The Three-Answer Test: How to Measure Founder Visibility in AI
Last updated: August 27, 2026
Maria Dykstra diagnoses why B2B companies are invisible to AI systems.
She is the creator of the Algorithmic Authority Stack™ and the Algorithmic Authority Index™: two waves of original research across 400+ assessed companies on ChatGPT, Perplexity, and Google’s Gemini.
The Three-Answer Test is a founder visibility measurement method: one buyer-intent question, asked on OpenAI’s ChatGPT, Perplexity, and Google’s Gemini, with the three answers compared against each other. It measures whether AI recommends you when a buyer has intent, instead of whether AI recognizes your name.
The one query they cannot lose.
Recognition.
The category.
“Best [what you do] for [who they are].”
Whether you are on the shortlist AI hands your buyer.
Five minutes. Zero cost. The first real baseline most founders will ever have.
You searched your own name in ChatGPT. It worked. AI knew exactly who you were.
That result feels like proof. It measures the wrong thing.
Your buyer never types your name. They type the problem. The question that decides revenue comes one step earlier: does AI introduce you when a buyer asks who to hire?
After 50+ B2B audits, plus the 420 structured assessments of Wave 1 of the Algorithmic Authority Index, the pattern is stable. Founders test recognition when they should test recommendation. Recognition means the machine has met you. Recommendation means the machine introduces you. Only the second one produces pipeline.
This post gives you the test that measures the second one. Five minutes, one question, three answers.
What is the Three-Answer Test?
Ask one buyer-shaped question on three AI platforms. Compare the three answers against each other. The comparison is the measurement.
The steps:
- Write the question your buyer asks before they know you exist. The problem plus the buyer: “best [what you do] for [who you serve].”
- Ask it, unedited, in OpenAI’s ChatGPT, Perplexity, and Google’s Gemini. Fresh sessions. No account history steering the answer.
- Save all three answers. Screenshot or paste them side by side.
- Read them as a set. Who appears in all three? Who appears in one? Where are you?
What it produces: your first real baseline on founder visibility. That alone puts you ahead of every company I have measured. In Wave 1, 0% of audited companies tracked their own AI citation frequency.
How to run these prompts (so you see what a buyer sees)
Your own ChatGPT account remembers what you have searched. If you have looked up your brand before, it will surface your brand for you. That is your history talking, not the model’s default answer. A buyer who has never heard of you starts with none of it.
To see what a buyer sees, strip out your history:
- Log out of ChatGPT, Perplexity, and Google.
- Open a private or incognito window.
- Ask one prompt. Read the answer. Close the window.
- Open a fresh private window for the next prompt. Do not ask follow-up questions in the same chat. Each follow-up teaches the model what you want to hear.
- Run the same prompt on ChatGPT, then Perplexity, then Google’s AI Mode. They pull from different sources and often disagree.
A warm account will look fine. A cold session shows how often your brand appears on its own, and how often a competitor answers instead. The gap between those two is the finding.
Why do I need three platforms instead of one?
Because the platforms disagree. In Wave 1, the three platforms agreed on a company’s authority 69% of the time. Nearly one answer in three contradicts the other two.
The disagreement is systematic. In my Wave 1 testing, three retrieval behaviors kept repeating. Platforms update constantly, so treat these as observed tendencies, and expect them to evolve:
OpenAI’s ChatGPT tends to behave like a cautious advisor. It lists peers without ranking them and prefers frameworks over vendor claims. Appearing here suggests the model can classify you.
Perplexity tends to behave like a transparent auditor. It shows citations, surfaces reviews, and filters hardest on third-party corroboration. Appearing here suggests independent sources back you up.
Google’s Gemini tends to behave like a confident recommender. It assigns authority language readily, drawing on Google’s index and knowledge graph. Appearing here suggests your category signal is strong enough to trigger a recommendation.
One answer tells you what one platform thinks. Three answers tell you whether your visibility is structural or accidental. A founder visible on one platform and absent from two has what the Algorithmic Authority Stack calls Surface Dependency. It looks like coverage on a dashboard. It behaves like invisibility everywhere else the buyer asks.
