AI Product Tools
Generators

Bulk Description Generator

Write short and long product descriptions and tags in bulk, choose how existing content is treated, and see the result on the storefront.

The Bulk Description Generator writes a product's short description, full description, and tags, for one product or thousands at once. This page covers what makes descriptions different from the other five generators: the content types, the existing-content options, Advanced-path settings, and the storefront result. For selecting products, running a batch, reviewing results, retrying a failure, and applying, see How bulk generation works: every step there applies here unchanged.

Where to find it

Open it from AI Product Tools > AI Generators, the AI Bulk Description Generator card, or go straight to admin.php?page=aipt-advanced-bulk-generator. The page itself is titled "Description Generator".

Before you start

  • A BYOK provider connected (OpenAI, Google Gemini, Claude, or OpenRouter) under Settings > Connection.
  • The Shop Manager or Administrator capability to generate content.
  • Decide whether you want this content type on its content structure (the default, recommended path) or on raw Custom prompts (Advanced); see How content structures work for the inline "1 Content structure" tab this page opens on.

Choose what to write

Three toggle chips above the product list, Short Description, Full Description, and Tags, decide which fields this run touches. At least one must stay on.

Description Generator toolbar with Short Description, Full Description, and Tags content-type chips, and a compact Generation Settings popover open showing only an Auto-approve switch and the same three chips
While this content type runs on its content structure, Generation Settings only offers Auto-approve on success; length, style, and format come from the structure itself.

The Generation Settings button opens a popover with an Auto-approve on success switch (once a cell succeeds, it is applied automatically, except a cell the content structure held back for review) and the same content-type chips. While descriptions run on their content structure (the default), that is everything in this popover: length, tone, and section layout belong to the structure's own editor, not to a per-run setting here.

The existing-content controls

Two controls above the product list, shared with the other generators but worth detailing here because of the extra Keep images in descriptions toggle next to them:

Run bar showing Start from set to Product data, Existing content set to Replace with a Keep images in descriptions checkbox, Results, Provider and Model
Start from, Existing content, and the description-only Keep images in descriptions checkbox, in the run bar.
  • Start from: Product data writes fresh from the product's own fields; Improve existing text rewrites what is already there.
  • Existing content: Replace (default), Fill empty only, Keep and add before, or Keep and add after.
  • Keep images in descriptions: on by default. When the current full description already contains images, turning this on keeps them in place instead of letting the rewrite drop them.

Tags follow the same Existing content choice: Replace overwrites a product's tags, anything else keeps the current tags and adds new ones alongside them.

Running on a content structure vs. Custom prompts (Advanced)

By default, descriptions are written from the content structure you edit on this page's own "1 Content structure" tab: an ordered list of sections (opening paragraph, features, specifications, and so on), a shared voice, and source data. This is the recommended path and needs no prompt writing; see How content structures work for the editor itself.

A content type can instead be switched to the older Custom prompts (Advanced) path from Settings > Advanced prompts, for raw prompt control. While Advanced is active, Generation Settings grows four extra fields and the toolbar shows the current length as a chip:

Generation Settings popover on the Advanced path, showing Full Length set to Long, Short Length set to Standard, Writing Style set to Professional, Output Format set to HTML (rich), Max Tags set to 3, and Additional Tags set to 2
On the Advanced (custom prompts) path, Generation Settings adds Full/Short Length, Writing Style, Output Format, and the two tag-count fields. A content structure ignores every one of these.
SettingOptionsDefault
Full LengthShort (~100 words), Medium (~250), Long (~500), Very Long (~800+)Long
Short LengthBrief (~20 words), Standard (~40), Detailed (~70)Standard
Writing StyleProfessional, Encouraging, Exaggerated, Friendly, Storytelling, Minimalist, Luxurious and Elegant, Adventurous, Educational, HumorousProfessional
Output FormatHTML (rich) or Plain textHTML (rich)
Max Tags1 to 20 total tags3
Additional Tags1 to 10 new tags, used only with "Improve existing text" plus Tags2
Output Format dropdown open showing HTML (rich) and Plain text options
Plain text asks the model to avoid HTML and Markdown entirely, and strips any that slips through.

None of these five fields are plan-gated: they are available on every plan, they simply stop applying once a content type is confirmed to run on its content structure.

Output language

The language a structure writes in is set once, store-wide, on that structure's own Rules tab (reached from "1 Content structure"), not as a per-run choice here. A content type still left on Custom prompts (Advanced) instead writes in the site's single default language, with no per-run language switch on this page.

Comparing the two paths. While a content type runs on Advanced, a "Compare with the content structure on 1 product" link appears next to the toolbar for anyone who can manage structure settings. It runs both paths once on a single product you choose, using two AI requests, and changes nothing until you decide to use the result.

Compare with the content structure panel: a product picker, a Compare on one product button, and a Keep custom prompts link
See what the content structure would write for one real product before switching a whole content type over to it.

A real run, start to finish

The screens below follow one real batch: five Accessories products, generating Short Description, Full Description, and Tags together.

Select products and content types, then generate

Check the products, tick the content types, and click Generate Bulk. A product that does not yet match its content structure shows a small warning badge and "Built-in structure" label right in the list, so you can see at a glance who is affected before you even generate; see the pre-flight check in How bulk generation works.

