ABOUT / PERSONASIGNAL

01

THE START

PersonaSignal startedwith a real operating question.

After years of working in ecommerce, a simple question emerged: when shoppers increasingly ask AI what to buy, how does AI decide which brands to recommend?

More shoppers are describing, comparing, and filtering products in natural language. What PersonaSignal studies is the observable recommendation behavior inside that process.

02

THE FIRST QUESTION

At first, I only wanted to know:
How does AI see a brand?

Early research started from real consumer scenarios and anonymized historical studies. The goal was never to document a career—it was to understand a question that kept coming up in operations.

  • 01

    Does AI know it?

  • 02

    How would AI describe it?

  • 03

    Would AI mention it?

  • 04

    What product information is worth adding?

But gradually, something became clear:

Studying how AI sees a brand
alone didn't seem to explain recommendations.

Shoppers don't always name the brand first. More often, they say:

  • “What my situation is”
  • “What I'm trying to solve”
  • “What my budget is”
  • “What I can't accept”
  • “Which fits me better”
03

A HUNCH

A PERSONAL OBSERVATION — THE START OF A RESEARCH HYPOTHESIS

Over time, I started to feel something strongly:

I started to get this strange feeling
that AI was somehow “labeling” each of us.

Sometimes I would compare how friends use AI. We could ask very similar questions and still get different product recommendations.

I started to wonder whether, through long conversations, repeated questions, and expressed preferences, AI had formed a rough sense of who each person was.

BOUNDARYI don't know whether that's literally how the model works internally. What mattered was that the observation changed the question I wanted to test.

INPUTSimilar questions
CONTEXTDifferent people, different situations
OBSERVATIONDifferent product recommendations
04

REFRAMING

To understand AI recommendations,you can't just look at the brand.You have to look at who's asking.

If different people, needs, situations, and constraints produce different recommendations, then a useful experiment can't only ask:

“How does AI see my brand?”

It also has to ask:

For different people
in different situations,
what would AI recommend?
WHO
Different people
NEED
Different needs
CONTEXT
Different situations
CONSTRAINT
Different constraints
05

THE PERSONA TURN

So I started simulating different shoppers.

I started small, with a handful of Personas. Then the test design expanded into dozens of scenarios, then around a hundred, and eventually hundreds of structured model runs. The point isn't the volume. It's making “who's asking” something that can be designed, compared, and re-checked.

STARTA handful of Personas

Describe one concrete shopping problem completely.

EXPANDDozens, then around 100 scenarios

Systematically vary person, task, preferences, and constraints.

RUNHundreds of structured runs

Compare conditions, models, and observable recommendation states.

  • Life stage
  • Use context
  • Purchase goal
  • Budget
  • Preferences
  • Constraints
  • Risk tolerance
  • Must-have criteria
  1. 01

    PERSON

    Person

    Who is deciding
  2. 02

    CONTEXT / TASK

    Context / Task

    What they are doing
  3. 03

    FULL NEED

    Full Need

    What they actually need
  4. 04

    SCREENING CRITERIA

    Screening Criteria

    How options are compared and ruled out
  5. 05

    PRODUCT MATCHING

    Product Matching

    Which options meet the criteria
  6. 06

    AI RECOMMENDATION

    AI Recommendation

    What appears in the end

TEST-DESIGN FRAMEWORKWe're not claiming AI internally follows these six steps. We use this structure to design external, repeatable recommendation tests.

06

THE NAME

The name PersonaSignal came from this same idea.

First, look at the need and context a Persona expresses. Then observe how AI maps those signals to brands and products.

PERSONA

Who is asking

SIGNAL

What need and context they express

OBSERVE

How AI maps it to brands and products

AT FIRST

“How does AI see a brand?”

LATER

“For different people, in different situations, what does AI put into the recommendation?”

07

HOW WE WORK TODAY

Today, PersonaSignal turns that question
into structured research.

Our research capabilities can cover different directions, but each project only uses the conditions that are relevant to the question and can be clearly defined and re-checked. Capabilities can be broad; public conclusions must be precise.

  • 01

    Recommendation Discovery

    Blind / Natural
  • 02

    Brand Understanding

    Brand-informed
  • 03

    Product Fit

    Product / shopper fit
  • 04

    Competitor Substitution

    Substitution
  • 05

    Repeated Runs

    Variation observed
  • 06

    Model Comparison

    GPT / Gemini
  • 07

    Search / Grounding

    Search / Grounding
  • 08

    Source Observation

    Source Review

FOUR RESEARCH PRINCIPLES

Turning an intuition into research that can be checked.

  1. 01

    Design before scale

    Define the Persona, scenario, need, and test conditions first, then scale the number of runs.

  2. 02

    Don't rely on one answer

    A single prompt can be sensitive to wording and randomness. What matters is what happens across repeated runs, contexts, conditions, and models.

  3. 03

    Distinguish recommendation states

    Mention ≠ Recommendation ≠ First Choice. Fit isn't a recommendation rate.

  4. 04

    Review public findings against evidence

    Automation scales the experiments. Evidence mapping, interpretation, boundaries, and public conclusions require Human Review.

View the research method
08

POSITIONING

We're clear about what we do,
and what stays outside.

IS

PersonaSignal is

  • Structured AI recommendation research
  • A research product
  • A professional diagnostic service
  • Evidence-aware
  • Persona-first
IS NOT

PersonaSignal is not

  • One-prompt verdict
  • Guaranteed ranking service
  • SEO / GEO agency
  • Hidden-reasoning decoder
  • Sales-growth guarantee
  • Mature self-serve SaaS
09

FOUNDING STAGE

FOUNDING / CUSTOMER VALIDATION

PersonaSignal is still early.
But it didn't start with a pitch deck.

It started with real operating questions, real experiments, and real data. What we're doing now is turning the methods, evidence management, and delivery process we've already used into a more stable research product.

RIGHT NOW, THE PRIORITIES ARE:

  • Continue running real research
  • Improve test design
  • Strengthen evidence traceability
  • Work closely with early brands on high-value research questions

FOUNDING PILOT

If you're asking the same question:

When shoppers ask with their real needs in mind, how does AI understand, filter, and choose your brand?