When buyers ask ChatGPT for the best tools in your category, does it name you — or your competitor?
51% of B2B buyers now start research in an AI chatbot, and 69% pick a different vendor after it answers.
— G2, 2026
From $19 · pay per scan · auto-refund if it fails
6 engines · 90–720 live answers per scan · ranges, not fake precision · from $19
Each LLMBrandScan run asks between 90 and 720 real AI questions across up to 6 engines — ChatGPT, Claude, Perplexity, Gemini, Grok, and Google's AI answers in Search (AI Overviews) — and reports the observed range instead of one lucky answer.
youtube.com — 188 citations, the top-cited source → earn a mention in a workspace-comparison video
ClickUp holds 20% share of voice, your closest rival → target head-to-head comparison prompts
reddit.com — 77 citations across category threads → join the conversation with a transparent comparison
The report you walk away with.
Real scan — Notion, July 2026 · 180 live answersA real scan we ran on Notion — this is the exact structure you get, not a stripped-down preview.
Share of voice
Wider ranges mean less stable answers — we show you the spread across samples, not a false-precision single number.
3 of 12 prepared prompts · 11 Strong, 1 Noisy overall.
Why now.
Buyers already ask an AI before they ask a salesperson. Here's what's happening in that conversation whether you're watching or not.
51%
of B2B buyers now start vendor research in an AI chatbot instead of a search engine.
— G2, 2026
92.8%
of ChatGPT shopping answers send the buyer nowhere — zero clicks to the open web.
— Profound, 2026
24% vs 11%
Brands buyers ultimately choose show up in roughly 1 of 4 AI answers; passed-over brands, barely 1 of 9.
— Profound, 2026
15.9% vs 1.76%
ChatGPT-referred traffic converts at 15.9% — Google organic search converts at just 1.76%.
— Airfleet, 2026
Built like a brand tracker.
Our three question lenses aren't invented — each one mirrors a construct brand trackers have measured for decades. The AI assistant is the new “mind” we're measuring, not a survey panel.
Unaided recall + mental availability via Category Entry Points — does the brand come to mind before anyone names it?
— Byron Sharp & Jenni Romaniuk, Ehrenberg-Bass
Aided recall — how the model describes the brand once it's named.
— Classic aided / prompted-recall canon
Consideration set — which brands the model puts in the choice frame together.
— Consideration-set theory; Share of Voice, Binet & Field
"Share of AI answers" — the next-generation Share of Search.
— Les Binet, EffWorks 2020
“LLMBrandScan measures the same thing brand trackers have measured for decades — whether your brand comes to mind when a buyer asks about your category — but the ‘mind’ is now the AI assistant your customers ask instead of a survey panel. It's a sample, not a census: we run your category's questions across multiple engines multiple times and show you the range, so you see both where you stand and how stable that standing is.”
A sample, not a census.
— Variance-decomposition analysis of 12,933 AI brand answers, 2026
What you pay for.
Tell us your domain, approve the questions we prepare, and we ask every engine live — every credit buys real model calls, not a subscription seat.
| Depth | Questions | Engines | Samples | Live answers | Credits |
|---|---|---|---|---|---|
| Light | 10 | 3 | 3 | 10×3×3 = 90 | 1 |
| Standard | 25 | 6 | 3 | 25×6×3 = 450 | 3 |
| Deep | 40 | 6 | 3 | 40×6×3 = 720 | 5 |
Deep adds questions, not repeats — phrasing diversity buys ~15× more signal than re-asking the same one. Samples stay 3 everywhere: they detect instability, not average it away.
$149 → 450 live queries across 6 assistants.
That's the real cost of running your questions live on every engine, 3 times each — not a margin we're hiding.
A failed call is 0¢. A fully failed scan refunds itself.
Enforced in the billing code, not a policy page — a dropped or errored engine call is never charged, and if every engine call in a scan fails, the whole scan refunds automatically.
One audit, one price. No seat you'll forget to cancel.
Monitoring tools charge $20–499 a month to keep a dashboard warm. This is a different purchase: a dated audit you run when the question comes up. Every category question is asked on six engines and sampled three times, so each number arrives with a confidence interval — not a single reading that moves when you refresh.
