Reputation — the fixed battery
12 questionsThe same twelve for every store on Burnish. All twelve fire the day you install, then the stalest question goes next.
Coverage
Five engines, read on a schedule.
Burnish asks the assistants your customers ask, and keeps every answer with its date and engine.
All plans
GPT-5.4 mini
ChatGPT The highest-traffic shopping assistant, and the one merchants are asked about first.
Read the engine page All plansSonar Pro
Perplexity Answers carry visible sources, so citation position is directly observable.
Read the engine pageEvery plan
What does not change.
The reputation battery is identical on all three plans. No tier buys a faster read of its own reputation.
- Questions in the battery 12
- A full cycle, every plan 72 hours
- Scoring window, rolling 21 days
- No confidence band below 10 samples
By plan
What the plan changes.
Presence, then Authority, then Omnipresence, in that order. These four lines are what a plan actually buys here.
- Engines sampled 3 · 5 · 5
- Questions on the board 50 · 50 · 100
- Minutes between firings 120 · 90 · 35
- AI answers a month 1,500 · 3,000 · 7,500
What a reading is
Two kinds of question, asked differently.
One track holds still, so a change in the answer means a change in your store. The other moves with your buyers.
What the twelve ask about
FixedLegitimacy, Safety, Reviews, Identity, Quality, Returns, Shipping, Warranty, Value, Support, Provenance, Standing. Only the brand name and category change.
Brand name shown, domain withheld
RecognitionThe engine never sees your URL, so it has to recognise you the way a customer would.
Discovery — your own board
50 or 100Open-ended questions a shopper would actually type, held on a tracked board rather than asked once and forgotten.
Weighted, so nothing starves
Worth × ageFiring order weights what a question is worth against how long since it was last read.
A question earns its place
14 daysA question stays on the board at least fourteen days before it can rotate out.
Five is the whole list. Three others are recognised when they send you traffic, but Burnish does not sample them and does not count them as coverage.
Monthly answer figures are the allowance a plan includes. Below ten samples Burnish draws no confidence band and says so.
Questions
Answered from what is already on this page.
Which assistants does Burnish actually sample?
Five: ChatGPT, Perplexity, Gemini, Claude and Grok. Three on Presence — ChatGPT, Perplexity and Gemini — and all five on Authority and Omnipresence.
Do you cover assistants outside these five?
No. Three others are recognised when they send you traffic, but they are not sampled, and Burnish does not count an unsampled assistant as coverage.
How often is a question asked again?
The twelve reputation questions complete a full cycle every 72 hours on every plan. Discovery fires one board question every 120, 90 or 35 minutes, by plan.
Does the engine get my domain?
No. It is shown your brand name and never your URL, so it has to recognise you the way a customer would.
How do questions get onto the Discovery board?
They are open-ended questions a shopper would type, held at 50 on Presence and Authority and 100 on Omnipresence. A question stays at least fourteen days before it can rotate out.
When does a score become trustworthy?
Scores are computed over a rolling 21-day window, and no confidence band is drawn on fewer than ten samples. Below that the app says it does not have enough data yet.