Case Studies

How an E-Commerce Brand Built an AI Content System That Scaled Product Copy

A mid-sized e-commerce brand had 2,000 SKUs and terrible product descriptions. Here's how they used AI to fix all of it — and what happened to their organic traffic.

MurphJanuary 18, 20256 min read

Product content is the unglamorous foundation of e-commerce SEO. Good product descriptions, accurate attributes, compelling category page content — all of this drives organic traffic, reduces return rates, and improves conversion.

Most mid-sized e-commerce brands have mediocre product content because writing it at scale is brutal. This brand had 2,000+ SKUs and descriptions that ranged from thin to non-existent.

Here's what they built.

The Starting Point

A home goods brand. Roughly 2,200 active SKUs. Mix of owned brands and wholesale products.

Content audit findings:

  • 40% of SKUs had descriptions under 50 words
  • 23% had no description at all (pulled directly from manufacturer feed)
  • Category pages had thin or no copy
  • Zero FAQ content on product pages
  • Image alt text was missing on ~60% of images

Organic traffic was underperforming significantly compared to domain authority. The reason was obvious in the audit: Google had almost nothing to index on most product pages.

Previous approach: one part-time copywriter writing 10-15 product descriptions per week. At that rate, improving the catalog would take four years.

The AI Content System We Built

Step 1: Content Brief Templates

For different product categories (furniture, lighting, textiles, kitchen goods), we built detailed prompts that defined:

  • The description structure (benefit-led opening, features, specs, care/use instructions)
  • Tone guidelines (warm but informative, specific to their brand voice)
  • Keyword integration approach (primary keyword, secondary terms, natural integration)
  • Length targets (200-350 words for standard products, 400-600 for hero products)
  • Mandatory elements (dimensions, materials, care instructions, compatibility where relevant)

These prompts were refined over two weeks until output quality was consistent and on-brand.

Step 2: The Generation Workflow

We built a Make scenario that:

  1. Pulled product data from their Shopify store (title, existing description, product type, tags, specifications)
  2. Fed that data into the Claude API with the appropriate category prompt
  3. Generated a draft description
  4. Pushed the draft into a review queue (a Notion database with the product info and draft)

The part-time copywriter's job shifted from writing from scratch to reviewing and editing AI-generated drafts. Her throughput went from 12 products per week to 85-100 per week.

Step 3: Category Page Content

Category pages in e-commerce are often purely navigation — a grid of products with no text. But Google uses category page content to understand what the page is about.

We built category page templates:

  • Opening section: 150-200 words introducing the category, buyer intent signals, and brand positioning
  • Buying guide section: "How to Choose" content specific to that product category
  • FAQ section: 4-6 common questions about the category with answers

AI drafted all of this from category briefs. Copywriter reviewed. Their 85 category pages got proper content in six weeks.

Step 4: Image Alt Text at Scale

Missing alt text was both an accessibility problem and an SEO problem. But manually writing alt text for 30,000+ product images wasn't feasible.

We used GPT-4 Vision to analyze product images and generate descriptive alt text automatically. Product name + visual description + key attributes. Pushed back to Shopify via API.

This was one of those wins that took a day to build and would have taken a year to do manually.

Results at 6 Months

Metric Before At 6 Months Change
Organic sessions/month 48,000 79,000 +65%
Pages with thin content (<50 words) 880 95 -89%
Average product description length 68 words 287 words +322%
Organic revenue (attributed) $31k/month $58k/month +87%
Content production rate 12 SKUs/week 90 SKUs/week +650%

The organic traffic results took time to accumulate — months 1-3 showed modest gains, months 4-6 accelerated as Google re-crawled and re-ranked the improved pages.

What Drove the Traffic Gains

Better product page content gave Google more signal about what each page was about — meaning it started ranking for long-tail searches it wasn't showing up for before.

Category page content improvements were faster to impact rankings — category pages with proper content started appearing for broader category searches.

The image alt text improvements contributed to Google Images traffic, which drove a small but measurable additional traffic stream.

No new link building. No changes to site structure or technical SEO. Pure content quality improvement.

The Ongoing System

Now the AI content system is part of their regular workflow:

  • New products get AI-drafted descriptions before they're published
  • The copywriter reviews all new drafts and maintains a library of approved brand voice examples
  • Category pages get quarterly refresh reviews
  • The AI is also used for ad copy, email marketing, and social content — same brand voice, same efficiency gains

Content that used to be a bottleneck became a competitive advantage.

The Takeaway

E-commerce brands sitting on large catalogs of thin or missing product content have an SEO opportunity they're actively leaving on the table.

AI doesn't replace the editorial judgment needed to keep brand voice consistent and catch errors. But it eliminates the blank-page problem and the volume constraint.

One brand voice guide. One set of category templates. One workflow. And 10x the content throughput.

That's the system.

Want this for your business?

Tell us what you're building. We'll map out exactly what to build and what it costs.

Start Your Project →

Frequently Asked

How does AI content production at scale affect e-commerce SEO?

Thin or missing product descriptions are one of the most common causes of underperforming e-commerce SEO. Google has almost nothing to index on product pages with sub-50-word descriptions. AI allows a brand to systematically bring every product page up to a quality threshold that makes it indexable — descriptions with sufficient length, relevant keywords, and customer-facing language. The organic traffic gains come from Google having substantive content to match to search queries.

How do you maintain brand voice consistency when using AI to write 2,000 product descriptions?

The key is a detailed brand voice document that Claude or another AI uses as a reference on every generation — not just a style guide, but actual examples of approved copy at different lengths, a list of vocabulary to use and avoid, and the specific emotional register the brand targets. With a strong voice document, AI output requires minimal editing for tone. Without it, every description needs significant revision.

What is a realistic timeline for using AI to rewrite product descriptions at scale?

With a well-designed system — a batch processing workflow using the Claude API, a quality review layer, and a publish pipeline — a team of one can review and publish 100-200 product descriptions per day. A catalog of 2,000 SKUs is a 2-3 week project rather than the years it would take with one part-time copywriter writing from scratch.

What content beyond product descriptions does AI improve for e-commerce?

Category page copy, image alt text (critical for both SEO and accessibility), FAQ sections on product pages, email sequences for abandoned cart and post-purchase flows, and collection-level content that explains the category context. Each of these has organic traffic and conversion implications — and all can be systematized with AI in the same batch process approach.

Jason Murphy

Written by

Murph

Jason Matthew Murphy. Twenty years building digital systems for businesses. Former CardinalCommerce (acquired by Visa). Now running VibeTokens — AI-built websites and content for small businesses.

The window is open.

It won't be forever.

Start Your Project →