What question should I ask?
The question your highest-value buyer asks in the week before their shortlist forms. Build it from the outcome they want, the category they would use, and the qualifier that describes them.
Examples of the shape:
- “Who are the best fractional CMOs for B2B SaaS companies under $10M?”
- “Best compliance consultants for medical device startups”
- “Who should a fintech founder hire to fix positioning?”
Use the buyer’s wording, including the sloppiness. Buyers describe pain in colloquial terms. If your visibility depends on a category term only you use, this test will show you what that costs.
The trap to avoid: retreating to the name query when the buyer query stings. “Who is [your name]” measures recognition, and recognition is table stakes. Passing it proves the machine has met you. It says nothing about whether the machine would ever introduce you.
This distinction decides revenue. 92% of B2B buyers have a shortlist before the formal purchase process starts (Omniscient Digital). The shortlist increasingly forms inside the three answers you are about to read.
How do I read the three answers?
Score each platform on three questions: Named? Category correct? Positioning correct? Then match your pattern to one of five outcomes.
Named? Yes / No
Category correct? Yes / No
Positioning correct? Yes / No
Presence: how many of three answers name you
Accuracy: whether the descriptions match reality
Consistency: whether the three answers agree
Five outcomes, each mapping to a different structural condition:
| Your result | What it means | Strategic priority |
|---|---|---|
| Named in all three | You hold Share of Model in your category: the percentage of AI answers that name you. Rare. In Wave 2, 73% to 86% of tested vendors across four industries received zero citations. | Defense. Record the answers, retest monthly, watch for drift. |
| Named in one, absent from two | Surface Dependency. One platform found a signal the others cannot verify. | Diagnose which trust surface carried you there. Build corroboration on the other two. |
| Absent from all three, competitors named | Category Absence: functionally invisible at the moment of shortlist. This loss produces no signal in your funnel. No lost-deal call. Nothing to tag in the CRM. | Entity and structure fixes, starting at Layer 1. More content will not fix this. |
| Grouped with the wrong companies | The model has no peer set for you, so it borrows an adjacent category. In Wave 2, roughly two in three returned vendors operated in a category adjacent to the one queried. | Sharpen the peer set. One category sentence, repeated on every surface. |
| Named, but described wrong | Outdated signals win. Rebranded companies get cited by their former names in 81% of rebrand queries, years after the change. | Update the surfaces machines actually read. The rebrand fix sequence is here. |
A note on the fourth outcome, because it flatters before it hurts. A solo consultant grouped with global firms is hearing a confession: the model cannot place them.
Why am I invisible when my expertise is real?
Because the machine cannot verify what it cannot trace. In my research, third-party sources consistently outweighed owned content when AI systems evaluated credibility.
One founder I audited had 54 LinkedIn articles, an active blog, and genuine expertise. The buyer query named three competitors and never mentioned her. The content was not the problem. The structure was: no author schema, no Person entity, a LinkedIn bio that disagreed with her website.
Six weeks after the structural fixes, the same query named her alongside two competitors and cited her work. The content never changed. The full expertise architecture fix is documented here.
That is the difference between publishing and being provable. Publishing more on a broken structure compounds the problem: each new post adds a classification decision the model gets to make about a company it cannot place. That mechanism is the Volume Paradox, and it is why the fix rarely starts with content. If your content ranks in Google and still never appears in AI answers, the extraction fix is here.
What do I do with my result?
Record it, date it, and rerun it before you trust it. A single run is a directional read. Three to five runs over a few days is a baseline.
AI answers drift between sessions. Model updates, retrieval changes, and randomness all move individual results. The pattern across several runs is the signal. Then retest monthly with the same question, because the platforms never stop changing underneath you.
Two honest limitations. The test is a snapshot of platform behavior, never a permanent ranking; the next model update can move any result. And in regulated categories, expect refusals. In my healthcare research, clinical queries were often refused outright. A refusal is data too. It tells you which evidence surfaces the platform demands before it will answer at all.
The Sequence
Test → Snapshot → Audit → Engine.
