Agentic AEO is live on the Shopify App Store

Attribution

A sensor in your storefront, not a guess.

Burnish runs a first-party pixel on your store, ties assistant visits to Shopify orders, and shows its working.

Run a free auditHow the score works
Representative demo · Threshold Footwear
Representative demo · Threshold Footwear

Five fields. Zero people.

The whole of what the pixel keeps, beside the whole of what it refuses to.

What it keeps

A closed record · five fields

The record
Event type, timestamp, page and referrer URL, an opaque session id, order value and currency. Nothing else.
The events
Three: page_viewed, product_added_to_cart, checkout_completed — a visit arriving, turning into intent, becoming an order.
The scopes
write_pixels and read_customer_events.
Sealed at Five fields

What it never touches

No container · never collected

The record
No name, no email, no phone, no address. Burnish does not request the scopes that would carry them.
The events
Nothing else on your storefront is subscribed to.
The scopes
No PII sub-scope is requested, so none is ever sent.
Empty by Design

The sensor, working.

Five channels report into one pixel. Orders land in the ledger as they close.

ChatGPT
Gemini
Grok
Perplexity
Claude
BURNISH PIXEL
$4,512.00 · 100 orders
sessions in orders out

Representative demo · seeded from the fixture’s own figures — $4,512 · 100 AI-referred orders

How a visit becomes a number.

Three readings, ranked. When they disagree, the ledger wins.

  1. The pixel classifies

    Thirteen AI referrer hosts, twelve crawler signatures. A match marks the visit AI-referred.

  2. Orders are the ledger

    Totals, currency, timestamps, refunds, test flag. Level 1, never a customer field.

  3. Journeys corroborate

    Shopify’s own first and last visit: source and referring URL, nothing else.

Representative demo · Threshold Footwear
Representative demo · Threshold Footwear
The AI-channel readings — Revenue · Threshold Footwear fixture

What this cannot tell you.

“Correlation-based, not randomized controlled trial” — the sentence the app itself publishes.

What it shows

That the AI channel grew: which days, against which baseline, at what confidence.

What it does not

That the change brought the growth about. A live store has no control group.

The same words

The sentence merchants see inside the product, not softened for a website.

Thirty days before, thirty after.

Measured inside the 60-day order history Shopify keeps by default, under three standing rules.

The 60-day cohort

The thirty days after a change publishes, against the thirty before — same cohort, same currency.

The publish moment

Recorded with a before-state, so the start of the window is not guesswork.

Seven days, three readings

Until a change clears both, the only number on the page is the days left.

Points before dollars

Lift ships in index points; the money column stays empty until a revenue source is connected.

Zeros stay zero

No smoothing, no projection, no illustrative figures standing in for yours.

Representative demo · Threshold Footwear
Representative demo · Threshold Footwear

Every figure carries its confidence.

Three labels, applied by the app itself, and printed beside the number.

ConfidenceWhat it rests onHow to read it
HighOrders and a connected GA4 property, agreeing within 15%.Two independent readings of the same period.
MediumShopify orders alone.The ledger, with nothing disagreeing with it.
LowGA4 alone, with no order data behind it.Directional at best, and labelled as such.

GA4 can be connected as a cross-check — a second opinion, never blended in. When two sources disagree the orders win, because orders are the ledger. Burnish names the source a figure came from rather than merging them into one confident-looking line.

See whether the AI channel is moving.

The free audit reads your store the way the assistants do. Whether the AI channel is worth measuring is a question the reading answers.

Run a free audit The changes being measured

Agentic AEO runs on Burnish.

Free AI auditFree AI audit