Can AI Assistants Tell Shoppers That My Product Ingredients Were Reformulated Accurately?
AI assistants can tell shoppers about a reformulation accurately, but only if the new formula is easier to find than the old one. Retailer listings, open product databases, review sites and photos of old packaging keep describing the previous ingredients, and assistants repeat them. Sometimes they mix the old and new lists into a formula that never existed. To get accurate answers, publish a reformulation page with both ingredient lists, the date of the change and how to tell the versions apart. Put the current list in text on the product page, and update feeds, retailers and product databases. Consider whether GS1's rules require a new barcode number. Treat any change to allergens as a safety matter, and test answers until they are right.
Why do assistants get reformulations wrong?
- Old listings outnumber new ones. Retailers, comparison sites and ingredient databases copied the old list, and many never update it.
- Ingredient lists in images. Photos of packaging are hard for assistants to read and date, and old photos stay online.
- No statement of change. If the product page silently shows a new list, nothing tells an assistant that the old one is out of date.
- Both versions on sale. During a transition, shops sell old and new stock side by side, and both descriptions are accurate for some units.
- Reviews describe the old formula, including complaints or praise about ingredients that have gone.
- Model memory holds the formula as it was during training.
How weak product content leads to wrong answers generally is covered in how poor product content leads AI to give shoppers incorrect information.
What should the reformulation page say?
| Element | Example |
|---|---|
| What changed | "We replaced palm oil with sunflower oil and removed added sugar." |
| When | "New formula shipping from March 2026." |
| Old and new ingredient lists | Side by side, in text |
| Allergen changes | Stated first and clearly, such as "Now contains sesame" or "No longer contains milk" |
| How to tell them apart | A "new formula" mark, a batch or lot code range, a new barcode, or a packaging change |
| Transition period | "Some shops may still sell the previous version until about June 2026." |
| Why | A short, honest reason, such as supply, taste or nutrition |
Link to this page from the product page and from your questions page.
What should the product page show?
- The current ingredient list in text, not only in a packaging image.
- A short note that the product was reformulated, with a link to the reformulation page.
- Allergen information in text, consistent with the label.
- Dietary claims such as vegan or gluten-free only if the new formula supports them. If a claim was lost, say so.
Does a reformulation need a new barcode?
Sometimes. GS1's GTIN management rules say a formulation change requires a new GTIN when two conditions are both met. The change must affect the legally required information declared on the packaging, such as the ingredient list. And the brand owner must expect consumers or supply chain partners to tell the difference. A new GTIN also helps assistants, retailers and databases keep the two versions apart, since each barcode then has one ingredient list. Check the rules for your case with GS1 or your GS1 member organisation.
Why are allergen changes different?
Because an error can cause harm. In the US, the FDA recognises nine major food allergens, with sesame added as the ninth from 1 January 2023. Other countries have their own lists, such as the 14 allergens in UK and EU rules. If a reformulation adds an allergen, an assistant repeating the old list could tell an allergic shopper the product is safe. For allergen changes:
- State the change prominently on the product page, the reformulation page and the packaging.
- Update every retailer and database listing as a priority.
- Test allergen questions specifically, such as "Does [product] contain sesame?".
- Report wrong answers through each assistant's feedback tools.
Handling false allergen claims more generally is covered in what food businesses should do when AI makes false allergen or gluten-free claims.
What about cosmetics, personal care and supplements?
The same problem appears outside food, with different sources. Cosmetics and personal care ingredient lists use standard international ingredient names, and ingredient-checker websites and apps copy them in bulk. Shoppers with sensitive skin rely on those sites, and assistants cite them. When you reformulate, check the main ingredient-checker entries for your product and submit the new list where they accept updates. Say clearly whether fragrance, preservatives or known irritants changed, since those are the questions shoppers ask. For supplements, the facts panel is the reference. Changes to amounts per serving matter as much as changes to ingredients, so state old and new amounts side by side. For both, "free from" claims such as "fragrance-free" or "no artificial colours" need rechecking after any reformulation.
What else needs updating?
- Merchant Center and other product feeds: descriptions, ingredient-related attributes and images.
- Retailer listings: send an updated specification with the new list, images and GTIN, and ask for confirmation.
- Open product databases, which anyone can edit. Update the entry with the new list and photos of the new pack.
- Review responses: where reviews discuss removed ingredients, reply briefly that the formula has changed.
- Social profiles and FAQs that mention ingredients.
How do you test what assistants say?
Ask ChatGPT, Gemini, Perplexity and Google's AI features:
- "What are the ingredients in [product]?"
- "Does [product] contain [ingredient or allergen]?"
- "Did [brand] change the recipe of [product]?"
- "Is [product] vegan?" or another dietary claim affected by the change
Run each several times. Record whether the answer uses the new list, the old list or a mixture, and which sources it cites. Mixtures are the most dangerous result, because they look plausible. Repeat every two weeks until answers are consistently correct.
What might an illustrative case look like?
Consider an illustrative snack brand that replaced milk powder with a plant-based alternative and began using sesame oil. Two months later, assistants described the product as containing milk and not sesame, citing two retailer listings and an open food database entry. The brand's product page showed only a photo of the new pack. It published a reformulation page with both lists and the sesame change stated first, and added the ingredients in text on the product page. It issued a new GTIN, as the change affected the declared ingredients, and updated retailers and the database entry. Within a month, most answers described the new formula and flagged the sesame content.
What do people ask about reformulations and AI answers?
How long do old ingredient lists keep appearing?
Search-based answers usually improve within weeks once the main sources are updated. Answers from model memory can lag for longer.
Should we remove old reviews that mention removed ingredients?
No. Reply briefly that the formula changed, and let the reformulation page do the explaining.
Does structured data help?
Keep product data accurate in feeds and markup, but the ingredient list in plain text on the page matters most.
Should we announce a reformulation publicly?
Yes, briefly, especially if allergens, dietary claims or key ingredients changed. A dated announcement gives assistants and shoppers a clear record that the old list is out of date.
Published by The Kovmere Team · updated Sep 29, 2026 · More resources