On Belay

Merchandising

Hundreds of products. Each one deserves expert copy. Nobody has the hours.

Merchandising writes brand-consistent product content for your whole Shopify catalog — FAQs, SEO meta, and full descriptions — and routes every piece through a review queue before anything goes live. You generate, revise, and decide what publishes — all included in your plan.

What it does

Merchandising is an AI content pipeline for Shopify stores. It generates three kinds of content per product:

  • FAQs — question-and-answer pairs written to Shopify as structured metaobjects and rendered on the product page through a Liquid accordion. Because they're real FAQ schema, Google can index them.
  • SEO meta — a Google-optimized title and meta description per product, sized to the limits search engines actually respect.
  • Product descriptions — full body HTML, including Schema.org structured-data markup.

Before it writes a word, it researches your brand. A Brand Intelligence pass reads your store, looks at competitors, scans review and community signals, and builds a brand profile so the content sounds like you — not like generic AI. From there the pipeline discovers products, generates the content types you've chosen, and holds everything in a review queue. Nothing is auto-published, ever.

Who it's for

Specialty Shopify retailers — brands where product knowledge drives the purchase. Espresso equipment, audio gear, outdoor kit, specialty food. If your catalog is large and your products need expert, brand-true copy, this is built for you.

Prerequisites (required integrations)

  • Shopify — required. This is the only integration Merchandising needs. Connect it first at Dashboard → Integrations. The store domain is read from your connected Shopify credential; you never paste keys into the fieldset.

If Shopify isn't connected when you enable the fieldset, On Belay tells you to connect it before you can run anything.

Setup (step-by-step)

  1. Enable the fieldset. Go to Dashboard → Fieldsets, find Merchandising, and enable it (admin only). You'll pass through the requirement check and consent step described in the fieldsets overview.
  2. Connect Shopify if you haven't already. Once it's connected, Brand Intelligence starts automatically in the background. It takes a few minutes; the card on the fieldset's overview page polls and shows phase-by-phase progress. You can keep using the page while it runs. If it fails, the pipeline falls back to default prompts.
  3. Calibrate on one product. Open the fieldset and run the calibration flow at Dashboard → Fieldsets → Merchandising → Calibrate. This generates content for a single product so you can see the brand voice in action and tune the prompts before committing to a full catalog run. Get one product right, then scale.
  4. Adjust settings and prompts. Visit the settings page at Dashboard → Fieldsets → Merchandising → Settings to set your run defaults, and the Prompts editor to refine the writing instructions per content type. Editable prompts are the heart of this fieldset — they're how the AI learns your voice instead of a template.

Configuration

Everything is managed from the Merchandising pages inside the dashboard:

How do I calibrate before my first run?

Calibration lets you prove the brand voice on a single product before you commit to a full catalog run. You open Dashboard → Fieldsets → Merchandising → Calibrate, generate all three content types for one product, read the output, tweak your writing instructions, and regenerate until it's right. Once you're happy, you unlock the full pipeline.

The steps:

  1. Pick a product. Search your Shopify catalog and select up to three products to calibrate on. (Most people use one.)
  2. Generate a sample. Click Generate Sample. It generates the description, FAQ, and SEO meta for the selected product using your current writing instructions. Generation takes about 10–20 seconds per content type; the result lands in your review queue.
  3. Review each content type. The output is shown across Description, FAQ, and SEO Meta tabs. Read each one, then click Looks good to mark it reviewed and advance to the next.
  4. Tune and regenerate if needed. A read-only panel shows your Writing Instructions alongside the output. If anything looks off, click Edit in Writing Instructions, make your changes, come back, and click Regenerate to see the updated output. Regenerating is unlimited and free.
  5. Run the full pipeline. Once all three content types are marked reviewed, click Run Full Pipeline to unlock the full catalog run.

Calibration is a one-time gate. Once you've run the full pipeline for the first time, the Calibrate page is permanently retired and you're taken straight to the fieldset. Any org member can calibrate.

