7 Claude Prompts Every Amazon Operator Needs Right Now

The difference between a useful Claude output and a hallucinated waste of time is almost always prompt quality - not the model's capability. Claude is reasoning over whatever context you give it. Give it vague instructions and generic data, and you get plausible-sounding but generic output. Give it structured context, a defined role, clear constraints, and an explicit output format, and you get analysis that a competent specialist would recognise as genuinely useful.

This is Part 2 of the Claude x Amazon series. In Part 1, we covered the five use cases where Claude is generating real results for Amazon operators right now. Here, we go deeper on the prompts themselves - seven field-tested structures covering listing audits, SQP analysis, competitor gap mapping, wasted spend, review intelligence, A+ content briefs, and bundle strategy.

Each prompt includes what data to provide, what to expect from the output, and one practical note on where it tends to fall short without the right context.

One Principle Before You Start

Every prompt below follows the same underlying structure: role definition, structured context, explicit task, specified output format. Claude performs significantly better when you define what it's acting as (“You are an Amazon SEO specialist”), what data it's working with (pasted inline, structured), what you want it to produce (a ranked list, a table, a rewritten title), and what constraints apply (character limits, marketplace, category).

The prompts below work with Claude Sonnet or higher - in the Claude web interface, the API, or any tool that exposes Claude. For large data sets - SQP reports with 500+ rows, search term reports with thousands of lines - work in batches or use the API with extended context. The analysis approach remains the same regardless.

Prompt 1: The Full Listing Audit

Use this when you want a comprehensive evaluation of an existing listing - what's missing, what's underperforming, and what to fix first.

What to expect: A structured gap analysis with ranked keyword additions, a use-case coverage map, and ready-to-use rewrites for the highest-impact elements. The quality scales directly with the richness of competitor listings you provide - the more complete, the better the gap identification.

Pro tip: Add your SQP data alongside this prompt (condensed to the top 30 queries by volume) and ask Claude to weight the keyword gaps by search volume. This transforms the gap list from a content exercise into a traffic-prioritised action queue.

Prompt 2: SQP Opportunity Map

Use this to extract the three highest-value insights from your Search Query Performance data: CTR gaps (high impressions, low click capture), conversion gaps (clicks without purchases), and PPC harvest candidates (high-converting organic terms not yet in campaigns).

What to expect: Three distinct, prioritised action lists - one for CTR improvements (usually listing and main image changes), one for conversion improvements (usually content and pricing issues), and one for PPC expansion. This is the fastest way to extract structured strategy from a raw SQP export.

Pro tip: The full SQP walkthrough - including how to export, clean, and structure the data for this prompt - is covered in Part 3 of this series.

Prompt 3: Competitor Keyword Gap

Use this to map which keywords your top competitors are covering that your listing doesn't address - ranked by commercial intent rather than raw frequency.

What to expect: A ranked list of keyword gaps with specific placement recommendations. The commercial intent ranking is Claude's inference - it's useful for prioritisation, but validate the top items against your SQP data before investing significant rewrite time.

Pro tip: Run this quarterly, not just at launch. Competitors update their listings continuously. A gap that didn't exist six months ago may now represent a meaningful traffic opportunity.

Prompt 4: Wasted Spend Audit

Use this to identify the search terms in your Sponsored Products campaigns that are absorbing budget without generating conversions - and flag which branded terms should be harvested to exact match before competitors capture them.

What to expect: A negative keyword list ready to upload, a branded harvest list, and - most valuably - any systemic campaign structure issues generating waste across multiple terms simultaneously. The structural diagnosis is often more impactful than the individual keyword flags.

Pro tip: Set your ACOS threshold at 1.5x your target ACOS, not your break-even ACOS. This catches the clear waste while leaving room for terms that are expensive but building ranking.

Prompt 5: Review Intelligence Brief

Use this to synthesise the buyer language in your reviews into a structured brief - one that translates directly into listing copy, A+ content themes, and positioning decisions.

