Choose the prompts to track for your business
Choose prompts that match your buyers using Erlin's suggestions, estimated demand and buying intent, then track them and choose where checks run.
Erlin's Prompt Discovery recommends questions for your business to track in AI answers. It combines partner data with Erlin's algorithm to identify relevant prompts, estimate demand and assess buying intent.
Use this guide to choose questions that matter to your customers, choose the language and run locations for your checks, and review your first results.
1. Start with a buying decision
Choose a product or service you want to understand. Then consider what someone needs to know before buying it.
For example, a website design business might want to know which agencies AI recommends for an ecommerce redesign.
Keep this decision in mind as you review suggestions. A question is useful to track when its answer could influence whether someone considers your business.
2. Review the suggested prompts
Open Prompt Discovery and review the suggestions for your business.
Look for questions about the problems you solve, the alternatives buyers compare and the requirements that affect their choice.
The following prompts illustrate different parts of a website buyer's decision. They are examples, not measured customer queries.
| What the buyer is deciding | Example prompt |
|---|---|
| What kind of help they need | What should I look for in an ecommerce website redesign agency? |
| Which approach to choose | Should I hire a freelancer or an agency to redesign my online store? |
| Which businesses to consider | Which agencies offer ecommerce website redesigns in Melbourne? |
Choose prompts that reflect services you actually offer. A popular question about an unrelated service will not help you understand visibility for your target customers.
3. Interpret volume and buying intent
Erlin estimates prompt volume using its algorithm and volumes supplied by data partners. The displayed volume is an estimate of demand, not an exact count of every conversation across AI platforms.
Buying intent describes how closely a question relates to a purchase decision. Someone asking which provider meets a specific requirement is making a different decision from someone learning what a service does.
Use volume and intent together. A lower-volume question can still matter when it closely matches your service and the customers you want to reach.
4. Choose the language and where checks run
The language picker in Discovery, Language for suggested prompts, sets the language Erlin writes suggestions in. Erlin offers 12 languages and saves your choice for the brand. It applies to the suggestions, not to each check.
Checks run from your brand's office locations. With two or more locations, Run prompts from lets you choose which ones. Pick locations that match the market you serve.
When comparing results over time, keep the prompt and the run locations consistent. If you change them, account for the change when interpreting the results.
5. Track your selected prompts and review the answers
Add your chosen prompts to tracking and run the checks.
Prompt Discovery estimates which questions are worth tracking. A tracked run collects AI answers to the questions you selected. Keep those two kinds of information separate when reviewing the results.
Read the answers alongside the metrics. Check whether your business appears, which alternatives are mentioned and which sources are cited where that information is available.
Once each prompt has run at least 4 times on different days in the last 90 days and you have picked competitors, Erlin suggests tasks. Until then, Find new tasks is off and says, for example, "Run your prompts at least 4 times first (3 of 12 ready)." Tasks are also found automatically after a scheduled run or a single-prompt run, and finding them costs no credits. Review the evidence behind a recommendation before deciding what to change.
Continue reviewing the same questions to understand how answers change. A result for one prompt and run location describes that check; it does not represent every buyer's AI conversation.