PERSONA-FIRST AI RECOMMENDATION RESEARCH

AI knows your brand. That doesn't mean AI will recommend it.

PersonaSignal tests what happens when different shoppers ask ChatGPT and Gemini for help with real purchase decisions: whether your brand gets recommended, which brands show up instead, and how different needs and contexts change the outcome.

You're not just asking, “Does AI know my brand?” You're asking, “When does AI actually recommend it?”

From “Why wasn't my brand recommended?” to “What should we improve next?”

Apply for Founding Pilot View Sample Report

Founding / Customer Validation

RECOMMENDATION SNAPSHOT

Purchase-context recommendation

ILLUSTRATIVE EXAMPLE

SHOPPER

42 · Denver

Dry, sensitive skin

NEED

A fragrance-free daily face cream for a dry climate

AI RECOMMENDATION

MODEL OUTPUT · SAMPLE

01Brand ARecommended
02Brand BRecommended
03Brand CRecommended
—Your BrandNot Selected

QUESTIONS WORTH INVESTIGATING

Is the product easy to identify?
Does AI understand the positioning?
What still needs to be clarified or tested?

Illustrative example · Not a customer result · Does not establish causation

THE QUESTIONS YOU'RE ASKING

These are the questionsPersonaSignal is built to investigate.

We design every project around questions like these — and answer them with structured evidence.

01
ChatGPT knows my brand. Why doesn't it recommend it?
02
Which competitors does AI recommend instead?
03
What kinds of shopper needs make my product more likely to be recommended?
04
Is the issue brand recognition, product information, or shopper fit?

PERSONA FIRST

Same purchase question.Different shopper.Different recommendation.

Many AI visibility checks start with the brand. PersonaSignal goes one step further: it also tests who is asking.

Whether your brand gets recommended can depend on more than the brand itself. We systematically vary these conditions and observe which brands get recommended, which alternatives appear instead, and how the outcome changes.

Who's asking?
What situation are they in?
What are they trying to solve?
What constraints matter?

METHOD NOTE

This is a test-design framework — not a description of hidden model reasoning.

Explore the method

RECOGNITION ≠ RECOMMENDATION

Brand recognition isn'tthe same as being recommended.

Asking, “Do you know Brand X?” mainly tells you whether AI can recognize or describe the brand.

PersonaSignal asks a different question: when a real shopper asks a specific purchase question, does the brand actually get recommended?

Being understood by AI and being recommended in a purchase context are different questions.

OBSERVABLE STATE 01

AI KNOWS YOUR BRAND

01Brand recognized
02Product understood
03Website discovered

OBSERVABLE STATE 02

AI RECOMMENDS YOUR BRAND

01Shopper asks a concrete purchase question
02AI compares available options

Your BrandORAlternative?

PersonaSignal studies the difference between those two observable states.

HOW IT WORKS

From one business questionto a structured diagnosis.

Three steps — from shopper contexts to what to improve next.

01
BUILD SHOPPER CONTEXTS
Define who's asking
Different people, needs, budgets, use cases, and constraints.
02
RUN STRUCTURED AI TESTS
Structured purchase questions
Test purchase scenarios across ChatGPT and Gemini, then record brand recommendations, alternatives, and how results change across conditions.
03
FIND THE RECOMMENDATION PATTERNS
Answer the five key questions
Who gets recommended? Who doesn't? What changes the outcome? Which competitors keep appearing? What should you investigate or improve first?

You get a written diagnosis — not a one-off AI screenshot.

PUBLIC EVIDENCE

100 shopper scenarios × 3 historical models= 300 model runs

AI can understand a brand in context without recommending it in natural purchase scenarios.

This is PersonaSignal's own research — it doesn't describe the market as a whole.

RESEARCH RECORD

PERSONASIGNAL RESEARCH

FOUNDING PHASE

MODELS: GPT
Gemini
TEST DESIGN: 100 shopper scenarios
× 3 historical models
RECOMMENDATION-LIST INCLUSION: 0 / 300
FIRST CHOICE: 0 / 300
GEMINI SAFETY-NOTE MENTIONS: 4 / 300
Mentions only
REVIEW: Human review

Research before optimization · Evidence before claims

Read the full Research Note

DIAGNOSTIC VALUE

The real value isknowing what to improve next.

PersonaSignal doesn't just show you who AI recommended. It helps you identify what may be worth investigating and improving next.

Find the signals shaping recommendation outcomes, so you can prioritize changes that give your brand a better chance of being recommended.

Is your brand information clear enough? Worth checking first
Is your product easy for AI to identify and distinguish? Worth verifying
Which shopper needs are the strongest fit? Worth testing
Which use cases are underrepresented? Possible gap
Which competitors are repeatedly recommended instead? Worth investigating
Does Search / Grounding change the answer or the sources? Next research step

WHAT YOU RECEIVE

What will the researchactually tell you?

You get a readable, traceable written diagnosis covering:

whether AI naturally mentions or recommends your brand
which shopper contexts change the outcome
which competitors are recommended instead
how AI describes your product and positioning
whether Search / Grounding changes the answer or sources
which gaps are worth addressing or investigating further

PersonaSignal

AI RECOMMENDATION DIAGNOSTIC

Brand recommendation research report

ILLUSTRATIVE SAMPLE

NOT A CUSTOMER CASE

SECTION 02

PERSONA MATRIX

ILLUSTRATIVE DATA

Compare brand entry, mention, and selection across defined shopper contexts.

Persona	Need	Your Brand	Main Alternative
P01	Need A	Recommended	—
P02	Need B	Mentioned	Brand X
P03	Need C	Not Selected	Brand Y
P04	Need D	Unknown / Insufficient Evidence	Brand Z

See the path from conclusion back to the supporting research record.

View Sample Report

RESEARCH BOUNDARIES

What these results can tell you— and what they can't.

The findings are only meaningful within the conditions we tested.

What they can show: What AI actually recommended under structured test conditions, and which conditions changed the outcome.
What they can't: They don't describe the market as a whole, and they don't prove causation.
Examples and snapshots: All report previews on this page are illustrative — not customer cases.
Research scale: Test volume varies by project and is confirmed before research begins.

FOUNDING PILOT

Want to know why your brand isn't being recommended —and what to improve first?

Tell us about your brand, your priority market, and the AI recommendation question you most want answered. Every application is reviewed manually, and scope and deliverables are confirmed before research begins.

Quick Scan

Focused scope

A compact first look at one clearly defined AI recommendation question.

Designed to help you decide whether a brand or product question warrants deeper research.

Gemini Deep Audit

Deep audit

FOUNDING PILOT

A fuller persona-first study covering recommendation discovery, product fit, alternatives, sources, and a written diagnosis.

Dual-model / Custom Evaluation

Custom scope

For projects requiring model comparison, broader test coverage, or a tailored research design.

ApplicationManual QualificationScope ConfirmationPayment / PilotResearchWritten Delivery