What to expect: A brand intelligence brief that usually reveals two or three things your listing is completely silent about - despite them being the primary reasons buyers chose your product over alternatives. The buyer language section is often the most immediately useful: verbatim phrases that belong in bullet points and A+ copy.

Pro tip: Run this on your top three competitor ASINs too. The gaps between what their buyers praise and what your buyers praise reveal positioning angles you're winning on that your current listing probably doesn't communicate clearly enough.

Prompt 6: A+ Content Brief

Use this to generate a module-by-module creative brief for Amazon A+ Content - structured well enough that a designer and copywriter can work from it directly without a lengthy briefing session.

What to expect: A module structure that a design team can work from in a first briefing session. The emotional hook section is often where Claude adds the most value - it identifies the non-functional purchase drivers that listing copy tends to underplay because they don't map to obvious keywords.

Prompt 7: Bundle and Variation Strategy

Use this to identify untapped variation and bundle opportunities based on actual buyer behaviour patterns in your reviews and search data - without relying on gut instinct or category guesswork.

What to expect: Specific expansion recommendations grounded in actual buyer language and search demand rather than category intuition. The quality of this output scales directly with review volume - the more reviews you include, the more grounded the variation suggestions.

What These Prompts Won't Do

These prompts are genuinely useful. They also have a structural ceiling that no amount of prompt refinement overcomes.

They work on data you export and paste manually. Which means every analysis is a snapshot - accurate when you ran it, increasingly stale as the days pass. Claude has no memory of your previous sessions: next month's SQP analysis starts completely fresh, with no visibility of whether last month's recommendations actually moved the metrics they were targeting. There's no connection between the recommendations Claude produces and the actions taken in Seller Central - no tracking of what was implemented, no measurement of what worked.

For one brand, once a month, this workflow returns real value. For a portfolio of brands, or a catalogue where weekly analysis would be the right cadence, the overhead of running the manual loop consumes a significant fraction of the value it creates.

That's the structural ceiling we examine in Part 4 of this series - and what changes when the data layer is continuous rather than export-dependent.

Frequently Asked Questions

Which Claude model works best for these Amazon prompts?

Claude Sonnet 4 or higher for most analytical tasks - SQP analysis, competitor gap mapping, wasted spend audits. It handles large data sets well and produces structured output consistently. Claude Opus produces marginally richer reasoning on complex prompts but the improvement is rarely worth the speed difference for routine analysis. The bigger variable is prompt structure rather than model version - a well-structured prompt on Sonnet outperforms a vague prompt on Opus.

How many rows of SQP data can I paste into Claude at once?

Claude Sonnet handles roughly 300-400 rows of SQP data comfortably in a single prompt using the standard interface. For larger data sets, reduce to the key columns (query, volume, impression share, click share, purchase share) before pasting - this roughly halves the token count. For very large catalogues with 1,000+ query rows, work in batches: top 300 by volume, then the next 300, and so on.

Can I use these prompts with ChatGPT instead of Claude?

Yes - these prompts work with any major LLM including ChatGPT-4o and above. Claude tends to produce slightly more specific and actionable output on analytical tasks (SQP analysis, wasted spend audit, competitor gap mapping) while ChatGPT is roughly comparable on copywriting tasks. Prompt quality remains the most important variable regardless of which model you use.

How often should I run these prompts?

SQP analysis: monthly, aligned with when Amazon makes the data available. Wasted spend audit: bi-weekly for active campaigns with significant spend. Listing audit and competitor gap: quarterly, or when you notice a ranking or CTR decline on specific queries. Review intelligence brief: when you accumulate 20+ new reviews, or quarterly for established products.

What if Claude gives keyword suggestions that aren't relevant to my product?

This usually means the competitive context isn't specific enough. If Claude's inferring what your competitors sell from vague titles rather than reading their full bullets and descriptions, it will generate plausible but misaligned suggestions. Fix: paste the full listing for each competitor (title + all bullets + description), not just the title. Also explicitly state your product category and target customer in the prompt.