AEVA · Answer Engine Visibility Audit

See what your customers' AI assistants actually tell them about you

AEVA simulates your real customers, human and AI agent alike, and asks their real questions across every major answer engine. You get a ranked, graded visibility map showing which businesses each persona actually sees, and where yours lands.

Start your first audit See how it works

Pay with tokens. No subscription. Your first audit is on us.

Live across OpenAI Anthropic Google Gemini

Query

3 engines · 3 samples

“Who is the best employment lawyer in Chicago?”

Law student · Chicago · EN

Harlan & Reed

HR manager · Miami · ES

Brightwater Legal

Procurement agent · JSON out

Lakeshore Law Group

Illustrative data for demonstration. Real reports are generated from live engine responses and cite every source.

The shift

Ten blue links became one answer. Nobody audits that answer.

When a buyer asks an assistant "who is the best X in Y?", they get a single synthesized recommendation — not a page of options. If you are not in it, you were never considered.

Answers shift with who is asking

The same question returns different businesses depending on language, geography, phrasing, persona context and which engine is asked. There is no single "ranking" to check.

A growing share of askers are not human

Shopping agents, procurement bots and travel planners now make shortlists on behalf of people. Almost no one measures visibility for that audience.

Generic prompts hide the truth

Most tools run persona-less prompts and report "does my brand appear?". That average conceals the segments where you are completely invisible.

The core insight

Visibility is not a number. It is a matrix.

persona × engine × query × time

Averages are a commodity. Cross-sections are the product.

Competitors track brand mentions at the prompt level and hand you one aggregate score. AEVA makes persona simulation the core primitive: demographic, geographic, linguistic and behavioral context is injected before every query, and AI agents are treated as a first-class customer type.

Every attribute — gender, age, role, language, geography, human or agent, engine, model — becomes a filterable dimension. You slice average positions like a pivot table.

How it works

Five steps from a seed question to a visibility cube

Step 01
Build your ICPs

Define ideal customer profiles as human personas (geography, language, age, role, budget, intent stage, query style) or agent personas (purpose, optimization objective, tool access, guardrails, autonomy level). Start from templates or build your own. Each one gets a face, so your team recognizes them at a glance.
Step 02
Expand the query set

You supply seed questions. AEVA expands them into a query matrix: translations, persona-natural paraphrases, funnel-stage variants and follow-up chains, all tagged by topic cluster and intent.
Step 03
Run every engine

Each persona context is injected as a system prompt and the query is sent, with live web search enabled, to OpenAI, Anthropic and Google Gemini. Each cell is run several times, because answers are stochastic; what comes back is a probability.
Step 04
Grade every answer

A pipeline of deterministic extraction plus LLM-as-judge pulls out every business mentioned, its order of appearance, sentiment and framing, factual accuracy about your brand, the sources cited, and exactly which competitor took the slot.
Step 05
Slice the cube

Everything lands in a multidimensional cube — position × persona attributes × engine × language × time — that you explore in the dashboard and export as CSV.

The filtering engine

One question. Twelve personas. Four very different answers.

An illustrative cohort asking one question, "Who is the best employment lawyer in Chicago?", across three engines. Change the filters and watch the leaderboard reorder. This is what a real report does with your data.

Query

“Who is the best employment lawyer in Chicago?”

Cohort

12 personas · 3 engines · 3 samples = 108 graded answers

Audience

Gender

Engine

Showing of personas

Business	Avg. position	Mention rate

What this slice says

Illustrative data for demonstration. Real reports are generated from live engine responses and cite every source.

Agent visibility

Half your future buyers will never see your website

They will send an agent. It reads your structured data, compares it against machine-readable competitors, and returns one shortlist to the human who asked. If your prices, credentials, availability and service definitions are not parseable, you are not on it.

AEVA audits that audience directly: agent personas with explicit optimization objectives, tool access, output formats and guardrails. The result is your Agent Gap — the distance between how machines rank you and how the people they serve rank you.

Agent Pick Rate

How often an agent selects you, not merely mentions you.

Agent Gap

Your average position for agents minus the same demographic's humans.

Agent-readiness

How well your web presence performs when the evaluator is a machine.

What you get

The numbers in every report

AEV Score

Composite 0–100 visibility score per brand, per persona segment.

Persona Coverage

Share of your defined ICPs whose answers include you at all.

Answer Stability

How consistently you survive repeated runs of the same question.

Share of Voice

Your presence against named competitors inside any slice.

Blind-Spot Segments

Filter slices where your mention rate collapses below threshold.

Hallucination Rate

False claims made about your brand, per hundred answers.

Citation Graph

Which sources the engines leaned on to pick the winner.

Position Drift

How your placement moves over time as engines and content change.

Pricing

Pay for what you actually run

Top up your balance with tokens and spend them per cell. A cell is one persona asking one query on one engine, once, grading included. Engines are priced individually, because a grounded answer costs us more on some than on others. No subscription, no seats, nothing expires.

personas × queries × engines × samples = cells

Starter

$30

5,000 tokens

one-off top-up

All three engines
Unlimited ICPs and saved audits
Full filtering cube and CSV export
Raw answers and citations retained

Create an account

Growth

Scale

$395

100,000 tokens

one-off top-up

All three engines
Unlimited ICPs and saved audits
Full filtering cube and CSV export
Raw answers and citations retained

Create an account

Every new account starts with a free token grant — enough for a first small audit before you spend anything.

Methodology

What this measures — and what it does not

Persona-conditioned probing, not impersonation

A real user has memory, location signals and an app-layer system prompt we cannot reproduce through an API. AEVA approximates that context deliberately and transparently: it is a directional signal about how engines respond to a customer type, not a recording of one person's screen.

Distributions, not screenshots

Every cell runs several times, so what you see is how reliably you appear across repeated asks rather than one lucky draw.

Official APIs only

We query documented, paid APIs with web search enabled. We do not scrape consumer apps. Persona simulation is framed and used as market research.

Your raw data is kept

Every raw response and citation is stored immutably against the run, so any number in a report can be traced back to the answer that produced it.

Find out who your customers are told to hire instead of you

Create an account, build your first ICP in two minutes, and run an audit on your free tokens.

Start free Talk to us first