How Importier's Avoid Words and Keywords Fields Shape AI Descriptions

A skincare brand migrates to Shopify and uses Importier to generate descriptions for 140 products. The descriptions read well and the structure is correct, but the vocabulary is wrong. The AI keeps reaching for "luxurious", "indulgent", "pampering": lifestyle spa language. The brand is clinical. Every product page on the website has language like "dermatologist-formulated", "clinically tested", "fragrance-free". The AI descriptions read like a different brand's copy.

The Brand Voice description field tells the AI what the brand is: "Clinical skincare brand, dermatologist-formulated, fragrance-free, suitable for sensitive and reactive skin." That context shapes the structure and angle of the descriptions. But it does not prevent the AI from reaching for a word it has learned is commonly associated with skincare. "Luxurious" is a frequent skincare descriptor in the AI's training data. The Brand Voice description context shifts the distribution toward clinical language but does not eliminate the tail vocabulary.

The Avoid Words field eliminates the tail. Once "luxurious", "indulgent", "pamper", "spa", and "treat yourself" are added to Avoid Words, those terms disappear from every description the account generates. Not reduced: eliminated. The AI produces clinical descriptions that match what the brand actually says on its packaging and website.

What the Avoid Words Field Actually Does#

The Avoid Words field is a negative vocabulary constraint. Every description the account generates is checked against this list. Any word or phrase on the list is suppressed regardless of how appropriate it might be given the product data.

The constraint is account-wide and applies across every surface: the import wizard's AI description step, the Store Scanner, the FAQ Generator, and Variant Descriptions on Scale and Enterprise plans. Configuring Avoid Words once in Brand Voice means the constraint is active on every future generation run from that account.

The field accepts single words and multi-word phrases. "Revolutionary" blocks that specific word. "Game-changer" blocks that exact phrase. Partial matches are not blocked by default: blocking "easy" does not also block "easily" or "ease". Merchants who want to block a word family should add each variant they want to eliminate: "easy", "easily", "ease", "effortless", "effortlessly" as separate entries.

The Avoid Words field is distinct from the guidance in the Brand Voice description field. The description field shapes what the AI aims for: it describes the brand's identity, customer, and positioning. Avoid Words enforces what the AI cannot use even when it would otherwise be contextually appropriate given the product data.

A useful mental model: Brand Voice description tells the AI "write like this." Avoid Words tells the AI "never write these words, regardless."

What the Keywords Field Does#

The Keywords field is the complement to Avoid Words. Where Avoid Words eliminates vocabulary, Keywords specifies terms that must appear in descriptions.

Keywords entered in this field are incorporated into every description the account generates. The AI weaves them in naturally rather than inserting them verbatim at a fixed position. The result reads like the keyword was chosen as the right word for that sentence, not forced in to satisfy a rule.

Common uses for the Keywords field:

• Brand-specific descriptors that the merchant uses consistently but that the AI would not reach for unprompted. A luxury goods brand might enter "hand-finished", "heritage craftsmanship", "numbered edition" as terms that should appear in descriptions for relevant products.
• SEO keywords the merchant is targeting for their category. Adding "organic cotton" to Keywords for a sustainable apparel brand means this phrase appears regularly across descriptions without the merchant manually editing each one.
• Compliance terms required in certain markets. A food supplement brand might add "not intended to diagnose, treat, cure, or prevent any disease" or an abbreviated equivalent as a required disclaimer phrase.
• Shopify collection and metafield anchor terms that reinforce filter values. If the merchant has a collection filter for "sustainable materials", adding "sustainable" as a keyword means it appears in the body text of relevant descriptions and reinforces the filter's relevance signal.

Building an Effective Avoid Words List by Product Category#

The most effective Avoid Words lists are category-specific. Different product categories attract different pools of marketing clichés. Identifying the overused vocabulary in your category and eliminating it produces descriptions that stand out from both supplier copy and generic AI output.

