Understanding enrichment
Plans: Starter (products only) · Growth · Pro · Agency
For step-by-step instructions (one entity, bulk, Teach the AI) see Enrich your catalog. This page explains what enrichment produces and what it does — and doesn't — do.
Each time Clione enriches an entity, it makes one AI call for that entity and saves the result in Clione. The same model applies to all four entity types: products, BigCommerce categories, Shopify collections and content pages.
What enrichment produces
| Output | What it is |
|---|---|
| Semantic description | What the entity is, who it's for, when to choose it, and the hesitations a shopper may have. Products also get what they are not for, when the content supports it. |
| Quality scores | Scores from 1 to 10 plus a price positioning. Which scores an entity gets depends on its type — see Understanding quality scores. |
| Search keywords | 12–18 terms shoppers actually type, in the same language as your content. |
| Meta title and meta description | Written to the length limits search engines show. |
| Vector embedding | Used for semantic search inside Clione. |
From these Clione builds the structured data (JSON-LD) your storefront receives when you publish: a Product graph for products, a collection graph for categories and collections, a WebPage graph for pages — plus FAQPage when the entity has approved FAQs.
What goes into the AI call
- Your catalog data: title, description, price, brand or vendor, categories, options.
- Your store-level context from AI Training: store instructions, your brand documents, and the Defaults below.
- The answers you gave in the entity's Teach the AI questionnaire, if any.
- When Google Search Console is connected, the real search queries that already bring traffic to that page.
The model doesn't see your reviews, returns, conversion rate or sales data. So:
- The description and reasoning are sharp — the model is good at summarising a product's character from its text.
- The quality scores are educated guesses from your catalog text, not verified performance data. Teach the AI → Verifiable evidence (warranty, reviews, certifications) is how you ground them in facts.
Store defaults
AI Training → Defaults holds the fixed facts about your store that the AI should never guess. They apply to every enrichment on the store — products, categories, collections and pages — from the next run after you save.
| Field | What it does |
|---|---|
| Brand name | The canonical brand for the catalog. Wins over the vendor on each product, which on multi-line catalogs is often a product line rather than a brand. |
| Market | United States, European Union, United Kingdom or Latin America. Decides the units and currency in generated copy. Leave blank to follow your source data. |
| Title format | Preferred order of the meta title, written with placeholders: {Brand}, {ProductName}, {SKU}, {Feature}. At least one placeholder is required; anything else in braces is rejected. |
| Keyword blacklist | Search terms the AI must never propose for this store, one per line. Removed after generation, on top of Clione's universal list. |
| Category blacklist | Admin categories (Sale, Clearance, staging buckets) the AI should not treat as shopper-facing. |
| Length preferences | Optional min/max characters for the meta title and meta description. Each max must stay inside the platform cap shown next to it. |
The two blacklists are filled only by your decision. Clione never pre-fills them: a term that looks generic from the outside may be exactly the audience a category was built to capture. Leave them blank until you've seen the enrichment propose a term you don't want traffic for.
Blank fields mean the default. Nothing already published changes until you re-enrich and publish again.
What happens on re-enrich
- A new version is saved; the previous output stays in the History tab.
- The 3 most recent versions are kept; older ones are removed automatically.
- You can Restore any kept version from History. The current version is saved first, so a restore can be undone.
- If the entity was edited by hand, the confirm dialog warns you that enriching replaces those manual edits.
Two counters, two questions
"Enrichment" appears in two places that answer different questions. They will diverge, and that's by design.
| Where | Question it answers |
|---|---|
| Enrichments counter in the store panel | How many enrichment credits have I used this month? Every run counts, including re-enriching the same entity. |
| Enrichment coverage on the store overview | What share of my entities has an enrichment saved? Re-enriching an entity that was already enriched doesn't raise it. |
Example. You re-enrich the same product 4 times in one month. Coverage goes from 0 to 1 product; the counter shows 4 credits used. To grow coverage, filter a listing to Never enriched and enrich those.
