How Fond works

Fond asks AI assistants the questions that your customers ask, again and again. It reads every answer for four things:

• whether the answer mentions your brand
• where in the answer your brand comes up
• which sites the answer drew on
• whether one of those sites was yours

Fond measures the visibility figures over the answers that it collected, not over a crawl of your site or your traffic. A few figures come from other sources, and each one says so where it appears:

• Your Google ranking comes from Search Console.
• AI-referred sessions come from Google Analytics.
• AI-referred requests come from your own edge.
• Agent Readiness comes from your site itself.

The cycleCopy Anchor Link

Each pass over your questions is a run.

1. Fond sends every question in your query bank to every assistant enabled for your runs, up to what your plan allows. By default, it asks each assistant each question three times. So 20 questions across 4 assistants give 240 answers. A run that you start through the REST API can ask a subset of the questions.
2. Fond reads each answer. It records whether the answer mentions your brand, and in which third of the answer your brand first comes up. It records how the answer speaks about you, the follow-up questions that it suggests, and every page that it drew on. It records whether your site was one of those pages, and whether the answer linked to it.
3. From the answers of the run, Fond calculates these figures:
   • mention rates
   • share of voice by question tier and by your topic clusters
   • citation counts for each site
   • the questions where you barely come up

With ongoing monitoring, Fond runs this cycle every other day at the same hour, unless it has set another interval for your site. The free audit plan has one run, which starts when you finish the website setup. Your plan sets a weekly run ceiling, and runs that you start count against it. To get answers sooner, click Run Now on the website.

Fond records the assistants, the market and the models on each run when the run starts. A later change applies to the next run, and never rewrites an old one.

For some assistants, Fond does not choose the model, and the assistant does not report which model it used. If the provider changes that model, your older runs show no sign of it. A step in that assistant's line can come from the model change, not from a change about your brand.

What a run isCopy Anchor Link

A run moves through a few working states. Then it stops at one of three terminal statuses:

Completed with errors is a normal status. Assistants time out and hit rate limits, and a failed request is not an answer that ignored you. So rates that count answers divide the answers that mention you by the answers that came back, and leave failed requests out.

Figures that count questions differ. Share of voice keeps a question whose answers all failed, as a question that you did not win.

The competitors leaderboard and the Gaps card leave that question out. The glossary says which figure counts which.

The Query Results page shows which requests failed, and offers a retry where Fond can run one again. Fond cannot retry two kinds, and the page says so:

• a request whose source Fond no longer collects from
• one question and assistant that Fond measured in two different ways in the same run

What Fond measuresCopy Anchor Link

The glossary defines each figure once, with the denominator that each page uses. The same figure can divide by answers on one page and by questions on another. Watch organic visibility most, because its two readings can differ by a lot on the same run.

• AI visibility is the share of answers that mention your brand. Organic visibility is the same rate over only the questions that did not name you.
• Share of voice is your mention rate, split by how competitive the question is.
• Average answer position is how early in an answer you come up.
• Citations and organic citations count how many answers drew on a site as a source.
• Average search position is where your pages rank in Google, from Search Console.

Where to startCopy Anchor Link

• Getting started adds a website, from the URL to the first run.
• How tracking works covers your questions, the assistants, the market, the run schedule and which run is the latest.
• Visibility is the page that you will read most. Its guide tells you how to read the trend.
• Citations covers which sites assistants draw on and where "Get cited on" suggestions come from.
• Competitors covers the leaderboard and how Fond decides that two names are one brand.
• Agent Readiness checks what an assistant can read and buy from your site. Site Audit is a one-off review of the site. Neither uses answers.
• The glossary defines the shared metrics as the app tooltips do. A figure that only one surface shows, such as the shopping-card rates, is explained on that surface.
• Plans and limits gives what each plan allows.
• The REST API reference tells you how to read your data and start runs with an API key.

What Fond does not measureCopy Anchor Link

The answers your shoppers gotCopy Anchor Link

Fond asks its own questions from its own market. An assistant answers the same question differently from one minute to the next. So Fond asks each question several times and reports a rate.

Every assistantCopy Anchor Link

A run uses only the assistants enabled for your runs, at most what your plan allows. Results on an assistant that Fond does not ask do not appear in your numbers.

Questions you do not trackCopy Anchor Link

Fond measures the visibility and competitor rates over your query bank only. When you add or archive questions, you change what the rates cover. So the visibility trend marks the runs where the question set changed. The AI traffic and crawler figures count your site's own traffic, not your query bank.

Whether AI sent you a saleCopy Anchor Link

Visibility shows what assistants say, not what shoppers do next. The GA4 and Cloudflare integrations report AI-referred traffic. Analytics counts sessions and the edge counts HTTP requests, and one session is many requests. Read each on its own terms, and do not compare or subtract the two. Edge measurement is still rolling out and is not on every account yet.