mapou
Log inStart free

Article

AI tells allergy shoppers the generic is the same. After a full conversation, 6 in 15 still bought a brand.

In one-off answers, ChatGPT, Gemini, Claude and Grok steer allergy shoppers to the generic 85% to 100% of the time. mapou ran 20 full shopping conversations to see what people actually ended up buying. It depended a lot on which assistant they asked.

purchases after a full conversation were the generic or a store brand
8 of 15
purchases after talking to Grok were generic
4 of 4
purchases after talking to Gemini were generic only
0 of 3

Published October 4, 2026. mapou allergy medicine study, September 2026: 20 shopper conversations and 55 single questions across four AI engines.

Ask an AI assistant about allergy medicine and it will usually tell you the generic is the same thing for less. In mapou's September study, ChatGPT, Gemini, Claude and Grok steered shoppers to the generic in 85% to 100% of one-off answers.

But a shopper rarely asks one question. So mapou ran 20 full shopping conversations, 5 with each assistant. Each simulated shopper kept asking until they decided to buy something or to see a doctor.

15 of the 20 conversations ended in a purchase. 8 of those were the generic or a store brand. 6 were a brand, and 1 was both.

The numbers

What shoppers decided after talking to each assistant
AssistantBoughtGeneric or store brandBrand, or brand and genericSent to a doctorSteered to generic, one-off answers
ChatGPT431185%
Gemini3032100%
Claude413185%
Grok4401100%

20 conversations, 37 assistant turns, September 2026. Steering is from one-off answers to 55 allergy questions.

Grok's shoppers bought the generic every time, 4 of 4. Gemini's never bought the generic alone: 2 bought a brand and 1 bought both. Claude's mostly bought brands too, 3 of 4.

So the assistant that steers hardest in a single answer isn't always the one that moves the purchase. Gemini steered to the generic in 100% of one-off answers, tied with Grok for the most. Over a full conversation, its shoppers still left with a brand.

Store brands came up by name. 3 purchases were a store brand, like Kirkland at Costco. Costco was also the 8th most-named name across all the one-off answers, brands included, in 14% of them.

Some shoppers were sent to a doctor. 5 conversations ended there, three of them for the 67-year-old on blood-pressure medication, who is told to avoid decongestants.

What each shopper decided, by assistant
ShopperChatGPTGeminiClaudeGrok
67, on blood-pressure medicationSee a doctorSee a doctorSee a doctorGeneric
38, severe hay feverGenericBrandBrandGeneric
27, new cat, new allergyBrandBrand and genericBrandGeneric
34, daily user, price shopperStore brandSee a doctorStore brandSee a doctor
41, buying for a six-year-oldGenericBrandBrandStore brand

Final decision of each conversation.

What it means

mapou's read: the generic pitch is simple and true, so AI repeats it in almost every single answer. But in a real conversation the shopper adds details, like a stuffy nose that needs a spray, a child's dose or a preference for one form, and the recommendation often lands on a named product. Which product depends heavily on the assistant.

What this means for brands. If you're a national brand, the single answer is lost to the generic, but the conversation isn't. Make the specific reasons to choose you (the form, the dose, the speed, the age range) clear on the pages AI reads, because that's what comes up once a shopper explains their situation.

What this doesn't show

This is a small sample, 15 purchases, so treat the differences between assistants as directional. The shoppers are simulated. It shows what AI recommends, not whether the advice is medically right.

How we measured this

Conversations
20 synthetic shopper conversations, September 2026: 5 allergy personas (for example a 67-year-old on blood-pressure medication and a parent buying for a six-year-old), each run once against ChatGPT, Gemini, Claude and Grok. The shopper keeps asking until it decides to buy or to see a doctor.
What was counted
Each conversation's final decision, hand-classified as generic (the ingredient name), store brand (for example Kirkland), brand (for example Flonase), both, or see a doctor.
Single answers
55 allergy questions asked once of each engine. Steering to the generic was detected with fixed phrase patterns (same active ingredient, bioequivalent and so on).
Engines
ChatGPT (gpt-5.4-mini), Gemini, Claude and Grok through one shared search layer with web search on.
Limits
Small samples: 15 purchases, 3 to 4 per assistant. This measures what AI recommends, not whether the advice is medically right.

Cite this article

mapou, “AI tells allergy shoppers the generic is the same. After a full conversation, 6 in 15 still bought a brand.,” mapou.ai, October 4, 2026. https://mapou.ai/articles/ai-sends-allergy-shoppers-to-costco

Reporters can reach mapou through for more on the method, other breakdowns or a comment.

Does AI send your shoppers to the generic?

mapou runs full shopping conversations in your category every month and shows how often each AI assistant moves shoppers from your brand to a generic or a store brand.

Monthly research newsletter

Get the next article by email.

Short findings like this one, plus the monthly panel report. Unsubscribe in one click.

We never share email addresses. One unsubscribe link in every send.

All articles

See where your brand stands

The workspace runs the same buyer questions for your brand and your competitors, every month. Free during early access.

Start free