How do your customers use ChatGPT, Gemini and Google's AI before they buy?
What do real GenAI conversations look like in your target groups? Almost every second AI conversation touches a product category. Murmuras shows how and when your brands and your competitors' brands are mentioned and recommended in ChatGPT, Gemini and AI Mode. Along the way, you receive key analyses of topics, intents and relevant product categories. Measured in real conversations with consent, not with test prompts.
- Mentions and recommendations of your brand, by target group and category
- Stage of the decision, from inspiration to purchase intent
- Competition in the same conversation: which brands appear alongside yours
- Ads in ChatGPT: who advertises, to whom and after which question
- free of charge
- reply within 48 hours
- no presentation meeting needed
- Real AI chats instead of test prompts
- Ads in ChatGPT measurable
- Every category can be analyzed separately
See what people ask AI before they buy, and which brands it names.
Our panel classifies real conversations: by product category, stage, brands mentioned and the AI's recommendation. Every product category can be analyzed separately, from groceries to travel.
Questions this answers
“How exactly do customers use ChatGPT & co. for our industry?”
In how many real conversations your brand is mentioned and recommended, per assistant.
In the example: Drugstore & beauty“Which competitor brands does the AI name next to ours?”
Brands in the same conversation and direct comparison recommendations.
In the example: Fast food“When in the purchase process do people ask the AI?”
Stage of the decision, observed from inspiration to purchase intent.
In the example: Household“Who advertises in ChatGPT, and after which prompts?”
newAds in ChatGPT with the advertiser and the conversation before them.
We analyze what people ask AI about your category and what it answers, with the participants' consent.
Four real research journeys with AI.
Straight from the panel, one case per category: several conversations by the same person about one purchase occasion. The guided walkthrough shows how the selection changes from conversation to conversation. The conversations are summarized. When you work with us, real, anonymized conversations are analyzed and delivered, and you can also read them word for word.
Real research · brands anonymized
Fast food: burgers, chicken or supermarket?
Four conversations with Gemini over just over two weeks in which fast-food chains are mentioned. Supermarkets or discounters appear next to the chains in every conversation; at the end, two burger chains are back in the conversation, with a comparison as the recommendation.
February 2026 · AI assistant: Gemini
- Inspiration
- Consideration
- Purchase intent (not in this case)
Four conversations in 16 days in which fast-food chains are mentioned. Go through the research station by station.
Tip: use the arrow keys ← → to step through
Time relative to the last conversation
First consideration
Four days later
Almost twelve days later
Guided walkthrough
Go through the research conversation by conversation.
For each station, we explain what happened and what it tells us.
- 4 stations
- Period: 16 days
- approx. 1 minute
What this case shows
- Fast food is weighed against grocery retail: discounters or supermarkets appear next to the chains in every conversation.
- The shortlist changes: first two burger chains, then a chicken chain, and at the end the two burger chains again.
- At the end, both burger chains are back in the conversation, with a comparison as the recommendation and the most questions.
What it does not show
- Purchase intent or a purchase. No conversation is classified as purchase intent.
Real research · brands anonymized
Coffee: from branded coffee to private label
Four conversations with ChatGPT in one week. At the start, three coffee brands are in the conversation; at the end, only a discounter and its private label.
June 2026 · AI assistant: ChatGPT
- Information (not in this case)
- Consideration
- Purchase intent
Four conversations in seven days, from consideration to purchase intent. Go through the research station by station.
Tip: use the arrow keys ← → to step through
Time relative to the last conversation
Consideration
Purchase intent
Guided walkthrough
Go through the research conversation by conversation.
For each station, we explain what happened and what it tells us.
- 4 stations
- Period: 7 days
- approx. 1 minute
What this case shows
- The private label displaces the branded products: coffee brands B and C drop out after the second conversation, coffee brand A after the third.
- The stage clearly moves forward: twice consideration with a comparison, then twice purchase intent with a specific recommendation.