Review each result

Click a product to open its side panel. Short, Full, and Tags are three separate tabs; each shows the generated text (or tag chips) next to the product's current value, with Apply, Reject, Regenerate, and Edit for that one cell.

Side inspector panel for a Cap product, Full tab active, showing the current lorem-ipsum description above the generated replacement with a Features bullet list
The Full tab compares the product's current description against the generated one before you apply anything.
Side inspector Tags tab for a Cap product showing eight generated tag chips, a Structure: Balanced Product Description label, and a Rewrite one section box above the Apply, Reject, Regenerate, and Edit buttons
The Tags tab shows every generated tag as its own chip, ready to apply, edit, or reject. The box below it is Rewrite one section, see below.

For a result written by a content structure, the panel also offers Rewrite one section (visible above, under the tag chips) for seven days after the run: pick one section, FAQ group, or companion from the structure and regenerate only that part, leaving the rest of the row exactly as it already is. A result from Custom prompts (Advanced) never shows this, since there is no structure section to target.

Click Edit to adjust a cell before applying. For Short or Full Description this opens a Visual/HTML editor; for Tags it opens a comma-separated tag list with a live count.

Editing description modal for a product's tags, showing eight tag chips and a comma-separated tags textarea with an 8 tags Saved indicator
Editing a generated result before you apply it. Tags get their own comma-separated editor; Short and Full get a Visual/HTML toggle.

Apply one, or apply several at once

Applying a single cell is immediate.

Product row for Belt showing green Applied status and checkmarks on Short, Full, and Tags after a single-cell apply
A single Apply writes immediately and marks the row Applied, with a Cell applied successfully confirmation.

Selecting two or more products with results waiting turns on the selection strip's Apply all (N) and Reject all (N):

Selection strip showing 2 selected with Apply all (2) and Reject all (2) buttons above two pending rows
With two or more succeeded products selected, Apply all and Reject all act on all of them together.

Acting on more than one product this way asks you to confirm first, since it overwrites existing product data:

Caution dialog: Apply to Selected Products, Apply pending content for 2 product(s)? This will overwrite existing product data, with Cancel and Apply All buttons
Applying to several products at once confirms first. A single product's Apply does not.

Check Content History and the storefront

Every apply captures the product's previous value in Content History first, labelled with which structure and version wrote the new one.

Content Bank History tab listing recent Description Generator snapshots for Cap, Beanie, and Belt, each with a Restore button
Every applied description and tag set lands in Content History before you ever need it, labelled by content type and the structure that wrote it.

The storefront reflects the change immediately (behind any page cache your host or a caching plugin keeps):

Belt product page on the storefront before generation, showing the WooCommerce sample text This is a simple product and no Tags row
Before: the WooCommerce sample short description, and no tags at all.
Belt product page on the storefront after generation, showing the generated short description and a Tags row listing everyday essentials, gift idea, leather belt, small batch, unisex
After: the short description and five tags, visible right under the price.

The full description renders separately, in its own Description tab further down the same page, never in the spot above:

Belt product page's Description tab showing the generated full description with a Features bullet list
The Description tab holds the full description. Short and Full never share one spot on the storefront.

Bulk tag suggestions

There is no separate tagging tool: Tags is simply one of this generator's three content types, alongside Short and Full Description. The AI Toolkit dashboard watches your catalog and, once more than 10 products have no tags at all, replaces its usual headline with a direct link back here:

AI Toolkit dashboard greeting reading 27 products have no tags yet, with a Tag them in bulk link
Once more than 10 products are tagless, the dashboard's headline links straight back to this generator.

Click Tag them in bulk, then tick only the Tags chip (toggle Short and Full Description off first) to run a tags-only batch across your filtered selection.

Reference

Content types

TypeWrites toNotes
Short DescriptionThe product's WooCommerce short descriptionShown near the price on the storefront.
Full DescriptionThe product's WooCommerce long descriptionThe main product-page description.
TagsWooCommerce product tagsAdded alongside or replacing existing tags, depending on Existing content.

Existing-content behavior

SettingEffect on Short/FullEffect on Tags
ReplaceOverwrites the current valueReplaces existing tags entirely
Fill empty onlyWrites only if currently empty; otherwise kept, shown as a skip rather than a failureWrites only if the product has no tags yet
Keep and add before / afterPlaces the new text immediately before or after the current valueNot applicable; tags have no order to insert into

Free vs Business

CapabilityFreeBusiness
Short, Full, and Tags, any provider, any writing style/length/format/tag-count settingYesYes
Show me first and Apply now execution modesYesYes
Send to review queueNoBusiness, see How bulk generation works
Example Mode: paste one ideal short and full description, and the AI mirrors its exact style, length, and structure instead of using length/style settingsNoBusiness

Legacy Pro accounts keep Example Mode, even though every other Business-only row on this page needs a current Business plan.

Example Descriptions section on Free, showing a teaser card instead of the example text fields
On Free, Example Descriptions is a teaser card rather than two locked fields, under Settings > Prompt library > Examples.
Example Descriptions section on Business, with real Example Short Description and Example Full Description text areas
On Business, paste a real short and full description here; every later run on the Advanced path mirrors it exactly.

Example Mode lives under Settings > Prompt library > Examples and only takes effect on the Custom prompts (Advanced) path; a content structure ignores it. If your plan loses access after examples were already saved, a run that would have used them falls back to the standard prompt instead, and the page shows an amber notice saying so.

Troubleshooting