Sample
$19
1 Light scan — 10 questions x 3 engines
Audit
$149
1 full Audit — 25 questions x 6 engines (or 3 Light scans)
Deep Audit
$249
7 credits = 1 Deep Audit (5) + 2 Light re-scans (1 each)
Agency
$990
10 full Audits — one per client
Failed engine calls are never charged — a fully failed scan refunds automatically. That's enforced in the billing code.
Your first scan is the baseline.
“Increase share of voice in AI answers from 14% to 22% by Q3” is how marketing teams already write the OKR. Re-scan monthly, or right after a campaign, to show the board a trend — not a single number.
The Deep pack already bakes this in: one Deep audit plus two Light re-scans, so you can prove movement without buying credits twice.
Who it's for.
Nobody owns AI visibility yet — SEO, brand, and content all get handed the question. Get a defensible number in an afternoon, self-serve — no procurement cycle, no onboarding call.
Is ChatGPT sending your buyers to a rival?
Run the scan before your next board update or fundraise — know your AI shortlist status cold, instead of guessing.
Prove your AI visibility to your CMO.
Sourced numbers and an observed range, not a hunch — a report built to drop straight into a QBR deck.
A per-client deliverable. No seat lock-in.
Buy scans in packs, brand the findings for each client, and skip the monitoring-tool contract neither of you wants.
Query your AI visibility from your own tools.
LiveThe pay-per-call MCP nobody else in the category ships. Plug check_brand_visibility straight into Claude, Cursor, or your own agent — no subscription, no seat required.
Questions, answered honestly.
Which AI engines do you check?
Six: ChatGPT (via API — more on that below), Claude (API-grounded proxy), Perplexity Sonar, Google Gemini, Grok, and Google's AI answers — the AI Overviews box at the top of Google Search. Those are the engines buyers actually use for AI SEO / AEO research today, not just the one your team happens to have open.
Is that the real ChatGPT?
We call OpenAI's API with search grounding enabled — the same model family, without the chat memory or plugins of the consumer app at chatgpt.com. We label every ChatGPT result an “API-grounded proxy” rather than pretend it's identical, because it isn't, and a tool that doesn't say so is being dishonest with you.
How accurate is this?
AI answers are non-deterministic, so we don't trust a single run. We sample each question 3 times and show you the observed range — never a confidence interval or a claim of statistical significance; 3 draws can't support one. Real accuracy comes from coverage: asking across more engines and more phrasings buys roughly 15× more signal than repeating the same question (a published variance study of 12,933 AI brand answers). It's a sample, not a census — an honest read on where you stand today.
Why is there no free scan?
Every scan runs 90 to 720 live paid queries across real AI engines — there's no way to run one without spending real API money. A free tier would mean cached, throttled, or fake results, and we'd rather charge $19 and run it live than pretend otherwise. If a scan fails outright, it refunds automatically — that's our substitute for a free trial.
Is this a subscription?
Scans aren't. You pay once per scan — packs start at $19 — and get a report plus a fix list; the credits are yours to spend whenever you like, with no seat and no dashboard you're billed for whether you open it or not. The one recurring thing we sell is optional monitoring: after a report you can subscribe to have that same scan re-run on a schedule, and cancel any time. Buying a scan never signs you up for it.
How is this different from AI-visibility monitoring tools?
Monitoring tools sell you a dashboard to check forever, starting on day one. We sell the diagnostic first: an honest, one-time answer and a ranked list of what to do next. Tracking is something you add afterwards, only if the report convinces you it's worth watching — and if you'd rather not commit to anything, just buy another scan when you want the next read: before a launch, a rebrand, a competitor's push.
What do I actually do with the report?
Every finding ships with a fix: which prompts you're losing, which competitor and source is winning them, and the concrete action to take — earn a review, publish a comparison, get cited on a page the AI engines already trust. Built to hand straight to whoever owns your AI SEO / AEO work, not to sit in a browser tab.
Do I control what it asks?
Yes. We prepare a tailored set of buyer questions from your domain, but nothing runs until you approve it — check the ones to test, edit anything that's off, add your own. You're never trusting a black box with your questions.
What happens if an engine fails?
You're never charged for it. A dropped or errored call to any engine is 0¢, and if every engine call in a scan fails, the whole scan refunds automatically — both enforced in the billing code, not a policy we're just promising.
See where you stand before your next buyer asks the AI.
From $19 · pay per scan · auto-refund if it fails
Not ready to run one?
Get occasional launch notes.