Multi-turn intent is the meaning a question takes from the conversation before it. Latent constraints are the conditions a buyer has in mind but does not state, such as budget, company size, location or tools they already use. AI assistants carry both into their searches, so a page that states who it is for and under what conditions it applies is easier to match to a real buyer’s situation.
How conversations change a question
In a traditional search box, each query stands alone. In a chat, the second question depends on the first:
- “What’s a good CRM for a small B2B team?”
- “Which of those work with Tally?”
- “And what would it cost for 20 users?”
The third question makes no sense alone. The assistant rewrites it using the earlier turns, into something like “cost of small-team B2B CRMs that integrate with Tally, for 20 users”, and searches for that. Google’s AI Mode, ChatGPT, Perplexity and Gemini all support follow-up questions in this way.
Some assistants also remember facts from earlier conversations or from saved preferences, such as a user’s industry or country. This can shape the searches they run even when the current question says nothing about it.
Latent constraints
Buyers rarely type everything that matters to them. Common unstated conditions include:
| Constraint | Example |
|---|---|
| Budget | Must fit a monthly spend the buyer has in mind |
| Size | Built for a team of 20, not 2,000 |
| Location | Available in India, prices in rupees, local support |
| Existing tools | Works with the accounting or CMS software already in use |
| Compliance | Meets data residency or industry rules |
| Skills | Usable without an in-house developer |
| Timeline | Can start or launch within a given period |
When an assistant knows or infers these conditions, it searches for sources that address them. A page that never mentions price, size, location or integrations gives it nothing to match.
Micro-personas (a Brilad working method)
We use the term micro-persona for a specific buyer situation defined by role plus constraints, for example “marketing head at a 50-person Indian SaaS company with no in-house SEO, needing results before a funding round”. It is our planning method rather than an industry standard. For each service, we list the micro-personas who buy it and check that the site answers their unstated questions somewhere a system can find.
How to write for it
- Say who a product or service is for, and who it is not for
- State prices, or at least price ranges and what changes them
- Name the tools and platforms it works with
- State where it is available and any location-specific details
- Answer the likely follow-up questions on the same page or link to them clearly
Each of these creates a passage that can match a follow-up question in a conversation, even if the first question was generic.
Frequently asked questions
Can I see the follow-up questions people ask about my topic?
Not from AI assistants directly. Sales calls, support emails, “People also ask” boxes and Search Console’s longer queries are the best sources.
Does stating who a service is not for lose customers?
It can reduce enquiries that would not have converted anyway. It also helps assistants recommend you to the buyers you actually suit.