How do I review and approve content?

Generated content collects in the Review Queue at Dashboard → Fieldsets → Merchandising → Review. The queue has three tabs — Products, By Collection, and FAQ Styling — and a filter bar for narrowing what you see.

Filtering the queue. On the Products and By Collection tabs you can filter by content type (All, FAQs, SEO Meta, Descriptions), by status (Pending, Approved, Flagged, Live on Shopify, or All statuses), and — when your catalog has them — by product type, collection, and vendor. The default view shows pending items. A Clear all link resets the filters.

Per-product detail. Each product card expands to show the content generated for it, split into FAQs, SEO / Meta, and Description sections. For each section you can:

  • Approve — marks that content type approved; it publishes to Shopify on the next nightly writeback. The card confirms "will be published to Shopify tonight."
  • Flag — marks the content type flagged so it won't publish, signalling it needs another look.
  • Edit — open inline editing. FAQs edit as question/answer pairs, SEO meta as title and description fields (with character counts and a Google SERP preview), and descriptions as a live preview or raw HTML. Edits save against the product.
  • Exclude — permanently remove the product from the pipeline (see the exclusion list).

Before and after. SEO titles, meta descriptions, and descriptions show the previous (live) value beside the new one, so you can see exactly what will change. After a content type has been published, a Revert control restores the previous version on Shopify and resets the status to pending for re-review.

Quality checks. An AI fact-checker reviews each piece against the product page. If it finds a claim it couldn't verify, the product is automatically flagged with a banner, and a Content Quality panel summarizes what to check before publishing. FAQ cards with an unverifiable claim are marked individually so you can spot them.

There is no auto-publish. A human approves before anything reaches your storefront.

What can I control in Content Settings?

Open Dashboard → Fieldsets → Merchandising → Settings (titled "Content Settings"). The form is one Save settings button for the whole page. It controls:

  • Brand Voice — free-text writing instructions appended to every generation prompt, plus a list of competitor brands to avoid naming in generated content. This is the most direct way to shape how the AI writes.
  • Target Markets — which regions your products sell in (US, EU, UK, AU, CA, Global). This drives geo-specific FAQ content like voltage and warranty law. Selecting Global generates neutral content; selecting specific markets generates region-specific content.
  • Content Types — turn SEO Meta, Product Descriptions, and FAQs on or off. Disabled tracks are skipped entirely — no tokens used.
  • Price Thresholds — a minimum price for descriptions & meta (default $400) and a separate minimum for FAQs (default $200). Products priced below a threshold are skipped, which is useful for accessories and low-margin items.
  • Exclude Patterns — substrings (for example, "Open Box" or "Refurbished") matched case-insensitively against product title and product type; matching products are skipped during generation.
  • Draft Products — by default Shopify draft products are excluded from the pipeline and review queue. Enable Include draft products to process them.
  • Excluded Collections — Shopify collection handles (lowercase, e.g. accessories, open-box) whose products are skipped during runs and hidden from the review queue. Matched exactly against the collection handle.
  • Notifications — the Slack channel where new-product alerts are posted. The On Belay Slack app posts to public channels without an invite; leave it unset to turn off alerts. Setting the channel is admin-only.

The exclusion list

The Exclusion List (at the bottom of Content Settings, and reachable via Manage exclusions) is a permanent product blacklist. Products on it are removed from all future pipeline runs and never appear in the review queue. You add a product to it from its review page using Permanently Exclude — you can attach an optional reason (saved for your reference only, never sent to Shopify). The Content Settings page lists everything excluded, with its handle, reason, and date, and a Remove button to undo an exclusion.

This is different from Exclude Patterns (which skip products by substring) and Excluded Collections (which skip whole collections) — the exclusion list targets individual products by name.

How do I run the pipeline myself?

