Guidelines

Guidelines tell Emfas how to write. They define tone of voice, terminology, formatting rules, and other standards that apply when AI generates content. Set them once and every automation, proposal, and bulk generation follows them automatically.

Guidelines are organized in layers. Each layer adds more specificity from organization-wide standards down to rules for a single attribute in a particular language.

How guidelines work together

When AI generates content, it combines all applicable guidelines for that attribute and language. The layers build on each other rather than overriding each other:

1. Brand guidelines apply to all AI generation across your entire organization
2. Language guidelines apply to all attributes generated in a specific language
3. Attribute guidelines apply to a specific attribute, with optional scoping by language, family, and channel

Start with broad rules at the brand level, then specialize at the language and attribute levels. Most customization should happen at the attribute level.

Brand guidelines

Brand guidelines define organization-wide standards. They apply to every piece of AI-generated content across all attributes and languages.

Use brand guidelines for rules that should never vary:

• Brand name formatting — "Always write Emfas, not emfas or EMFAS"
• Universal tone — "A professional but approachable tone across all content"
• Standard terminology — terms that must be consistent across every attribute

Keep brand guidelines minimal. The more specific you need to be, the more likely it belongs at the attribute or language level instead.

Setting brand guidelines

1. Go to Settings then Brand
2. Add your global instructions in the Tone of voice field
3. If you have multiple channels configured, use the channel selector to set channel-specific brand instructions

Instructions scoped to a specific channel only apply when generating content for that channel.

Language guidelines

Language guidelines apply to all attributes generated in a particular language. Set them for market-specific tone, cultural adaptations, and localization standards.

Examples

Market tone

For German, use a formal tone. Avoid contractions.

Cultural adaptation

Adapt idioms to sound natural in French rather than translating them directly.

Localization

For the Swedish market, emphasize sustainability and eco-friendly materials.

Setting language guidelines

1. Go to Settings then Languages
2. Click a language to open its settings
3. Add instructions in the Instructions field

Attribute guidelines

Attribute guidelines are the most specific layer. They apply to a single attribute and can be scoped further by language, product family, or channel.

This is where most guideline work happens. Common attribute guidelines include:

• Format — "Start with one sentence overview, then three to five bullet points"
• Length — "Keep between 80 and 120 characters"
• Content focus — "Emphasize materials and fit over style"
• Terminology — "Use slim fit instead of skinny. Never use flattering."

Language scoping

By default, attribute guidelines apply across all languages. Switch to a specific language to set rules that only apply when generating in that language — useful when a particular attribute needs different formatting or tone per market.

Family scoping

If you have families configured, you can set different guidelines per family. For example, your Description attribute might use one set of rules for Clothing products and a different set for Accessories.

Select a family from the family bar above the instructions field to view or edit family-specific guidelines.

Channel scoping

If you have multiple channels, attribute guidelines can be scoped per channel. Guidelines scoped to a channel only apply when generating content for that specific destination.

Setting attribute guidelines

1. Go to Settings then Attributes
2. Select an attribute
3. Open the Instructions tab
4. Add your guidelines in the instructions field

Use the language, family, and channel selectors to scope guidelines as needed.

References

References give the AI concrete examples of what good output looks like. Instead of describing requirements in words, you show the AI existing content that matches your standards.

References are added per language and per attribute. They work alongside written guidelines — the AI uses both when generating.

When to use references

References work best when:

• Your brand voice is distinctive and hard to describe in instructions
• You want the AI to match a specific formatting pattern precisely
• You are generating for multiple markets with different writing styles
• Your existing catalog content does not reflect your current brand direction

Adding references

1. Open an attribute in Settings then Attributes
2. Go to the Instructions tab and select a specific language. References require a specific language, not "All languages".
3. Click + Add Reference
4. Paste example content that represents ideal output for this attribute and language

Add several references to give the AI a better picture of what to aim for. The more consistent your examples are, the better the AI can match the pattern.

Guidelines overview

Navigate to Guidelines from the side menu to see where guidelines are configured across your workspace:

• Brand — global tone of voice and organization-wide instructions
• Attributes — per-attribute formatting and content rules
• Translations — per-language tone and localization standards

This page is a navigation hub. Click through to each section to configure guidelines for that area.

Best practices

Start broad, then specialize. Set minimal brand guidelines for organization-wide rules. Add language guidelines for market-specific tone. Use attribute guidelines for everything else.

Be specific. Vague instructions like "write well" do not help. Instructions like "start with the material, then describe the fit in one sentence" produce consistent results.

Use references for distinctive voice. If your brand style is hard to articulate, references are more effective than lengthy written instructions. A few strong examples communicate more than paragraphs of description.

Test before scaling. When you add or change guidelines, create a proposal for a few products to verify the output meets expectations before running bulk enrichment or enabling automations.

Separate concerns by layer. Do not repeat the same instruction at multiple levels. If a rule belongs at the brand level, remove it from individual attribute guidelines. Duplication adds noise without improving results.