Run Check and the Report

Press Run Check in the brand's page header. There is one button for everything; the Report reloads by itself when the check finishes. A check covers:

• Agent Search Visibility: your questions, asked to ChatGPT search and Perplexity search.
• AI Chat Visibility: the same questions, collected from the ChatGPT and Gemini apps.
• Local & Maps: for local businesses with a target location, a local question asked to Google AI Mode.
• Google AI Overviews: if you turned it on, Google Search with its AI answer.
• AI Readiness: a fresh read of your website (up to 10 pages), page speed, the checks for AI agents and, for local businesses, your Google Business Profile. After the answers arrive, S6S also reads the pages AI cited most.

Each Agent Search question is asked 3 times on each engine, because AI answers vary from one ask to the next. Which questions a check asks

If some kinds of check return results and others don't, the check ends as Completed with partial results and the Report shows what arrived. It ends as failed only when nothing usable came back.

• A check uses 10 credits, charged once when it starts. The price is the same whatever engines or kinds of check your brand has.
• Pressing Run Check again while a check is still running returns the same check and charges nothing.
• Each brand can run up to 20 checks in any 24 hours.
• Some partner accounts include free checks: one free full check per business, up to a monthly number. Their quote shows 0 credits.

Through the API or the Visibility MCP, a free quote (GET /api/visibility/brands/{brandId}/report/quote) shows the credits, your balance, the daily limit, and the exact questions and engines before anything runs. It asks no engine and writes nothing.

In each of these cases nothing is created and nothing is charged:

A brand has four tabs: Report, Fix Plan, Progress and Questions. The Report reads top to bottom:

The verdict

One sentence, driven by the kind of check that matters most for your business: for example “AI knows you — but doesn't lead with you yet”. A measured result leads it; an early read leads only when nothing is measured yet, and says so. It never contradicts the count under it: it says “doesn't mention” only when no answer named you. The Fix Plan and the Progress tab open with the same count.

AI Visibility: answers by AI company

The main card counts answers, not scores: “Named in X of Y answers · N AI companies”. Each company lists every way S6S checks it:

A row reads like “named 1/3”, “not named 0/9” or “unclear 0/1”. A count only compares within its own row. A row built from too few answers is marked early read. Answers S6S couldn't read are counted apart and never shown as “not named”. A row that wasn't refreshed by this check shows the date it was last checked.

An answer names you when it uses your name, one of your other names (brand settings), or your name's short form: your name without the words for your kind of business, its legal form or your town (“Hoekstra & Van Eck” for “Hoekstra en Van Eck Makelaars”), and only when the short form isn't followed by another name (“… Hypotheken” is another business). “&”, “en” and “and” count as the same word. When Share of Voice lists a name that looks like yours, This is us? adds it to your other names; adding or changing other names recounts the latest check from the answers already collected, with no credits and nothing asked again. Earlier checks keep the count they were made with.

Scores by kind of check

Under the main card, each kind of check gets its own row and its own state. The one that matters most for your business is tagged Matters most and listed first once it has answers. A number out of 100 appears only for a measured result, or as a muted, directional number for an early read. How the score is calculated

Normal swing

AI answers move around even when nothing changed. Under the leading score, S6S compares this check with earlier checks that ran exactly the same way (same questions, engines and settings), using at most the 5 most recent from the past 42 days. It needs 3 of them before it gives a label:

• “Within this brand's normal variation”: the change is no bigger than how much those earlier checks differed.
• “Higher” or “Lower than this brand's normal variation”: a bigger change. A change of up to 5 points always counts as normal.
• “Normal swing not known yet: k of the 3 comparable checks needed”: not enough comparable checks yet.

This reading is available for Agent Search Visibility and Google AI Overviews; the other kinds of check say it isn't tracked yet. Checks run before full check settings were recorded (25 September 2026) are never compared. A changed set of questions starts a new baseline. The swing never says “improved” or “declined”, and a recorded fix between two checks is noted, not credited.

More on the Report

• By buying situation: when a check asked questions for your buying situations, a breakdown shows how often you were named for each, and for your category in general. These are counts, not scores.
• Measured on N questions · Review: how many of your questions got usable answers, with a link to the Questions tab.
• Share of Voice: your share of the business names counted in the answers, next to each competitor's. Every answer, from AI Chat, Agent Search and Local & Maps, is read for the businesses it names. Only direct competitors the answer named, and the competitors you set, are counted; product codes, standards and cited sites are not. A competitor you set counts in every answer whose text names it, even when the reader couldn't classify the name (since 21 Oct 2026). Until an answer has been read, the card says so instead of saying nobody else was named.
• Sites your AI answers cite: which sites each engine cited for which question. They feed the Reputation part of AI Readiness.
• AI Readiness: a separate letter grade for the inputs you control, with a free “Check readiness” button.
• What's holding you back: the biggest gap S6S found, with a link to the Fix Plan. It sits right after the AI Readiness summary, before the lists of cited sites and answers.
• Before the first check, the Report shows one panel: what the first check asks, what it costs and how long it takes.
• While a check runs, the panel under the page header lists each part (Agent Search, AI Chat, Local & Maps, AI Readiness, the Report) and where it stands, and the Report reloads as each part finishes. A check with AI Chat answers usually takes 10 to 20 minutes (S6S's own checks, October 2026).

The Progress tab lists every past check with its Agent Search Visibility score, the number the Report showed for that check, how long the whole check took and whether part of it couldn't be collected, and compares two of them. When two checks asked different questions, the comparison is marked “Not a before/after”.

Share / Export in the page header opens a branded report for your client, with a share link that works for 30 days, print-to-PDF and email. It shows every kind of check the Report shows (Agent Search, AI Chat, Local & Maps and Google AI Overviews), each by engine with the Report’s own numbers and never added together, then the Report’s Share of Voice (you ranked among your competitors) and its AI Readiness grade. The question table covers only the questions the latest check asked, and the 90-day line only the days that asked exactly those questions; when your questions changed, the report says so instead of drawing a line through answers to other questions. Its CSV lists every answer behind those numbers. For developers it offers the Fix Plan as a downloadable bundle, or as text to paste into an AI coding tool. Fix Plan