Are you named?
Mention rate
The share of AI answers that bring up your brand when the shopper didn’t ask for it by name.
Why it matters. If AI doesn’t say your name, nothing else you do in this channel counts.
AI visibility measurement & message testingMeasure how AI recommends your brand.
Measure how AI recommends your brand.
Test which message moves you up its shortlist.
We measure how AI describes and recommends your brand, then test your current message against a new one under the same conditions. You get the evidence, the comparison, and the limits.
QuestionsRepeated answersEvidence
Questions in. Repeated AI answers. A report you can check.
What you get when you work with us.
Five steps, from a short brief to proof of what changed. You see the evidence at every one.
Brief
You tell us your category, your competitors and your markets.
You getA question set written the way your buyers ask AI.
AI brand audit
We ask every question five times, across the AI models your buyers use.
You getHundreds of stored answers, not one screenshot.
Report
How often AI names you, where you rank, what it says and what it gets wrong.
You getThe numbers, the quotes and the gaps.
Message test
We test your current message against a new one, before you spend on it.
You getWhich message moves you up AI’s shortlist.
Re-audit
The same questions, a quarter later.
You getProof of what changed.
Steps 2 and 3 · The audit
First, we measure how AI treats your brand.
Four questions, four numbers. The examples below use made-up brands, so you can see what your report will look like.
Are you first?
Position and first-pick rate
Where you appear in the answer, and how often AI makes you its first recommendation.
Why it matters. Shoppers act on the first name or two. Third place is a polite mention.
What does AI say about you?
Associations and tone
The reasons AI gives for choosing you, and the ones it never gives.
Why it matters. The gap between your pitch and AI’s version of it is your content plan.
Is it true?
Accuracy and sources
Every checkable claim AI makes about you, compared with your facts, and where it found them.
Why it matters. Out-of-date prices and wrong policies cost sales you never see.
Step 4 · The message test
Then, before you spend, we test the fix.
An audit tells you where you stand. Before you invest in GEO or brand marketing, find out which message actually moves you. We run your current message against a new one, two ways.
AI shortlist test
We give AI each version of your message and ask it for a ranked top five, again and again. Does the new message get you onto the list, and closer to first pick?
You learn which message makes AI more likely to recommend you. How shortlist lift works
Buyer response test
We have the model role-play your buyer types and react to each version in their own words, then score how interested each reaction sounds.
You learn who the message works for, who it loses, and what they ask next. How we score it
Same questions, same conditions. Only the message changes.
Why brands run this.
Five reasons, each with something we actually found.
See how AI ranks and describes you today.
We asked where to buy jewellery in Singapore 139 times. AI named SK Jewellery in 97% of answers and ranked it first in 0 of 140 shortlists.
Read the jewellery study →Find the gaps in your GEO and SEO.
We asked DeepSeek to describe Insta360 25 times. The newest story the brand tells about itself came up 0 times. For SK Jewellery, “trusted family jeweller” never appeared.
Read the action camera study →Know who AI sends your customers to instead.
For gold savers, AI’s first pick was BullionStar or UOB 80% of the time. For engagement rings it was Love & Co. or JannPaul 98%.
See who wins each need →Test a message before you pay for it.
A Leica partnership line that promised better image quality lifted buyer interest on both AI models. A bare “Leica certified” line lowered it. A hotel group’s lounge rewrite scored no better than the original, so it was shelved.
See the Leica test →Be ready when AI does the shopping.
Morgan Stanley expects 10–20% of US e-commerce to run through AI agents by 2030. Gartner expects traditional search volume to fall 25% by 2026. The brands AI recommends first will take that traffic.
Read our research →Latest research.
Original studies of how AI talks about brands. New work most weeks.
Questions.
Is Windtunnel a GEO or AEO tool?
GEO and AEO describe work intended to improve how brands appear in AI answers. Windtunnel is the measurement and testing layer for that work: it measures AI visibility, citations, competitive ranking, and brand perception, then tests messages. It does not publish content, build links, or promise rankings. Agencies use the audit to set a baseline and the test to show what their work changed.
How does Message Lift work?
Message Lift begins with a stored AI answer from the same audit. One Current message and one New message then go through shared contexts, so only the message changes. Buyer-response and AI-recommendation results are both labeled simulated.
Why run the same audit prompt more than once?
LLM answers are probabilistic. One answer is an anecdote. Repeated answers support a rate, with a confidence interval where the math supports one.
Do results match a consumer chat interface exactly?
No. Windtunnel uses disclosed API/model routes and settings. Those can differ from a personalized consumer chat session. The site makes no consumer-interface equivalence claim.
How is this different from a free “what does ChatGPT say about my brand” checker?
A one-off chat check is one anecdote. Windtunnel samples every prompt repeatedly, reports rates with their sample size and a confidence interval where the math supports one, and stores each answer so every figure traces to its source.
Why do AI visibility scores change from day to day?
Answers are sampled from a distribution, so a single run is an anecdote, not a fact about your brand. Windtunnel reports rates over repeated samples, states the sample size, and timestamps every stored answer.
Does this work for Singapore and APAC markets?
Market context is part of every prompt. Results are labeled by market, route, and date, so a Singapore answer is never pooled with one from another market.