dashboard — probably empty. That's fine. The account exists, and that's what matters.

We found two old campaigns from a previous experiment — both ended, neither had meaningful results. Starting fresh with a strategy this time.

Step 2: Create an Amazon Developer Account

To access the Ads API (which is how an AI agent talks to Amazon Ads), you need a developer account at developer.amazon.com.

This is where we hit our first wall. The registration form requires a US state and zip code. Our client is in Puerto Rico. Puerto Rico wasn't in the state dropdown. We used Florida with a Miami zip code as a workaround — Amazon accepted it.

Then came identity verification. Amazon asks you to upload a government ID, and your name has to match exactly. Not "close enough" — exactly. Our client's legal name on their ID was different from the name they use professionally. We had to use the full legal name from the ID. Verification took a few hours.

Step 3: Create a Login with Amazon (LwA) Security Profile

This is the OAuth layer that lets your application (the AI agent) authenticate with Amazon's API.

Go to developer.amazon.com/loginwithamazon/console. Create a new Security Profile. You'll get a Client ID and Client Secret — save both. These are the keys your agent uses to request access tokens.

Under Web Settings, add a redirect URI. We used https://www.amazon.com as a simple callback URL since we're doing a manual OAuth flow, not building a web app.

Step 4: Apply for Amazon Ads API Access

Go to advertising.amazon.com and find the API application page. Select "Direct Advertiser" — you're advertising your own products, not managing ads for others. Choose "Author" as the business type.

Amazon reviews the application. They say up to 72 hours, but ours was approved the same day. Once approved, go back to the LwA console and assign the advertising::campaign_management scope to your security profile.

Step 5: Complete the OAuth Flow

This is the most technical part. You need to get a refresh token that your agent stores permanently.

The flow works like this: you visit a specific Amazon URL with your Client ID, the user (you) clicks "Allow," Amazon redirects back with a temporary authorization code, and you exchange that code for an access token and a refresh token.

The refresh token is the golden ticket. It doesn't expire (unless revoked) and lets your agent get fresh access tokens whenever it needs them.

One gotcha: the authorization code expires in 5 minutes. Have your token exchange ready to go before you click "Allow."

Step 6: Get Your Advertising Profile ID

Once you have a working access token, your agent calls the profiles endpoint to get the advertising profile IDs associated with your account. Each marketplace (US, Canada, etc.) has its own profile ID. The US profile ID goes into the Amazon-Advertising-API-Scope header on every subsequent API call.

Step 7: Set Up Budget Controls

Before the agent creates any campaigns, we set up guardrails:

Portfolio with monthly cap: We created a portfolio with a $200/month budget cap. Once total spend across all campaigns in the portfolio hits $200, Amazon pauses everything automatically. This is the hard ceiling.

Daily campaign budgets: Each campaign also has its own daily budget.

Agent rules: The AI agent has a three-tier permission system:

• Tier 1 (autonomous): Read reports, generate analysis, post metrics. No approval needed.
• Tier 2 (propose and wait): Bid changes, keyword pauses, ad group adjustments. Agent proposes with data, waits for human OK.
• Tier 3 (never autonomous): New campaigns, budget increases, campaign deletion. Human only.

Step 8: Reporting Schedule

The agent runs a reporting cron on Monday and Thursday mornings. It pulls performance data from the API — impressions, clicks, spend, ACoS, ROAS — and posts a formatted report to a dedicated Discord channel.

If ACoS exceeds 50% on any keyword for 7+ days, the agent automatically proposes pausing it. If ROAS exceeds 3x, it flags the opportunity to scale. Both require human approval before action.

What We Learned

The API bureaucracy is the bottleneck, not the integration. Three sequential gates — developer account, identity verification, API application — each taking hours to days. The actual API calls are straightforward once you have credentials.

Budget controls need to be structural, not just rules. Telling an AI agent "don't spend too much" is not a strategy. The portfolio cap is a platform-level kill switch that works regardless of agent behavior.

The three-tier autonomy model is essential. An agent that can freely create campaigns and adjust budgets is terrifying. An agent that can only read data and propose changes is useful. Know which one you're building.

Start conservative. $5-10/day per campaign, $200/month portfolio cap, manual keyword targeting. You can always scale up once you see what works.

The Bottom Line

An AI agent managing Amazon Ads doesn't mean "set it and forget it." It means the agent does the tedious work — pulling reports, analyzing keyword performance, catching underperformers — while the business owner makes the strategic decisions.

For an indie author spending 10+ hours a month manually checking Amazon Ads dashboards, this is a real time savings. For one who never ran ads because the learning curve was too steep, this makes the entire channel accessible.

We built this system for our own client. We can build it for yours.

AutoPilotPR builds AI agent teams for businesses in Puerto Rico and beyond. Want to see what an AI-powered Amazon Ads setup looks like for your business? Book a free discovery call.