The Three-Answer Test tells you whether the problem exists. The Snapshot names the failing layer. The Audit maps all seven and sequences the fix. The Engine builds the durable evidence on top of a resolved foundation. Skipping the diagnosis and buying tactics is the fastest way to spend twelve months optimizing a company AI still cannot classify.
If your three answers came back clean, protect the position. If they came back the way they do for most founders, the question changes. Whether you have a visibility problem is settled. Which layer is producing it becomes the work. That diagnosis is the next post in this series: what an AI visibility audit actually reveals, and what comes next.
You now know how the machine answers when your buyer asks. The founders in those three answers know something you are about to learn.
AI Visibility Snapshot
The formal version of the test you just ran.
The Snapshot runs your company against the same 7-layer framework used across 400+ assessed companies in the Index research. Written diagnostic in 48 hours: the layer breaking first, the structural reason, and the competitors holding the slots you should hold.
$497. No discovery call. The price credits toward whichever next step you take.
FAQ
What is the Three-Answer Test?
A founder visibility measurement method. One buyer-intent question, asked on OpenAI’s ChatGPT, Perplexity, and Google’s Gemini, with the three answers compared against each other.
It measures recommendation, the moment AI names you on a shortlist, rather than recognition.
Why test three AI platforms instead of one?
The platforms disagree systematically. In Wave 1 of the Algorithmic Authority Index, the three platforms agreed on a company’s authority 69% of the time.
Three answers reveal whether visibility is structural or accidental. One answer only reveals what one platform thinks.
What question should I ask?
The question your buyer asks before they know your name: the outcome they want, the category they would use, and a qualifier that describes them.
A name query only measures recognition. Your buyer already knows who you are before typing your name, so passing it proves nothing about recommendation.
How often should I run it?
Run it 3 to 5 times over a few days for a reliable baseline, because answers drift between sessions.
Then retest monthly with the same question. Record and date every result. That log is Layer 7 of the Stack: Visibility Measurement.
Related reading:
- The Algorithmic Authority Stack: 7 Layers Between You and AI Visibility
- The Algorithmic Authority Index: Waves 1 and 2
- Why Doesn’t AI Recognize My Expertise? The Expertise Architecture Fix
- The Two-Loop Problem: Why AI Still Trusts Your Old Identity After a Rebrand
- The 43,000:1 Problem: Why Volume Makes AI Invisibility Worse
- The AI Visibility Snapshot
Similar Posts
What Does an AI Visibility Architect Actually Do?
B2B buyers now build shortlists inside AI systems before they visit your site or book a call. Nobody on your existing team has structural authority over how machines resolve, classify, and retrieve you. That is the gap the AI Visibility Architect fills.
What Does E-E-A-T Look Like for AI Models vs Google?
Google’s E-E-A-T evaluates individual pages. AI models evaluate entities across the entire web. That difference changes everything. Search trust = page-level signals. AI trust = cross-platform entity consistency. Your existing E-E-A-T investment is the foundation. It is not the destination. Learn all of the key differences.
Layer 2: Expertise Architecture — How AI Decides If You Know What You Claim
Last updated: June 26, 2026 Last updated: April 2026 You claim expertise on your about page. AI checks if your content proves it. Claims without evidence register as noise. Your company spent years building genuine expertise. You hired the right people. Solved hard problems. You have 15 years of client outcomes, a methodology that works,…
AI visibility for PR agencies: which sources should you target?
Before you build the next media list, ask AI a question your client’s buyers would ask. Open the sources behind the answer, then check who wrote them and who controls them.
Does Reddit Still Matter for B2B AI Visibility: The Citation Strategy Most Brands Get Wrong
Reddit used to be the most-cited source across major AI platforms. But then the citations dropped suddenly on August 14, 2026. This blog explains what you need to know now, whether Reddit still matters, how to check on where your citations sources are.
Why Your AI Citations Dropped (And How to Get Them Back)
Three different events get reported as a citation drop. Rotation happens daily and means nothing. Removal is close to permanent. The diagnostic that separates them takes twenty minutes.
