Stop Guessing at eBay Titles: The Exact Step-by-Step Claude Workflow for Keyword Research That Actually Ranks

Most guides to using AI for eBay keyword research stop at "paste in some competitor listings and ask Claude what patterns it sees." That is not a workflow, it is a single step dressed up as a whole process. Here is the full six-step process, with the exact prompt to use at each stage.

Why This Task Fits Claude Specifically

eBay Cassini search weighs listing titles heavily, and good title research means looking at a genuinely large set of top-performing listings in your exact category. Claude long context window can hold twenty, thirty, even fifty pasted competitor titles and item specifics in a single conversation without losing detail, and Projects keeps that research context persistent across sessions.

Step 1: Set Up a Dedicated Project for the Category

Create a Claude Project for the specific category you are researching, named specifically, "Vintage Camera Lenses Title Research" rather than "eBay Research." Add context in the description: typical item condition, general price range, and category-specific terminology.

Step 2: Gather 20 to 30 Real, Currently Active Listings

Search your exact category on eBay, sort by best match or watchers/sold count, and pull the title and item specifics from the top 20 to 30 listings. Twenty is the practical floor, thirty a reasonable ceiling. Copy into a plain text document, one listing per block.

Step 3: Extract the Title Pattern, Not a Single Title

Paste your listings and use this prompt: "Here are the titles and item specifics of the top 20 to 30 active and recently sold listings in [your category]: [paste]. Identify the specific words and their typical order that appear most consistently across the highest performers, brand placement, condition terminology, specification ordering, size or measurement format. Do not give me a single example title. Give me a reusable title structure, in the format of a template with bracketed placeholders, that I can apply across my inventory."

The critical instruction is the last line. A structure generalizes across your inventory. A single title does not.

Step 4: Stress-Test the Structure Against Your Actual Item

Run this prompt: "Here is my draft title for a specific item using the structure you proposed: [paste your draft title]. Compare it against the original 20 to 30 listings I gave you. Flag any specificity that is present in the top-performing listings but missing from my draft, and flag anything in my draft that does not fit the pattern you identified."

Step 5: Audit Item Specifics With the Same Rigor

Paste your current item specifics alongside the researched listings item specifics and use this prompt: "Here are my current item specifics fields: [paste]. Here are the item specifics fields used by the top-performing listings in this category: [paste]. Flag any field I have left blank or filled in less specifically than what the top performers use, and note which fields appear most consistently across all of them."

Step 6: Test on a Small Batch Before Rolling Out

Apply the new structure to five to ten listings first with existing traffic or sales history. Track impressions, click-through, and sales over the following two weeks against that same batch own prior baseline. If it improves, roll out to the rest of the category. If not, return to step three with a refreshed set of competitor listings.

A Worked Example, Start to Finish

To make the six steps concrete, here is what this looks like applied to an illustrative example, a seller listing a used mechanical keyboard.

Draft title before running the workflow: "Mechanical Keyboard Blue Switches Great Condition." Vague on brand, no measurement or layout information, condition buried at the end.

After step two, the seller has pasted in 25 active and recently sold listings for similar keyboards. Step three identifies a consistent pattern: Brand, then Model, then Switch Type, then Layout (Full Size, TKL, 60%), then Condition, then any notable accessory. Proposed structure: [Brand] [Model] [Switch Type] Mechanical Keyboard [Layout] [Condition], [Included Accessory if any].

Running step four against the draft title, Claude flags two gaps: no brand or model at all, and no layout information, both present in nearly every top-performing listing. Revised title: "Keychron K8 Blue Switches Mechanical Keyboard TKL Layout, Excellent Condition, RGB Backlit."

Step six means listing five to ten keyboards with the new structure, watching impressions and click-through over two weeks against those listings own prior baseline, and rolling out further only once the data confirms it helped.

What This Workflow Cannot Do

Claude has no live connection to eBay actual current search rankings or Cassini own weighting logic. Every recommendation is a pattern inference, not a verified ranking factor. This is exactly why step six, small-batch validation, is not optional.

Turning This Into a Repeatable Habit

Revisit each category Project every few months, rerun step two with a fresh pull, and compare against what you found previously. Verify Claude context window and Projects details against Anthropic own documentation. Marketplace Pulse tracks broader cross-platform search trends.

Pair this with our eBay product research prompts, and check our newsletter for ongoing marketplace coverage.

Frequently Asked Questions

How many competitor listings do I actually need to gather before starting?

Twenty is the practical floor. Below that, the pattern signal gets noisy. Thirty is a reasonable ceiling.

Do I need to redo all six steps for every single new listing?

No. Steps one through three build a reusable structure for an entire category. New items just need step four.

What if my category has very few active listings to research?

For fewer than fifteen or twenty comparable listings, skip pattern extraction and do a manual individual review instead.

How do I know if the new title structure actually worked after step six?

Compare the test batch against its own prior baseline over the two weeks before the change, not against untouched listings.

Can I skip Projects and just do this in a regular Claude conversation?

For a single one-off session, yes. Projects becomes valuable once you research multiple categories over time.

Takeaways

• This is a six-step workflow: Project setup, gather listings, extract structure, stress-test, audit item specifics, validate on a small batch.
• Ask for a reusable title structure with bracketed placeholders, not a single finished title.
• Item specifics deserve the same research rigor as titles.
• Never skip small-batch validation before full rollout.
• Revisit each category Project every few months as patterns shift.