Original post by Sergei Rogulin
30-second rundown
Key learnings:
- AI prompt research should focus on questions where buyers compare options and the system recommends specific brands.
- Detailed buyer constraints reveal the situations in which an AI system considers a product a good fit.
- A small, decision-focused prompt set is easier to monitor and act on than hundreds of generic questions.
This week: Choose one high-value service, write down one buyer’s budget, risk, context, and must-have constraints, convert those into 15 decision-stage questions, test them in two AI systems, and record which brands appear and why.
Coming 3 months: Use an AI search monitoring tool to track the approved decision-stage questions weekly, record which brands are recommended and why, and alert one marketer to meaningful gains or losses. Review a monthly summary to see which buyer constraints increasingly trigger your brand and which content gaps still leave it outside the shortlist.
Sergei Rogulin walks into the prompt-tracking carnival with a knife and cuts away most of the tent. His argument is brutally useful: track the questions where an AI system compares choices and recommends brands, not every harmless request for an explanation.
1. Recommendation Moments Matter Most
Prompt research identifies questions that make AI systems evaluate options and suggest specific brands. Broad informational prompts may generate plenty of answers but little buying influence, while decision-stage prompts expose whether your brand even enters the room when money is close to changing hands.
2. Constraints Force the Machine to Choose
A generic buyer description produces generic recommendations. Detailed constraints such as budget, experience, risk, location, compliance needs, or allergies push the system from explanation into comparison, turning a foggy question into a real decision with winners and losers. Compare the foggy “What is the best project-management tool?” with “What is the best project-management tool for a 20-person UK healthcare team, costing under £15 per user per month and meeting ISO 27001 requirements?”
3. Product Facts Must Match Buyer Problems
Rogulin recommends connecting concrete features to outcomes, use cases, risks removed, and signals of fit. Instead of merely listing “SAML single sign-on,” explain that it lets a healthcare IT team securely onboard employees through its existing identity provider, reducing password risk and showing the product fits a regulated organization. Those details must stay consistent beyond your own site because AI systems compare information across reviews, discussion sites, and third-party pages, a scavenger hunt with no respect for your brand guidelines.
Small teams can start with two or three priority offers, one or two sharply defined buyer types, and roughly twenty prompts tied to actual purchase pressure. That beats building a glorious warehouse of synthetic questions that nobody reviews after Tuesday.
