This Seller Cut Keyword Research From 6 Hours to 45 Minutes: The Exact AI Workflow They Built

A mid-size Amazon seller running roughly 40 SKUs used to spend close to six hours every product launch cycle on keyword research. After rebuilding the process around Claude, the same research now takes about 45 minutes. Here is the exact six-step workflow, not a general description of what AI can theoretically do.

Why the Old Process Took Six Hours

The original workflow involved manually copying competitor titles and bullets into a spreadsheet, then eyeballing repeated words while switching between browser tabs and search term reports, losing track of which terms had been evaluated.

Step 1: Pull 15 to 20 Competitor Listings in One Pass

Open the top 15 to 20 organic and sponsored results for your target term, copy title, bullets, and backend terms into a single document labeled by ASIN. Fifteen is the practical minimum, twenty a reasonable ceiling.

Step 2: Extract Candidate Keywords With a Specific Prompt

Prompt: "Here are the titles and bullet points from the top 15 to 20 listings for [target search term]: [paste]. Extract every distinct keyword phrase that appears in at least 3 of these listings. For each phrase, note how many listings it appears in and whether it shows up in the title, bullets, or both. Do not include generic words like the product category name alone."

The 3-listing threshold filters out one-off phrasing from a single competitor.

Step 3: Cross-Reference Against Your Actual Search Term Report

Prompt: "Here is my search term report for the last 60 days: [paste]. Compare it against the candidate keyword list you just built. Flag any candidate keyword not currently appearing anywhere in my report, and flag any term with meaningful impressions but zero or near-zero clicks."

Step 4: Prioritize Into Three Tiers

Prompt: "Sort the full keyword list into three tiers: Tier 1, appears in 8 or more competitor listings and not currently in my search term report. Tier 2, appears in 4 to 7 listings with low click volume. Tier 3, everything else. Present as three separate lists."

Step 5: Draft Placement, Not Just a List

Prompt: "For each Tier 1 keyword, suggest whether it fits better in the title, a specific bullet, or backend search terms only, based on how it appeared across the original competitor listings."

Step 6: Time-Box the Whole Process

Run steps one through five inside a 45-minute window. Twenty listings and one search term report pass is enough signal.

A Worked Example From Step 2 Through Step 4

Researching "stainless steel french press" turned up 18 competitor listings. Step two surfaced candidate phrases including "double wall insulated," "borosilicate glass," "4 cup capacity," and "cold brew compatible," each appearing in at least 5 of the 18 listings. Cross-referencing in step three showed "cold brew compatible" was not appearing anywhere in the seller current terms, a clear Tier 1 gap, while "double wall insulated" had 340 impressions but only 2 clicks, flagging a possible relevance mismatch.

Step four sorted "cold brew compatible" into Tier 1 with a bullet point placement recommendation, since it appeared in bullets across 7 of the 8 listings that used it, never in a title.

What Changed in Practice

The time reduction came from eliminating manual pattern-spotting, not from removing judgment about final implementation. Explore our Keyword Density Checker to verify your finished content reflects the priority keywords at healthy density.

What This Workflow Cannot Verify

None of these steps confirm actual current search volume or ranking behavior. Treat the output as evidence-based, not a guarantee.

Check our newsletter for ongoing AI research workflow coverage.

Frequently Asked Questions

How many competitor listings do I need to gather for step one to work well?

Fifteen is the practical minimum, twenty a reasonable ceiling.

Why require a keyword to appear in at least 3 listings in step two?

This filters out phrasing idiosyncratic to one competitor rather than genuinely common across the category.

Do I need Brand Registry or Amazon Brand Analytics to run this workflow?

No, the core workflow uses your own search term report and public competitor listing content.

How often should this six-step process be repeated?

Before every new product launch, and every 60 to 90 days for existing high-priority ASINs.

Takeaways

• This seller cut keyword research from roughly six hours to 45 minutes using a six-step process.
• Gathering 15 to 20 competitor listings with a 3-listing minimum filters noise from genuine patterns.
• Cross-referencing against your own 60-day search term report connects research to real performance.
• Sorting into three tiers turns a keyword list into a prioritized action plan.
• Time savings came from eliminating manual pattern-spotting, not removing seller judgment.