Your sales team knows a buyer's objections inside out. But what does that buyer ask an AI assistant first? That conversation happens before the first contact, and if you are not in the answer then, you don't make the shortlist. For B2B lead generation, that is exactly the moment you want to measure. It only works with questions that sound like those of a real buyer, not like your own product brochure. In this guide, you build a prompt set in the GripIQ Intent Engine per buying stage, per country and with tags. So on Monday, you know in which stage you drop out of the picture.
A prompt is a buyer's question, not your slogan
In GripIQ, a prompt is a question we regularly put to AI models, so you can see which brands appear in the answer. You find them under AI Visibility, in the Prompts tab.
The mistake we come across most often is a prompt set full of brand names. A question like "Is Company X a good supplier of quoting software?" mainly tests whether a model knows you. A buyer who doesn't know you yet asks something quite different: "What should I look for when choosing a supplier of quoting software?" That question contains no brand name, and the answer determines whom the buyer looks up next.
For sales, that has a direct consequence. A prompt set without brand names measures your chance of being discovered. A set with brand names only measures how well known you are among people who already know you.
B2B lead generation per buying stage: build the set
In each stage, a buyer asks a different question. So create a short series of prompts per stage, in the words of the persona who buys from you. The four stages below are a way of working, not a screen in GripIQ. We use quoting software as an example:
- Orientation: the buyer has a problem and is looking for an approach. "How do I shorten the lead time of our quoting process?"
- Comparison: the buyer knows the category and compares options. "How do quoting tools for B2B companies differ from each other?"
- Purchase decision: the buyer is looking for certainty. "What should I look for in the contract and pricing of quoting software?"
- Aftercare: the buyer is a customer and wants to know whether they made the right choice. "What support can I expect after implementation?"
Start with three or four prompts per stage. That gives you twelve to sixteen: enough to see patterns, and few enough to still read every answer yourself.
What does that deliver for your pipeline? For each stage, you see whether your brand is mentioned. If you are not in the answer in the orientation stage, the buyer only hears your name when someone else drops it. If you are in it in the comparison stage, but described differently from how you sell yourself, your team knows which picture it needs to correct in the first conversation. You read the answers themselves under Answers, with the brand mentions highlighted in the full text.
Let GripIQ suggest a first set
You don't have to start from a blank page. Under Generate smart prompts, GripIQ suggests 15 to 20 search questions based on the available insights. If a buyer persona and a market analysis have been created, they form the basis. You find them in the insights tab of AI Visibility.
The suggested questions are worded in the language of your project, contain no brand names and match the buying stages described above.
They don't go straight into your measurement either. Each question first appears as a Suggestion, in the Suggestions sub-tab next to Active and Archive. You review them and approve what fits your buyer. Questions already in your set are skipped.
Read the suggestions through the eyes of your best salesperson. Would a buyer really type this? Cut whatever sounds too general or too theoretical, and add the questions your customers ask in conversations. You type a prompt under New prompt, or import a list via CSV. Does your team already have a list of frequently asked questions from quoting conversations? Then you have your best source right there.
Set the country and group with tags
A buyer in Ghent phrases their question differently from a buyer in Rotterdam, and a model can give a different answer per market. That is why every prompt has a country: Belgium, the Netherlands, Germany, France, Luxembourg, the United Kingdom or the United States. You choose the country when you create the prompt.
If you sell in both Belgium and the Netherlands, add the same question for both countries to the set. The difference in answers shows whether your position differs per market. Useful for a sales rep who covers both markets, because a reference that convinces in one country may not do so in the other.
Tags let you group the prompts. You create them in the Tags tab and assign them to a prompt. For sales, these groups work well:
- a tag per buying stage, for example "Orientation" and "Comparison";
- a tag per segment or product, so each team sees its own figures;
- a tag per persona, if you serve more than one type of buyer.
Then filter the dashboard in AI Visibility on those tags. That way, you see Share of Voice, Rankings and Sentiment for one buying stage, instead of one average figure in which everything gets lost. In AI model comparison, you read for each prompt what ChatGPT, Gemini and Claude answer side by side.
Need more background? Why the orientation stage carries so much weight is explained in Lead generation starts before the first contact. And how to set the results against those of your competitors is covered in Competitor analysis: the model, an example and what AI answers add.
What you can do now
- Together with a colleague from sales, write down what your best customer asked before they got in touch. Divide those questions across the four stages.
- Let GripIQ suggest a first set under Generate smart prompts, and in Suggestions only approve what a buyer would really type. Add your own questions, with the right country and a tag per stage.
- Filter the dashboard in AI Visibility on one tag, and in AI model comparison read the answers from the stage where you are weakest.
Would you rather set this up together with us for your own market? Book a demo.