Personas & Context

QueryBurst Answer Engine uses persistent personas to represent the background context of the person using AI. A persona is not a mega-prompt. It is a stable profile that exists outside the current message and shapes how the system researches and answers every turn.

What is a persona in Answer Engine?

A persona represents the relatively stable background context of the person using AI.

That might include:

  • profession or role
  • company or industry
  • location
  • broad goals
  • persistent preferences
  • constraints
  • relevant history
  • known circumstances

That context exists separately from the current user message.

So instead of asking:

"I am a 45-year-old marketing manager in New York running a six-person team with a $3,000 monthly budget. Which CRM should I buy?"

the message can simply be:

"Which CRM should I buy?"

Because the system already knows who is asking.

Should a persona contain everything you know about the buyer?

No.

In fact, a useful persona should generally be a little underbaked.

A persona should describe who the person is and the relatively stable context they bring into a conversation.

It should not contain the entire future buyer journey.

Things such as:

  • which products have already been recommended
  • what the user has ruled out
  • which objection appeared three turns later
  • whether pricing suddenly became important
  • whether cancellation terms became a concern
  • which product is currently leading the shortlist

belong to the conversation, not the persona.

A simple rule:

The persona tells us who the buyer is. The journey tells us what has happened to them.

Why is an overbaked persona a problem?

Because you start encoding journey state as persona state.

Imagine a marathon runner persona that already specifies:

  • injury history
  • current shoe rotation
  • race goal
  • preferred stack height
  • attitude toward carbon plates
  • durability requirements
  • budget
  • the fact they already own a daily trainer

Now change one of those situational details.

Do you need another persona?

What about two details?

Or five?

Very quickly you need a separate persona for every possible variation of the buyer's state.

That is exactly the job the conversation and its branches should be doing.

If you need a new persona every time the context changes, your persona is doing the job of the journey.

Doesn't adding personas to a prompt tracker solve the context problem?

Only partly.

It solves one component:

who is asking?

It does not solve:

  • what they have already asked
  • what the AI previously told them
  • what products have entered the conversation
  • what they have rejected
  • which objections have emerged
  • which criteria have become important
  • which path through the decision they are currently following

Those are all forms of conversation state.

Persona context and conversation context are not the same thing. QueryBurst’s definition of AI visibility treats both as part of the problem — visibility is conditional on who is asking and what has happened in the conversation so far.

You need both.

How is Answer Engine different from "persona stacking"?

Some prompt trackers now allow a persona description to be injected around every tracked prompt.

That improves on testing completely anonymous prompts, but the basic architecture is still:

static persona block + isolated prompt → answer

Every prompt remains an independent event.

Answer Engine instead models:

persistent persona + accumulated conversation state + branch history + current message → answer

The distinction is important.

Adding a persona to every prompt gives each question a richer starting condition.

It does not turn those questions into a conversation.

Why not simply inject all relevant context into every prompt?

Because that assumes you already know which context should exist at that particular point in the journey.

It also makes every piece of context simultaneously present on every run.

That is not how a real conversation develops.

A user may reveal something on turn three.

A product might enter the shortlist on turn four.

A cancellation concern might not arise until turn seven.

Those details should affect subsequent turns because they happened, not because somebody anticipated the entire journey and stuffed it into the first prompt.

Better prompt construction can improve an isolated prompt.

It still does not turn it into a conversation.