- Long deliberation, quick decision: eight questions in the second conversation, one in the last.
What it does not show
- Whether anything was bought at discounter A. The classification shows purchase intent, not a purchase.
Real research · brands anonymized
Skincare: from 19 mentions to purchase intent
Six conversations in Google Search's AI Mode over nine days. First brands, pharmacies and shops, then a shortlist, and finally a conversation with purchase intent in which brand A and, for the first time, brand D are mentioned.
June 2026 · AI Mode in Google Search
- Information
- Consideration
- Purchase intent
Six conversations in nine days, each with a specific recommendation from the AI. Go through the research station by station.
Tip: use the arrow keys ← → to step through
Time relative to the last conversation
Broad search
Checked one by one
Purchase intent
Guided walkthrough
Go through the research conversation by conversation.
For each station, we explain what happened and what it tells us.
- 6 stations
- Period: 9 days
- approx. 1 minute
What this case shows
- A clear funnel: from 19 mentions via a shortlist of brands down to one or two mentions per conversation, in nine days.
- The candidates are checked one after another: first only brand B, then only brand A. Brand C no longer appears after the third conversation.
- All six conversations contain a specific recommendation from the AI; most questions come early, not at the end.
What it does not show
- A purchase. The classification shows purchase intent.
Real research · brands anonymized
Cleaning: furniture store vs. drugstore private labels
Five conversations with Gemini. In one evening, the furniture store and two private labels drop out; a day and a half later, the conversation is about exactly the remaining private label.
August 2026 · AI assistant: Gemini
- Information
- Consideration
- Purchase intent (not in this case)
Five conversations in a day and a half, four of them in one evening. Two conversations a few minutes apart are combined into one station. Go through the research station by station.
Tip: use the arrow keys ← → to step through
Time relative to the last conversation
First evening
A day and a half later
Guided walkthrough
Go through the research conversation by conversation.
For each station, we explain what happened and what it tells us.
- 4 stations
- Period: 1½ days
- approx. 1 minute
What this case shows
- Within 24 minutes, the selection shrinks from six to three mentions: the furniture store and two of the three private labels drop out.
- The choice is between two retailers and their own products.
- The shortlist holds: a day and a half later, the conversation is about exactly the remaining private label.
What it does not show
- Purchase intent or a purchase. The last conversation is an information question.
Individual cases show what research looks like. How common such patterns are only becomes clear in the analysis across all participants. We only show conversations from participants who consented to the analysis, and only their classification.
Measured, not asked.
Murmuras passively measures what actually happens on the smartphone. Around 3,000 participants in Germany have voluntarily installed the Murmuras app for this purpose. They were recruited with quotas for age, gender and region, and are compensated for the digital measurement. In Switzerland, we run a similar digital measurement with around 1,000 participants through our partner Media Focus. Our on-screen technology mainly captures consumer-relevant content directly from the phone screen, as it appears in everyday life. Participants have given their explicit consent and can withdraw it at any time. We process the data on servers in Germany.
All sources come from the same people. That is why they can be linked into one path and analyzed by age, gender and other characteristics.
What we measure
- AI conversationsChatGPT, Gemini, AI Mode in Google Search
- Ads in ChatGPTAds in the chat and product cards
- Social media adsFacebook, Instagram, TikTok, YouTube, Pinterest
- Coupons and offers in retailer appsLidl Plus, REWE, PENNY, EDEKA, Netto, Kaufland, Mein dm, Rossmann, Müller
- Online shoppingUse of shopping apps such as Amazon
- ReceiptsIn-store purchases, digital or as a photo
- OtherApp usage, web and search
What does this look like for your category?
Name a category or brand. You will receive a free extract with sample conversations.
or write directly to info@murmuras.com
The whole path to the checkout
Advertising, AI conversation, coupon and receipt as one sequence. A product in its own right, in which AI conversations are one touchpoint.
Learn more