E-commerce
June 29, 2026
"Is it peanut-free? My son is allergic." The agent replies from memory: "Yes, normally it's fine." The product contains sesame, which is not mentioned on the obsolete sheet. An unwise formulation in support costs more than a return: it can cost trust, or even much more.
EU Regulation 1169/2011 requires the declaration of 14 major allergens before online purchase (EU Commission, FIR 2011). Lynkpim reminds us that allergens in free text are not enough: structured, auditable attributes are required on each SKU (Lynkpim, food taxonomy 2026).
This guide #261 covers allergy and dietary restriction support in depth. Distinct from regulated products (#119) (transversal framework): here, food protocols, macros, and escalations only.
Summary
Why do allergies represent the main food safety risk?
Allergy support requests are not like other product questions. An error can have health consequences.
Three levels of risk
Pre-purchase: customer hesitates, needs factual certainty
Post-purchase doubt: label received vs website info
Reported reaction: empathy urgency + quality escalation
Typical food DTC volume
15 to 35% of pre-purchase tickets relate to allergens, diets (vegan, gluten-free) or traces. AdsX notes that AI assistants increasingly recommend based on dietary restrictions: your support responses feed or corrupt this layer (AdsX, AI allergens 2026).
Delicatessen example
Brand with 280 tickets/month, 22% allergen questions. Agents without a protocol: 4 "risk-free" formulations in 30 days. Deployment of ALLERGY macros + metafields: 0 formulation incidents, allergy segment CSAT 4.6, pre-purchase allergy conversion +18%.
How does it differ from neighboring guides?
Four contents, four scopes.
Regulated products (#119)
Regulated (#119): cosmetics, food, supplements, REG disclaimers. #261 goes deeper into allergens and diets alone.
Cosmetics advice (#146)
Cosmetics (#146): skin INCI, skin reactions. #261: 14 food allergens, not perfume or hyaluronic acid.
Supplements (#147)
Supplements (#147): EFSA claims, dosage. #261: foods and snacks, with a focus on factory cross-contamination.
Allergies bot (#262)
Allergies bot (#262) automates allergen_check and diet_tag_query intent with RAG metafields, safety filter, and medical handoff. #261 sets the human policy and macros that bot #262 will apply.
Conversion signals (#260)
CRO signals (#260): tag signal_ingredient reports spikes in allergen questions.
Which allergens and dietary restrictions should be structured into data?
The allergen support matrix relies on Shopify fields, not on agent memory.
14 EU allergens (FIC 1169/2011)
Gluten, crustaceans, eggs, fish, peanuts, soy, milk, nuts, celery, mustard, sesame, sulphites, lupin, molluscs. US FALCPA: 9 major + sesame (FASTER Act 2021).
Recommended metafields by SKU
allergens_contains[]: official list presentallergens_may_contain[]: documented tracesdiet_tags[]: vegan, vegetarian, gluten_free_cert, halal, kosherfacility_notes: shared or dedicated linenutrition_per_100g: macros if keto/FODMAP question
Three values per allergen
Lynkpim: Contains / May contain / Free from. Never leave blank if active food product. Monthly audit: top 20 revenue SKUs, check PDP consistency = metafield = photo label.
Common customer restrictions
Vegan, vegetarian, gluten-free (certified vs. naturally), lactose-free, halal, kosher, paleo, keto, FODMAP, diabetes. Distinguish medical allergen vs. diet preference: different tone and caution level.
What is the 7-step agent protocol for an allergen query?
The agent allergen protocol standardizes each ticket without cold rigidity.
