What you can configure
The overall voice for generated copy (for example professional or casual). You
can also supply a short custom description to steer the tone more precisely.
Free-text guidance MerchantOps should honor when writing — phrasing to prefer
or avoid, claims to steer clear of, and other house rules.
The shortlist of models your organization has picked, searched from the model
catalog. It is narrower than it sounds: the list populates the LLM model
dropdown when you set a per-property model override in the Property
Dictionary, and saving a property with a model that isn’t on the list is
rejected. It is not consulted at generation time, and a property with no model
override uses the default model configured for its prompt whether or not that
model appears here. An empty list disables the check rather than blocking
everything.
Which product attributes MerchantOps uses to build the brand-site search query
during enrichment — for example a gender-bearing title field for footwear or
apparel. Their values are joined in order; leave it empty to fall back to brand
and name.
Content settings shape how content is written, not whether a given field is
overwritten. That is decided per property by
source_priority and
overwrite_merchant_input — and note that description, features, and meta title
ship with overwriting enabled, so enrichment may replace merchant-entered
text on those three. See How enrichment works for the
full model.Grounding thresholds
Grounding decides whether a property has enough real source data behind it for AI to write it at all. Every property that generates declares the context fields it is written from (set per property in the Property Dictionary), and these three controls say how many of those fields have to carry a value. Each one is optional. Leave a box blank and the platform default applies, so an organization that never opens this section keeps the built-in behavior.default 0.5
The share of a property’s context fields that must have a value before it is
generated — half of them, by default.
default 3
A number of populated context fields that is enough on its own, however long
the list. This exists because a long context list makes the ratio unreachable:
a 32-field list would need 16 populated fields, which a single product may
never have.
default 0.25
Below this share, the property is still generated, but deliberately shorter and
kept strictly to the context that exists. It is marked “thin evidence” so you
can tell those values apart when reviewing.
Pick these numbers against your own catalog. An evidence floor of 3 means
something very different against a 32-field context list than against a
four-field one.
Scope
Content settings apply at the organization level — one configuration for your whole organization. Tone of voice and brand guidelines have no per-product-type form: all generated content uses your organization’s values for those. Other parts of enrichment are configurable more narrowly. The prompt, context fields, model, source priority, and overwrite behavior for an individual property can be overridden for one product type or a whole category — see Enrichment configuration and overrides. The structure a description is organized into and how features are presented (including caps on feature count and description length) still shape generated content, but they run on built-in defaults and aren’t editable from the Content Settings page yet. If you clear your organization’s content settings, MerchantOps falls back to sensible built-in defaults rather than leaving content generation unconfigured.Related
How enrichment works
Where generated content fits in the enrichment pipeline.
Quality review
Advisory checks that inspect the copy these settings shape.
Product types & properties
Per-property enrichment configuration and product-type overrides.
Team & roles
Who can view and change organization settings.