Fashion and apparel: The AI frequently generates "chic", "on-trend", "statement piece", "boho", "wardrobe staple". For brands with a specific aesthetic (technical sportswear, minimal basics, heritage outerwear), these phrases feel generic. Add the ones that do not fit your brand's voice to Avoid Words.

Consumer electronics and tech accessories: "Revolutionary", "game-changing", "cutting-edge", "next-gen", "future-proof". These are the most overused technology descriptors. Most electronics brands want to move toward specific performance claims. Blocking them forces the AI toward factual descriptors: "dual-band", "USB-C compatible", "lossless audio", "MIL-SPEC rated".

Food and beverage: "Artisanal", "handcrafted", "small-batch", "farm-to-table". Depending on the brand, some of these are true and worth using; others are generic claims the product cannot support. Brands that need accurate, specific product language add the ones that do not apply to their range.

Beauty and personal care: "Luxurious", "indulgent", "radiant", "glowing", "pamper". Clinical brands, dermatology brands, and men's grooming brands often want to avoid the lifestyle-beauty register these words carry. Skincare brands with specific active ingredient positioning typically want descriptions to reference the ingredient and its function, not a feeling of self-care.

Home goods and furniture: "Cozy", "rustic", "farmhouse", "charming". Modern minimalist brands, industrial design brands, and commercial furniture brands consistently want to avoid the lifestyle-staging vocabulary that surrounds home decor AI descriptions. Their vocabulary is more functional: "solid hardwood", "powder-coated steel frame", "stain-resistant finish".

1. 01
   Open Brand Voice in Importier. You will see four fields

   Description, Keywords, Avoid Words, and Example Phrases.
2. 02
   In the Avoid Words field, enter the terms you want to eliminate from your descriptions. Start with the 5-10 words or phrases that appear most often in AI descriptions but do not fit your brand's voice. Separate each entry with a comma or press Enter after each one.
3. 03
   In the Keywords field, enter 3-8 terms or phrases that should appear regularly in your descriptions. These are brand-specific descriptors, SEO targets, or compliance terms. Do not overload this field

   3-5 specific terms produce better results than 15 generic ones.
4. 04
   In the Description field, write 2-4 sentences describing the brand's identity, target customer, and positioning. The Description gives the AI context for the overall angle and structure; Keywords and Avoid Words constrain the vocabulary within that angle.
5. 05
   Generate 5-10 sample descriptions with these settings before running the full catalogue. Review the sample outputs for

   does the vocabulary match? Are Avoid Words absent? Do Keywords appear naturally?
6. 06
   Adjust and re-test if the sample outputs still contain unwanted vocabulary or if Keywords feel forced rather than natural in the text. Avoid Words lists usually need a second pass after reviewing real output, as the AI has many synonyms for common concepts and you may need to block variants as well.

How Keywords and Avoid Words Interact With Persona Selection#

Brand Voice vocabulary control works alongside persona selection but does not override it. The 156 expert personas across 43 industry categories each have their own vocabulary range. A "Perfumer" persona writes about fragrance notes and sensory properties; a "Materials Scientist" persona writes about substrate composition and performance properties.

When Avoid Words are applied on top of a persona, the AI maintains the persona's knowledge framework while constraining its vocabulary. A Perfumer persona with "indulgent", "luxurious", and "sensory journey" in Avoid Words still writes with fragrance expertise; it still uses the vocabulary of top notes, base notes, and fragrance families, but the overlay of spa-lifestyle language is removed. The result is an expert fragrance description with a more factual, less promotional register.

This interaction works predictably for most Avoid Words. The exception is when Avoid Words blocks terms that are central to the persona's core vocabulary. If a Food Scientist persona has "artisanal" in Avoid Words, the AI avoids it as expected. But if the blocked word and its alternatives are both in the AI's high-frequency vocabulary for that category, all of them may need to be added. Testing the persona-and-Avoid-Words combination on sample products before the full batch identifies these edge cases.

The interaction with Keywords is simpler: Keywords are additive. The persona provides the vocabulary framework, and Keywords specifies particular terms that should appear within it. A "Sports Nutritionist" persona with Keywords set to "NSF Certified", "informed-sport.com verified", and "heavy metal tested" will incorporate these certification markers into the description when the product data supports them.