Credits
Each enrichment of one entity uses one credit; a bulk run uses one per entity it processes. The allowance is counted per calendar month in UTC and resets on the 1st, whatever day you signed up.
| Plan | Credits per month |
|---|---|
| Starter | 50 (products only) |
| Growth | 200 |
| Pro | 500 |
| Agency | 500 |
Credits belong to the organization on every plan: all its stores draw from the same allowance.
When you run out, single enrichments are refused and bulk runs enrich only what fits, skipping the rest. Everything already enriched and published keeps working. To lift the cap without changing plans, buy Expand credits from Billing & Plan → Add-ons — see Billing.
Enrichment never writes to your storefront
Enriching an entity does not:
- Write anything to your storefront. The meta title, meta description and structured data stay in Clione until you publish them.
- Change your catalog's price, inventory or theme.
- Notify your customers.
- Change your search ranking by itself — search engines only see a change after you publish it and they re-crawl the page.
Enriching does:
- Save the new output in Clione, where you review it on the entity page.
- Use one credit.
- Save a version.
There's no setting that publishes automatically after an enrichment. Inspecting first is the default: enrich, read the output on the entity page, and publish from the Publish tab when you're happy. See Publish.
Where enriched content ends up once published
| Where | BigCommerce | Shopify |
|---|---|---|
| Meta title, meta description | Meta description and meta keywords are written to the product, category or page. Meta title is always written to products, and to categories and pages only when Push meta_title to categories, collections, and pages is on (Store Settings → General). Search keywords are written to categories and pages, never to products. | Meta description is written to the product, collection or page SEO fields. Meta title is always written to products, and to collections and pages only when Push meta_title to categories, collections, and pages is on (Store Settings → General). |
| JSON-LD | A BigCommerce Widget API placement per entity, rendered server-side by Stencil. | Stored on the entity and rendered by the Clione SEO app embed of Clione's Shopify app. |
| Visible FAQ accordion | Rendered by the Schema Injector script. | Stored on the entity (clione.faq_html) and rendered by the Clione FAQ app embed. |
llms.txt and one .md document per entity | On the domain you point at Clione and verify. | On your storefront under /apps/clione/, through Clione's app proxy. |
The Shopify storefront output — both app embeds and the app proxy — comes with Clione's own Shopify app. Headless storefronts read the same signals through the SEO signals API or the @clione/seo SDK.
The </> tab on every entity page shows the structured data — what crawlers and AI agents see.
Language and markets
The model writes in the same language as your content: an English description gets English copy, a Spanish one Spanish copy. Search keywords follow the same rule — they're not translated.
On Shopify, a store that sells in several languages can manage them under Store Settings → General → Markets. Read markets from my store loads the languages your store sells in; turn on the ones you want. For each market that's on, Clione writes a translated title, description, FAQs and structured data, adapting the copy you've already approved rather than generating new facts. Localizing uses one credit per entity per market. Your plan includes a number of markets; each one beyond that is the Extra market add-on. Markets aren't available on BigCommerce.
Progress bar
While an enrichment runs, a bar at the bottom of the dashboard shows it:
- Running — bulk runs show
processed / total. - Done — stays visible until you move to another page, start another operation, or dismiss it.
- Error — stays visible until you dismiss it.
The bar ends when the enriched copy is saved in Clione. Publishing is a separate step.
Per store
Enrichment is per store. The same SKU on a BigCommerce store and on a Shopify store is enriched separately, and the results may differ slightly because their source data differs.
Troubleshooting
"Quota reached" after a bulk run — The run enriched what fit in your remaining quota and skipped the rest. Your selection is kept: run it again after the reset on the 1st, upgrade, or buy Expand credits.
A re-enrich gives different scores — The model doesn't return identical text on every run. Fill Teach the AI → Verifiable evidence to anchor the scores — see Understanding quality scores.
Enriched, but the storefront still shows the old meta tags — The enrichment hasn't been published. See Publish, then check Verification.
Shopify: JSON-LD is missing from the live page — The Clione SEO app embed isn't enabled in your theme. Open the theme editor → App embeds and turn it on. See Schema Injector — Shopify.