Welcome: empathy, take the allergy seriously
Identify exact SKU: variant, batch if already ordered
Open PDP + metafields: single source of truth
Quote official list: contains + may contain, word for word
Never guarantee "risk-free": traces possible if documented
Recommend checking received label: packaging takes precedence over site if there is a discrepancy
Refer to doctor if severe allergy or past reaction
Forbidden phrases
"Normally it's fine for you"
"I'm sure there are no nuts"
"Our factory is super clean, no traces" (without doc)
"My allergic colleague eats this"
Validated phrases
"According to the official product sheet, the declared allergens are: [list]. Possible traces: [list or none documented]. Check the label upon receipt and consult your doctor if you have a severe allergy."
Which ALLERGY-* and DIET-* macros should be used?
Eight major food allergies cover 90% of tickets. Completes REG-FOOD-001 from #119.
ALLERGY-LIST-01 (pre-purchase)
"Hello [First Name], for [product], declared allergens: [contains]. Possible traces: [may_contain or "none documented"]. Source: product sheet [link]. Check the received label."
ALLERGY-COMPARE-01 (two SKUs)
"[SKU A] contains: [X]. [SKU B] contains: [Y]. I cannot recommend "the safest": compare with your allergist."
ALLERGY-LABEL-MISMATCH-01
"The received label takes precedence. Send photo of batch + date. Quality escalation within 24 hours. Do not consume if in doubt."
ALLERGY-REACTION-01 (post-consumption)
"I am sorry. Stop consumption. If symptoms occur: call 911 or your doctor. Batch number + photos: immediate quality escalation."
DIET-VEGAN-01
"[Product]: status [certified vegan / not certified vegan]. Declared ingredients of animal origin: [list or none according to PDP]."
DIET-GLUTEN-01
"[Certified gluten-free / contains oats / contains gluten]. Regulatory threshold < 20 ppm if official claim. No celiac diagnosis."
DIET-HALAL-KOSHER-01
"Certification [organization] visible PDP: yes/no. Without certificate: "not certified", do not infer."
DIET-CHECKOUT-NOTE-01
"Indicate your allergies in the "Order Notes" field at the checkout step. We cannot modify the formulation, but the preparation team will be alerted if applicable."
How to handle cross-contamination and traces?
The cross-contamination support is the most contentious area if poorly formulated.
May contain rule
Only state if documented by the factory or supplier. Otherwise: "No trace information documented for this product. Contact us if you have a severe allergy." AdsX: clearly separate contains vs may contain for AI and clients (AdsX, may contain 2026).
Dedicated vs shared line
Metafield facility_notes: "produced on peanut-free line, facility processes peanuts" = may contain peanut. Do not promise a 100% free facility without a certificate.
Oats and gluten
Naturally gluten-free oats are often contaminated. State factually: "contains oats" or "certified gluten-free" according to the PDP. Do not say "gluten-free" if the only mention is "oats".
Multi-allergen orders
Customer orders a mixed basket for a household: explain that warehouse preparation does not physically separate unless a documented service is provided. Honest macro, no false promises.
How to route bot and agent on sensitive food intents?
The allergy routing bot prioritizes accuracy over deflection.
Authorized bot intents (locked corpus)
allergen_check: pull metafields, cite listdiet_tag_query: vegan, halal, gluten certingredient_list: PDP ingredients word for wordnutrition_macros: values /100 g if requested
Mandatory human handoff
Allergic reaction, illness, emergency
Allergic child, pregnant/breastfeeding woman
Request "can I eat this with my medical condition?"
Discrepancy: label vs website
Model confidence < 90% on allergen intent
Append disclaimer REG-GEN (#119) on all bot allergen intents. Prepare bot #262. See bot governance (#142).
How to manage vegan, halal, and dietary restrictions without over-promising?
Dietary restrictions support mixes certification, preference, and allergy.
Vegan / vegetarian
PDP logo certification = yes. Otherwise list declared animal ingredients (milk, honey, gelatin). Do not assume "probably vegan".
Gluten-free / lactose-free
Regulatory claim or ingredients list. Keto / FODMAP / diabetes: cite nutritional macros, refer to a dietitian or doctor, zero personalized meal plans.