Example Phrases: Anchoring the AI to Your Actual Writing Style#

The fourth Brand Voice field (Example Phrases) is the most precise vocabulary control tool in the set. Unlike Keywords (which ensures certain terms appear) and Avoid Words (which ensures certain terms do not appear), Example Phrases shows the AI actual examples of how the brand writes. The AI uses these as stylistic anchors.

An example phrase like "Engineered for sustained performance across long trail runs" teaches the AI several things simultaneously: sentence structure (active verb plus benefit clause), register (technical-athletic rather than lifestyle), vocabulary level (sustained, engineered, trail runs rather than comfortable, great, outdoors), and the brand's tendency to lead with the benefit rather than the product feature.

According to Shopify's guidance on product page content, the most effective product descriptions speak directly to the customer's use case rather than describing the product in isolation. Example Phrases that demonstrate how the brand connects product features to specific customer scenarios give the AI clearer guidance than abstract brand personality descriptions.

Two or three well-chosen example phrases, combined with a tight Avoid Words list, often produce better vocabulary consistency than a long Brand Voice description field with no negative constraints.

“Avoid Words does not reduce: it eliminates. A word on the Avoid Words list will not appear in any description the account generates, even when the AI would otherwise use it for contextual reasons.

Testing Vocabulary Control Before Running the Full Batch#

Before applying Brand Voice vocabulary settings to an entire catalogue, testing on a representative sample is worth the few minutes it takes.

The test process:

• Select 5-8 products that span the range of the catalogue: the most unusual product name, the most generic product type, the most technically complex product, and a few average products.
• Generate descriptions for this sample with the current Brand Voice settings.
• Review each description for: Avoid Words appearing (they should not), Keywords appearing (they should appear at least once per description if the product supports it), and overall vocabulary register.
• Check for synonym leakage: the AI sometimes uses a near-synonym for a blocked word. If "luxurious" is blocked and the output says "sumptuous" instead, "sumptuous" needs to be added to Avoid Words.

The synonyms problem is the most common source of vocabulary control gaps. A skincare brand blocking "luxurious" often also needs to block "opulent", "sumptuous", "lavish", and "rich" to fully clear the luxury register from clinical product descriptions. The AI does not intentionally circumvent the Avoid Words list: it simply has a range of terms for each concept, and if the first choice is blocked it reaches for the next. Adding the full synonym family to Avoid Words closes that gap.

This testing process takes 10-15 minutes and catches synonym leakage before it appears across 200 or 2,000 product descriptions. Running the fix on a full catalogue after the fact costs more than catching it in the sample stage.

Key Takeaways for Shopify AI Product Description Avoid Words#

Vocabulary control in Brand Voice is the mechanism that turns the AI's generic product-category knowledge into brand-specific language. The four Brand Voice fields work together but serve different functions.

• Avoid Words eliminates vocabulary at the account level: any word or phrase entered in this field will not appear in any description the account generates, across every surface (import wizard, Store Scanner, FAQ Generator, Variant Descriptions).
• Keywords specifies must-appear terms: enter brand-specific descriptors, SEO targets, or compliance phrases. The AI incorporates them naturally, not as literal insertions at a fixed position.
• The two fields are complements, not alternatives: Avoid Words restricts; Keywords adds. Most effective Brand Voice configurations use both.
• Partial matches are not blocked by default: blocking "easy" does not block "easily" or "ease". Add each variant of a term separately if the whole word family needs to be excluded.
• Test on 5-8 products before the full batch: synonym leakage is the main source of vocabulary control gaps. Identifying which synonyms need blocking in the sample stage prevents cleaning up vocabulary issues across a full catalogue.
• Example Phrases give the most precise vocabulary guidance: showing the AI how the brand actually writes produces better vocabulary consistency than describing the brand in abstract terms, particularly for unusual brand voices or technical product categories.

Configure Brand Voice vocabulary control for your Shopify store at importier.app.