482 products with thin copy
For Shopify merchants · Private alpha
Make your catalog legible to AI.
AI shopping agents are already picking products for your customers. If your catalog isn't legible to them, you're invisible — regardless of how good your products are. Legible scores yours on the six signals that decide whether you get picked.
trylegible.com/store/acme/readiness
62
/ 100
material, gender, audience missing
702 images unlabeled
product_type consistent
18 near‑dupes to review
fr-FR partial, de-DE missing
Why now
Shoppers aren't browsing anymore.
They're asking agents.
ChatGPT, Perplexity, Rufus, Gemini, Claude — each one is turning into a storefront. When a buyer asks "find me a linen shirt, breathable, under $80, ships to Paris," the agent reads your product data and decides in milliseconds whether to surface you. Legible tells you, line by line, where that decision goes wrong.
6pillars
The signals agents actually read before recommending
~1second
Average time an LLM spends evaluating a candidate product
0writes
We read. You fix. You stay in control of every edit.
The six pillars
What an AI agent sees
when it reads your catalog.
Critical · fixes first
Descriptions that actually describe.
We score every product description with Claude Haiku against a rubric that mirrors what LLMs look for: concrete attributes, use cases, fit, material, what problem it solves. Fluff gets a low score. Copy that sells to a human but fails an agent gets flagged.
- Per-product quality score 0–100
- Distribution chart across your catalog
- Ranked list of the 50 that'll move your score most
High impact
Metafields an agent can actually query.
Agents don't read prose to find sizing. They query structured fields. Legible audits your metafield coverage across the namespaces AI shopping systems actually pull from — material, audience, gender, origin, care — and shows you exactly which products are invisible on each.
- Coverage matrix by namespace
- Missing-field heatmap by collection
- Priority list weighted by traffic
Silent killer
Near-duplicates, grouped and ranked.
We embed every product with Voyage AI's voyage-3-lite and cluster variants, re-releases, and accidental duplicates. Agents punish catalogs that look cluttered — you'll know which SKUs to merge, which to retire, and which to just re-name.
- Confidence score per cluster
- Diff view of the overlap
- One-click export to CSV for your ops team
Accessibility + agents
Alt text that does double duty.
Alt text matters for screen readers. It also matters for multimodal agents that caption your images before ranking you. We find every unlabeled image, score weak ones, and suggest rewrites that work for both audiences.
- Per-image coverage + quality
- AI-drafted suggestions
- WCAG-aligned
Structure
Taxonomy that doesn't contradict itself.
"Tee", "T-shirt", "Tshirt", "Graphic Tee" — four product types, one actual thing. Legible finds taxonomy drift across your catalog and proposes a canonical mapping an agent can trust.
- Product-type collisions
- Collection consistency score
- Suggested canonical labels
Global reach
Multilingual, audited honestly.
If you ship to France, an agent querying in French will read your fr-FR copy — or fall back to your English, or skip you. Legible checks the translations you already have, flags placeholder strings and missing locales, and shows you where each market is exposed.
- Locale coverage per pillar
- Detection of machine-translated-and-forgotten copy
- Per-market readiness score
How it works
Score in five minutes
— fix on your own schedule.
-
Install from Shopify
One click. Read-only scopes. No write permissions, ever, in v1. Your catalog stays yours.
-
We read your catalog
Products, variants, metafields, translations, images. First pass runs in minutes for most stores, background sync after that.
-
Fix in priority order
The opportunity list is ranked by how much each fix moves your score. No spreadsheets, no guessing — just the next best thing to do.
I built Legible after watching my own store disappear from AI shopping results while competitors with worse products kept showing up. The difference wasn't quality — it was data legibility. Six specific signals, the same six every agent reads. This app scores them, in order of what matters.
Hamza Ziouine
Founder · Legible
Pricing
Flat monthly
— 7-day free trial, no credit card.
One optimized product description from a freelancer costs around $100. Legible tells you which 50 products to fix first — for $39 / month.
Starter
$39/ month
Up to 500 SKUs
- All six pillars
- Weekly re-sync
- Per-product opportunity list
- CSV export
Growth
$59/ month
Up to 5,000 SKUs
- Everything in Starter
- Daily re-sync
- Multilingual audit across locales
- Duplicate clustering with confidence
- Priority email support
Scale
$149/ month
Up to 50,000 SKUs
- Everything in Growth
- Hourly re-sync
- API access (read)
- Dedicated onboarding call
Over 50K SKUs? Get an enterprise quote →
FAQ
Answers
— before you ask.
Does Legible change anything in my store?
No. V1 is strictly read-only. We request the minimum scopes to read products, variants, metafields, translations and images. Write-back is a V2 feature and will be opt-in per action.
How long does the first scan take?
Under 500 SKUs: about 2 minutes. Up to 5,000 SKUs: under 15 minutes. Above that, the first scan runs in the background and the dashboard fills in progressively — you don't have to wait for it.
Do I need a developer to use it?
No. Install is one click from the Shopify App Store. The dashboard, scores, and opportunity list are designed for merchants — no CLI, no SQL, no GraphQL. A dev helps if you want to bulk-apply fixes, but none of V1 requires one.
What happens when the trial ends?
You pick a plan to keep scoring your catalog, or the app stops syncing and your last report stays visible for 30 days. Nothing auto-charges — Shopify always confirms the subscription before billing starts. Cancel from the Apps screen in your admin.
How is the score calculated?
Your score is a weighted average of six pillars. Weights reflect what actually changes an agent's decision — descriptions and metafields carry more weight than alt text or taxonomy. The breakdown is visible per product on every report.
What data leaves my store?
Product text and image URLs are sent to Claude Haiku (Anthropic) for scoring and Voyage AI for embeddings. No customer or order data, ever. Processors are disclosed in our privacy policy and can be switched off in enterprise plans.
What if I have more than 50,000 SKUs?
Email hello@trylegible.com — we'll size a plan for you. Enterprise includes a dedicated sync schedule and SSO on request.
Does it work for non-apparel catalogs?
Yes. The pillars are catalog-agnostic — we've built the rubric against home goods, electronics, beauty and food as well as apparel. Weights auto-tune to your product type over the first week.
Agents are already shopping your catalog.
Find out what they see.
Already a Shopify partner? Book a 15-min demo