Open Dashboard → Fieldsets → Merchandising → Run. You pick a scope, choose which content types to generate, and start the run, then watch live progress.

  1. Choose a scope. Pick By Collection, By Product Type, or All Products. For By Collection, search and select a Shopify collection (it shows the product count); for By Product Type, pick from your store's product types. (If you arrived from a "new products" alert, a pre-scoped "specific products" option appears too.)
  2. Choose content types. Toggle FAQs, SEO Title + Meta, and Product Descriptions. FAQs and SEO meta are on by default; descriptions are off by default. You can preview the active writing instructions for the selected types inline, or jump to Edit.
  3. Start the run. Click Start Run.

The pre-flight cap warning. Runs are capped at 200 products each. When a run could exceed that cap — any All Products or By Product Type run, or a collection larger than 200 — On Belay opens a confirmation dialog that counts the eligible products and tells you exactly how many it will process this run. If your catalog is larger than the cap, the dialog explains how many runs it'll take to cover everything and that each run picks up where the last left off (already-generated products are skipped). The count is generation-free: it makes no AI calls and writes nothing. Small collections and specific-product runs launch directly without the dialog, since they can't be truncated.

The dialog is honest about cost: it tells you the run generates content for that many products using Claude Opus and may take a while, and reminds you generation is free — you only pay when you approve and publish.

How do I edit the writing instructions (prompts)?

Editable prompts are the heart of this fieldset — they're how the AI learns your voice instead of a template. Open Dashboard → Fieldsets → Merchandising → Prompts (titled "Writing Instructions"). A left rail lists one prompt per content area, grouped under Product FAQs, SEO Copy, and Collections; pick one to edit it in the main editor.

  • Edit and save. Change the prompt body in the editor and click Save. On Belay seeds proven defaults, so you're always editing from a working starting point. If a prompt draws on your Brand Intelligence profile, that context is shown in a read-only panel above the editor so you can see what the AI already knows.
  • Version history. Click History to see every saved version with its date and editor. Each save adds a version; you can Restore any earlier one. The current version is marked.
  • Reset to default. Reset to Default replaces your current instructions with the original default. Your current version is saved to history first, so a reset is always reversible.
  • Variants. Variants forks the prompt into a named alternative for side-by-side comparison. Variants run in parallel during the next generation cycle so you can compare approaches. Variant editing is admin/API-only for now.

How it runs (schedule)

Merchandising runs as a nightly background pipeline:

  • Product discovery runs daily around 6am UTC — it syncs your Shopify catalog and detects new products, posting a Slack alert when it finds them.
  • Generation runs daily around 7am UTC for enrolled orgs with pending products.

You can also kick off a run yourself from the Run page any time. Discovery and generation only produce drafts — they never publish on their own.

Outputs & where they land

Approved content is written back to Shopify by the nightly writeback pass:

  • FAQs become Shopify metaobjects assigned to the product, rendered through the FAQ accordion Liquid snippet on the product page.
  • SEO meta updates the product's title tag and meta description.
  • Descriptions replace the product's body HTML, including Schema.org markup.

Each write is stamped (for example, faqWrittenAt) so the same content is never published twice. Publish history is visible on the fieldset's Publish History page.

Billing

Merchandising is included in your On Belay plan — there are no per-fieldset, per-publish, or usage charges. Generate, revise, and publish as much as your catalog needs; the only thing you pay for is your base plan tier, billed through Shopify Managed Pricing on top of your 14-day free trial. See Plans & billing for tier details.

Tips

  • Calibrate before you scale. One good product tells you whether the prompts are right. Fix the voice on a single item, then run the catalog.
  • Lean on the prompt editor. The default prompts are a strong start, but the brands that get the most out of this fieldset tune the prompts to their own voice.
  • Start narrow. Run By Collection or By Product Type first to build confidence before an All Products run.
  • Generation is unlimited — use it. Regenerate as many times as you like; it's all included in your plan, with no per-publish charge.

Related