Know what ChatGPT says about each client. Fix it on their store.
ChatGPT and Claude already recommend stores by name. Every week, each client's buying queries are run against them and the standings are recorded per client — who is cited, at what position, against which competitors. When a query goes to a competitor, the fix is drafted on the client's site and waits for your sign-off.
Already have an account? Set this up on a client store →
24 queries × ChatGPT, Claude, Gemini, 5 runs each. Visibility 41% (+6 pts). Bluefin entered the list for “best reef-safe sunscreen” at position 3; lost “rash guard for kids” to Harbor & Co.
Your clients are being recommended — or not — and nobody is watching.
A shopper asks an assistant which product to buy and gets a named list. That list is the new first page, it changes week to week, and there is no console for it. Checking it by hand for ten clients is not a process anyone keeps up.
No instrument
There is no Search Console for ChatGPT. Unless someone runs the query, you don't know whether the client is cited — or that a competitor took the slot last week.
The miss has a cause
An LLM skips a store because the page never answers the buying question directly, the spec is missing, or the category is thin. The gap is on the site, and it's fixable.
Clients ask, you guess
“Are we showing up in AI search?” is now a standing question on every call. Without a tracked number per client, the answer is an anecdote.
One scan policy per client. Fixes ask; the scan doesn't.
The scan is deterministic and cheap to re-run, so it goes on autopilot on a weekly schedule. The fix — new copy on a product or category page — is a write to the client's store, so it asks.
- Scoped to one client: its queries, its brand, the 3–10 competitors it actually compares against.
- Runs on your own OpenAI, Anthropic or Google key. Models are pinned and disclosed, so a change is attributable to the content, not a silent model swap.
- Scan step on autopilot: five runs per query per engine, averaged. Digest of what moved lands in the inbox.
- Fix step asks: the draft is written on the client's Shopify, Shopware or WordPress and waits for sign-off — web or mobile.
Schedule → scan → draft → verify → ledger.
Monday, per client
Each client's policy runs on its own schedule with its own query set — the queries it already ranks for in Search Console, or the ones you typed. Up to 50 per account.
Standings, week over week
Visibility (share of query × engine pairs where the brand was cited) and average position (slot in the LLM's ranked list, same convention as Search Console). What moved up, dropped off, or entered for the first time.
Every miss traced to a cause
For a query lost to a competitor, the agent reads the client's page and the competitor's answer, names the gap, and drafts the fix on the actual store — as a draft, never live. You approve, edit or deny.
Detected → fixed → verified
After approval the page is re-crawled and the fix confirmed live. Standings, fixes and who signed off are on the client's ledger; a public report link shows the client the trend without a login.
Typed visibility actions — the scan is the instrument, not the product.
You approve “update the description on /products/reef-safe-spf50 with the draft below”, never “let the AI improve our AI presence”. Scans read. Fixes write, and ask by default.
- toolRun the weekly scan across ChatGPT, Claude & Gemini
- toolTrack the client's brand plus its competitors
- toolSuggest queries from the client's Search Console
- toolDraft product & category copy that answers the buying question
- toolRe-crawl and verify a published fix
- toolPublish a per-client standings page, no login required
Put every client's AI visibility on a weekly scan.
Start with the free check — a real crawl of one client site, results in about a minute. Then attach the scan policy per client and approve the first fix. Free for one store; agency plans from the second.
Free for one store · 14-day trial on paid plans · no card · bring your own AI key