AI search visibility, answered without the hedging.

A few minutes. You add your brand, confirm what we read from your site, and approve the competitors, topics and prompts we suggest — it's confirming, not filling in forms.

The dashboard fills as answers arrive: first metrics after the first collection cycle, and a trend worth acting on after one to two weeks. If you'd rather see the product before signing up, book a demo.

Your brand name, your domain and the market you sell in. The rest we read from your site and put in front of you: what you do, your industry, the brands you compete with, the topics your buyers ask about, and the prompts to track them. You confirm or change anything that doesn't look right.

There's no prompt research to prepare beforehand, and no list of name variations to write out — we work those out for you.

Yes. Your brand details, name variations, competitors, topics and prompts all stay editable in the product. Onboarding is a one-time pass to get you running, not a decision you're locked into.

Expect weeks, not days. Crawlers have to fetch the change, assistants have to incorporate it, and we have to collect enough answers for the trend to separate from noise. Changes surface faster on assistants that lean on live sources than on those answering from internal knowledge.

If you want the full checklist, we published a step-by-step AI visibility audit.

Visibility is the share of collected answers that mention your brand: answers mentioning you ÷ answers received × 100. Collect 200 answers for your prompts in a week, get named in 46 of them, and your visibility is 23%. A prompt that returned no answer in the period is excluded from the denominator rather than counted as a miss, so a collection gap can't quietly depress the number.

It deliberately weights nothing else. Where you appear inside an answer is Position, how much of the conversation you hold is Share of Voice, how you're described is Sentiment. Collection runs on a schedule, typically daily, so read a 7- or 30-day trend rather than a single run — the whole metric set is broken down in our guide to AI visibility metrics.

In every answer we order the brands named in it by where they first appear in the text: the brand named earliest gets rank 1, the next rank 2, and so on. Your Position is the average of those ranks across the answers where you appear — lower is better, and 1.0 means you're named first every time. Named first in one answer, third in the next and second in the third, your Position is 2.0.

Answers that don't mention you don't drag Position down; they show up in Visibility instead. Each value carries a reliability tier, because an average over a handful of mentions isn't stable yet.

Every answer that mentions your brand gets one of four verdicts — positive, neutral, mixed or negative. Those verdicts are combined into a 0–100 score where 50 is the balanced midpoint: neutral and mixed answers count in the denominator, so they pull the score toward 50 rather than up or down. Ten scored answers — six positive, one negative, three neutral — come out at 75.

Small samples swing hard, so read the trend line rather than the daily value.

ChatGPT, Perplexity and Gemini run by default; Copilot is available on Enterprise. During the trial, collection runs on ChatGPT only, and other assistants can be added on request depending on your plan — what each plan includes is on the pricing page.

Every data point comes from running your prompt on schedule: one assistant, one prompt, one country, one point in time.

Because assistants aren't deterministic: the same prompt asked two days running can name different brands, in a different order. A one-day move is sample variance, not a signal — compare 7- or 30-day periods instead.

The smaller the prompt set, the louder the noise: on 20 prompts one changed answer moves visibility by five points; on 200 it moves it by half a point. When a jump needs explaining, check the period first, then whether prompts were added, then the assistant filter — in that order. Our AI visibility audit guide walks through the same checks.

A mention counts only when the form used in the answer matches one of your brand's names or variations. We work those out from your site, but assistants shorten, translate and misspell — if you spot a form we missed, add it in your brand settings and those mentions start being counted.

You set the market and the language for your brand, and prompts are asked in the language they're written in — so the same category can be tracked in two languages side by side. Starter and Growth cover Ukraine and the USA; Enterprise has no country limit.

Both. Alongside the answer data we audit your own site the way an AI system reads it: structured data, the technical readability of each page, and whether robots.txt actually lets AI crawlers in.

We also log which AI crawlers fetch your pages, so you can see the supply side of your visibility — not just the answers, but what the assistants had to work with. That half of the product is Source Gap Analysis.

A small analytics snippet, so we can log which AI crawlers fetch your pages. And an MCP server, so you can query your own visibility data from your AI client instead of opening the dashboard.

No. SEO helps pages rank higher in search engines. AI search works differently: instead of just showing a list of sites, it composes a ready-made answer from a mix of sources — websites, reviews, forums, documentation, media and public brand mentions.

So a high Google ranking doesn't guarantee that ChatGPT, Perplexity or Gemini will mention your brand. In AI search, what matters isn't only being found, but being understood correctly, cited and recommended.

Most tools answer one question: are we mentioned? We start a step earlier — which questions your buyers actually ask, which of them you're absent from, who the assistant recommends instead of you, and which sources shaped that answer, down to the pages AI crawlers fetch from your site. Those two halves are Prompt Research and Source Gap Analysis.

The analysis is never better than the data underneath it. If the prompt set, the sources and the competitive set are captured loosely, the dashboard will be confidently wrong.

Yes. AI assistants pull from Google and from the open web. SEO is necessary but no longer sufficient. Treat AI visibility as a parallel channel with its own playbook.

One prompt is one question, written the way a customer would ask it. Your plan caps how many prompts you track — 50 on Starter, 100 on Growth, 300+ on Enterprise — and how many answers we analyse per month: 1,500, 9,000 and 36,000 respectively. On the trial you get 50 prompts for the whole account.

The two limits are linked: one prompt, run against one assistant, in one country, on one day, produces one answer. Adding assistants or countries spends the answer allowance faster than adding prompts does — the current limits are on the pricing page.

Yes. Write to us or use the contact details in your settings and we'll move you to another plan; the change applies from your next billing period.

15% off when you pay annually — the plans and prices are here. Talk to us for multi-year terms.

You can cancel at any time, and your plan stays active to the end of the period you've paid for. If you need a refund on an annual plan, write to us and we'll sort it out.

The trial runs on ChatGPT, with 50 prompts for the account and a fixed allowance of scheduled requests — whichever runs out first ends it. Site audits and team access are included.

When it ends, collection stops, but your prompts, competitors, metrics and audit results stay in the account.

The answers we collect about your brand are yours, along with everything built from them — mentions, citations, competitor comparisons and audit results. We keep them for a year, so you always have a rolling year of history to look back on.

If a trial or a plan ends, collection stops, but what's already been collected stays in your account. You can also pull it into your own AI client through the MCP integration.

On our managed cloud infrastructure. If you have specific residency or processing requirements, tell us before you buy and we'll confirm in writing what we can commit to.

Every plan has unlimited seats — invite the whole team and the people on the client side, on any plan. Roles control who can change prompts and who can only read the dashboard.