Answer-engine visibility

Your buyers are asking AI. Does it say your name?

More buying decisions now start with a question to an assistant instead of a search box. If the model doesn't name you, you never make the shortlist it hands over — and no analytics tool will show you that, because there was no click to miss.

Ask for an early scan Free while in early access. One reply from a person.

From our own measurement

Same 32 buying questions, asked three times, to two assistants. One named the brand half the time. The other named it once.

Three pinned runs of 32 questions per assistant, 96 answers each. Claude named the brand in 48 of 96 answers; ChatGPT named it in 1 of 96. Claude 48 of 96 answers Claude — named the brand in 48 of 96 answers (50.0%) ChatGPT 1 of 96 answers ChatGPT — named the brand in 1 of 96 answers (1.0%) 0 96 answers
One B2B brand, three repeat runs on a pinned question set, July 2026. The narrowest single-run gap was 46.9 points — that is the figure we stand behind, not the average. Your category will differ; this is the kind of split a scan is built to find.

How a scan works

Three steps, and you confirm the middle one.

A scan takes minutes, not a sales call. Nothing to install, no tag on your site, no access to your analytics — we only need the address a shopper would land on.

1

We read your storefront

The scan works out what you actually sell, who buys it, and which names you also go by — then shows you that profile and waits for you to correct it. Every number downstream depends on getting this right.

2

We ask the buyer's questions

A battery of buyer-shaped questions goes to ChatGPT and Claude — the kind your customers actually type. Not your brand name: nobody discovers you by searching for you.

3

We count who gets named

How often you appear, where in the answer, and which competitors take the slot when you don't. Then we re-run the same pinned questions, so next month's number means something.

The questions

Written the way a buyer types, not the way a marketer does.

Every question we ask is in the report, with the full answer underneath it. You can read exactly what the assistant said and check the count yourself.

Best leather weekender bags for the money — what would you pick?
Who actually makes good leather bags these days?
Which leather weekender is worth paying more for, if I want it to last?
Carry-on for a three-day work trip — what should I buy?

Those are real frames from the battery, filled in for a leather-goods brand. They are templated on purpose: identical across customers and identical across re-runs, so the same question is being asked every month.

The wording adapts to who is buying — a shopper choosing a bag and a VP choosing software do not phrase things the same way — and a handful of category-specific questions get generated on top, because a template cannot know your category's language.

What we never do is ask a question with your name already in it. That measures whether the model has heard of you, not whether it recommends you.

Why the number holds up

We measured our own repeatability before we sold anything.

The uncomfortable finding stayed in the product: some of what a single run tells you is noise, so the report marks it as noise instead of selling it to you.

Ask a model the same question twice and you can get two answers. Any tool that reports a single run as fact is showing you a coin flip with a decimal point on it.

So we ran the same battery repeatedly and checked what survived. The headline score is stable. Individual competitor names are not — a name can appear in one run and vanish in the next.

The report says which is which. Names we can't stand behind are marked as one-offs rather than quietly printed alongside the solid ones, and a month-over-month change smaller than the measurement's own resolution is reported as no change at all.

A tool that only shows you its best run is measuring itself, not your market.

Every figure in a Connexion report is read straight out of the run that produced it. None of them are typed by hand, and the report refuses to assert a competitor name the data doesn't support.

Pricing

What it will cost when the scan opens publicly.

No tiers, no seats, no annual contract. The audit is a fixed price and the tracking is month to month — cancel it the month it stops telling you anything new.

One-off audit

$99

The full battery across both assistants, every answer included as evidence, the competitors taking your slot, and what would move the number.

Tracking

$39 / month

The same pinned battery re-run every month, with changes reported only when they exceed what the measurement can actually resolve.

Early-access scans are free right now while we work with the first brands.

Get your number

Send us your storefront. We'll run it and send you what it says.

Ask for an early scan

No list, no sequence, no signup — one reply from a person.