Halal / kosher
Organization certificate + number if PDP. "Not certified" if absent. Support does not validate religious compliance without proof.
Gift basket / discovery box
List allergens of each SKU included. Customer gifting to an allergic person: recommend recipient verification, no global box guarantee.
WooCommerce best practice
"Allergies or diet" checkout field visible (explicit placeholder). Feeds preparation if process is documented (WPSlash, menu allergens 2026).
How to escalate an allergic reaction and quality incident?
The allergic reaction escalation follows a strict timeline.
Timeline 0-24 h
T0: ALLERGY-REACTION-01, empathy, health first
T+15 min: tag
incident_allergy, notify leadT+1 h: collect batch, label photos, reported symptoms
T+4 h: 6-field quality sheet (#119 section 10)
T+24 h: product return, goodwill gesture according to policy, no admission of liability without legal approval
Recall or suspect batch
If same batch × 2 reports: pause SKU sales, internal RECALL macro, proactive bot message to customers of the affected batch. Document in decisions (#237).
Social Media
Story "your bar made me sick": short public response + move to private DM. No public debate on ingredients without quality approval.
What KPIs and quality controls for allergen support?
Measuring the allergen support KPI prevents formulation drift.
Leading KPIs (weekly)
Allergen tickets / total pre-purchase
% responses with metafield citation: target 100%
Prohibited formulation incidents: target 0
Allergen bot handoff: rates and reasons
Lagging KPIs (monthly)
Allergy segment CSAT
Batch quality incidents
Allergy pre-purchase conversion (tagged sessions)
PDP vs metafield audit: discrepancies resolved
Reinforced QA
Audit 100% of incident_allergy tickets. Sample 20% of allergen_query tickets vs 5% standard. Feed into the monthly audit (#259).
How does Qstomy secure allergy responses?
Qstomy reads Shopify allergen metafields and quotes only validated data, without generative medical advice.
Allergy features
Allergen RAG locked: metafields + PDP only
Intent allergen_check: ALLERGY-LIST flow
Blocklist: risk-free, guaranteed, safe for you
Disclaimer REG-GEN auto append
Emergency handoff: reaction, child, pathology
Macros ALLERGY-* sync Gorgias
Audit export: allergen tickets for QA
Quantified DTC Scenario
Snacks brand, 190 allergen tickets/month, 2 junior agents. Before Qstomy: 3 "it's fine" wordings / month, allergen FRT 4 h. After metafields + locked bot + macros: forbidden wordings 0, FRT 42 s bot + 12 min human, allergy CSAT 4.7, allergy pre-purchase conversion +21%, batch quality escalations −40% (label discrepancies detected earlier).
Explore AI customer support, Shopify, request a demo.
Which playbooks can be deployed in 2 weeks?
Playbook 1: top 30 SKU metafields (1 day)
Fill in contains, may_contain, diet_tags section 3. Quality validation + label photo. PDP Sync.
Playbook 2: ALLERGY + DIET macros (3 h)
Import 8 macros section 5. Link Gorgias tags. Agent training: 7-step protocol section 4.
Playbook 3: agent quiz 15 scenarios (1 h)
Peanut, traces, vegan, gluten, reaction, label mismatch. 90% score required before solo allergen tickets.
Playbook 4: allergen_check bot pilot (1 week)
10 SKUs, 20 test questions, handoff rules section 7. Audit 0 forbidden formulation before go-live.
Playbook 5: incident escalation (2 h)
Document timeline section 9. Notify Slack quality. Suspect lot simulation test.
Playbook 6: monthly review
KPI section 10 + audit of 20 allergen tickets. 1 PDP/metafield fix minimum. Prepare bot #262.
Useful linking
Responding to a food allergy in support is not about reassuring at all costs. It is about transmitting the exact information, recognizing the limits of your role, and escalating when health comes into play. The allergic customer who gets a factual and careful response often becomes your most loyal customer.
Enzo
June 29, 2026
