Purchase Recommendations

Measuring product purchase questions in your monitoring

How to use your own prompts to measure whether AI systems mention your brand when people ask purchase and product questions in your category.

Last updated 11 days ago

KI Score doesn’t track individual products. With your own prompts in monitoring, though, you can measure whether AI systems mention your brand when someone asks about products in your category. This article shows how to set up those prompts, how to read the results, and where the limits of this measurement lie.

Writing purchase and comparison questions

Phrase your prompts the way prospective buyers actually ask. Good examples:

  • Recommendation questions: “Which running shoes are good for beginners?”
  • Comparison questions: “What’s the difference between trail and road running shoes, and which brands are good?”
  • Problem questions: “Which shoes help with knee pain when jogging?”

Add the prompts under Prompts and set Messland (Country) to the countries where your customers buy. The article Adding your own prompts and using suggestions covers the details.

To keep several product categories apart, organize the prompts with topics and tags, for example one topic per product line. In the prompts table, metrics such as Erwähnung % (Mention %) and Tonalität (Brand sentiment) each refer to a single prompt.

Why leave out your brand name

Write purchase prompts without your brand name. A question like “Are [your brand]’s shoes any good?” almost always produces a mention of your brand. That pushes up your mention rate without telling you whether the AI recommends you on its own.

KI Score doesn’t block brand names in your own prompts. It’s your call. KI Score marks such prompts automatically: in the Tags column of the prompts table, prompts that name your brand show a “B” (Branded), all others an “N” (Non-branded). If the detection is off, for example because of a product name, click into the Tags column to correct the type. Use Filter hinzufügen → Nennt Marke (Add filter → Mentions brand) in Prompts and Responses to keep branded prompts apart from your category questions so they don’t skew them.

Brand level, not product level

Whether your brand counts as Erwähnt (Mentioned) in an answer is determined by your brand name and aliases appearing in the answer text. Product names only count if you add them as aliases of your brand. In that case, every mention of a product counts as a mention of the whole brand. KI Score doesn’t report separately which product was named.

Note: Monitoring metrics are calculated with your current brand and alias list. If you add or remove aliases, the values of answers already stored change as well.

The KI Score itself also measures at the brand level: manufacturers are compared as a brand in a market for their product category, and retailers in a market for their retail sector.

Reading results in Prompts and Responses

  1. Open Prompts to see the metrics for each purchase prompt, such as Erwähnung % (Mention %), Erwähnungslücke (Mention gap), and Tonalität (Brand sentiment).
  2. Open Responses to read the stored answers themselves.
  3. Use Filter hinzufügen → Prompt (Add filter → Prompt) to select the prompt you’re interested in.
  4. Check the Erwähnt (Mentioned) column to see which answers your brand was detected in.
  5. Open an answer to read the context in which the AI names your brand or your competitors.

The Tonalität column shows the share of classified answers in which the AI presents your brand positively, for example by recommending or praising it, and how often it only mentions it factually or criticizes it. The bar appears from five classified answers.

Limits of this measurement

  • No product level: KI Score doesn’t capture product placements, prices, or shopping cards. KI Score doesn’t offer product or shopping tracking. The Shopping item under Scoreboard is marked Bald (Soon) and can’t be used yet.
  • No consumer apps: AI providers are queried through their APIs with web search, not through the apps end users use. Some providers are explicitly labeled as approximations.
  • Cost: Every answer costs credits: one credit per prompt, AI provider, and run. If a prompt is measured in several countries, each country counts like a prompt of its own. Many purchase prompts across several providers and countries use correspondingly more.
  • No effect on the KI Score: Your own monitoring prompts don’t change your KI Score.

What teams say about KI Score

“We finally see week by week where our brand shows up in ChatGPT and Perplexity.”

Lena Vogt, Head of SEO

“The score finally gave us one number the whole team actually understands.”

Tobias Reimann, CMO

“Instead of guessing which prompts matter, we now have real data.”

Franziska Aldag, Head of Content

“Within the first week we already saw which competitor was beating us in AI answers.”

Malte Ohlsen, CEO

“Setup took one afternoon, not weeks of onboarding.”

Insa Brechtel, Head of Marketing

“Finally a tool that takes AI Overviews as seriously as classic SEO.”

Jonas Peiter, SEO Manager

Measured where your customers ask

ChatGPT
Perplexity
Gemini
Claude
Copilot
Grok
AI Overviews
DeepSeek

Ask our experts

Our team is happy to help personally — reach out anytime.