このガイドは英語でご覧いただけます。

AI visibility

Track whether ChatGPT, Perplexity, Google AI Mode, Gemini and Claude mention and cite your brand, who they recommend instead, and how to fix it.

More and more buyers ask AI assistants instead of searching Google. AI visibility tracks what those assistants say when your customers ask the questions that matter to you. It shows whether your brand is named in the answer, whether your site is linked as a source, and which competitors and websites get recommended instead.

It's for anyone working on generative engine optimization (GEO): brands checking their presence in AI answers, and agencies reporting it to clients. Open it from Project → AI visibility, or open it directly.

How it works

  1. We ask AI your customers' questions. Each tracked prompt is sent to the engines you picked: ChatGPT, Perplexity, Google AI Mode, Gemini and Claude. The country and language come from your project.
  2. We check whether you're in the answer. An answer mentions you when it names one of your brand names or your domain. It links you when your domain is among its cited sources.
  3. We show who's recommended instead: the other brands named in the answers, and the sites the engines cite.

Engines per plan. Starter includes up to 3 engines per prompt (ChatGPT, Perplexity and Google AI Mode are pre-selected). Pro and Agency include all 5. You choose the engines when you add prompts.

How often prompts are re-asked. On Starter, every prompt is re-asked every 2 weeks. On Pro and Agency, it's every week. If a prompt's last three answers on every engine agree (same mention, position, link and competitors), it's marked Stable and re-asked half as often until something changes.

Prompt limits apply across all your projects: Starter 25, Pro 100, Agency 250. You can add +50 AI prompts for $29 per month on the Billing page.

Add prompts step by step

  1. Open Project → AI visibility and scroll to Discover prompts.
  2. In Category / what you sell, describe your offer, for example "CRM software for real estate agents". It's pre-filled with your first tracked keyword.
  3. Click Discover prompts. You get buyer-intent questions written in your project's language, built from your brand, competitors, tracked keywords and country. Each one is tagged discovery, comparison, evaluation, local or problem. Prompts you already track are left out.
  4. Untick the ones you don't want, choose the engines (up to your plan's limit) and click Add selected. The prompts run right away.
  5. To write your own, use Add prompts below: one question per line, in the language your customers use. Pick the engines and click Add and run.
  6. Wait a moment for the answers, then review the Results by prompt table. Click any cell to read that engine's full answer.
  7. For any answer that leaves you out, click Why not visible? to open the diagnosis (see below).

You can run Discover prompts up to 10 times an hour. To stop tracking a prompt, click Delete on its answer card.

What the numbers mean

All figures use the latest answer for each prompt on each engine.

Metric Meaning
Brand mentioned Share of AI answers that name your brand
Site linked Share of answers that cite your domain as a source
Avg. position Your average position when you are mentioned
Prompts Prompts in use against your plan limit, and the number of answers checked
  • Position – if the answer is a numbered or bulleted list, it's your item number in that list. Otherwise it's the order in which you're named among you and your competitors.
  • Visibility trend – the share of answers that mention you, per day of runs.
  • By engine – mention rate and link rate for each engine.
  • Results by prompt – #2 means you were mentioned in position 2 (a link icon means you were also cited). Only linked means you were cited without being named, and Not in answer means neither. Top sources AI used lists the domains each prompt's answers cite.
  • Brands AI recommends – other names highlighted in the answers. These are often competitors you aren't tracking yet.
  • Sites AI cites – the domains engines use as sources most often. Getting featured or reviewed on them helps you show up more.
  • Share of voice – how your mentions compare with your competitors' mentions across the answers. Add competitor domains in Project → Project settings to see it.
  • Sentiment – Positive, Neutral or Negative, judged from the wording around your first mention.
  • Branded prompts contain your brand name. Any answer repeats that name, so for those prompts only your domain or your other brand names count as a mention.

When an engine starts or stops mentioning you compared with the previous run, you get an AI visibility changes alert. Pick its channels in Settings → Notifications.

"Why am I not visible?" diagnosis

Click Why not visible? (or Analyze when you are mentioned) on an engine's answer. You can also open the diagnosis page directly and pick a prompt from there. We fetch your page and up to 5 of the pages that engine cited, then score each one on GEO signals. This usually takes 5–15 seconds.

  • Your page is picked automatically: the ranking URL of a tracked keyword that matches the prompt, or your homepage. Enter another URL and click Compare to test a different page.
  • The cards show Your brand (mentioned or missing, cited or not), Your page score (0–100), Cited pages avg. and Third-party sources (review, community, media and reference sites among the citations).
  • What to do lists fixes ranked High, Medium or Low by expected impact. Examples: answer the question in the first 100 words, get listed on the review sites the engine cites, add an FAQ or Organization schema, or add concrete numbers.
  • Your page vs. cited pages compares the signals side by side. The GEO score adds up: direct answer up top (20), FAQ section (10), FAQPage schema (5), entity schema (10), headings (10), lists and tables (10), statistics (10), word count (10), freshness (10) and outbound citations (5).
  • All cited sources labels each source as a review site, community/UGC, reference, media/listicle, competitor or your own site.

Tips

  • Write prompts the way a buyer would ask them ("best X for Y", "X vs Y", "is X worth it"), not as keywords.
  • Keep your brand names unique to you. A generic word like "CRM" will match almost every answer.
  • Add your real competitors so share of voice and competitor badges mean something.
  • If your brand names change, editing them in Project → Project settings re-analyzes your existing answers for free.
  • Run AI readiness too. If AI crawlers can't reach your site, the engines can't cite it.
  • Want a quick one-off check for a prospect? Try the free AI visibility checker.

FAQ

Which AI engines are tracked?

ChatGPT, Perplexity, Google AI Mode, Gemini and Claude. Starter covers up to 3 engines per prompt, and Pro and Agency cover all 5.

How often are my prompts re-asked?

Every 2 weeks on Starter and every week on Pro and Agency. Prompts marked Stable are re-asked half as often until an answer changes. New prompts run as soon as you add them.

Why does a prompt show "Branded"?

The prompt itself contains your brand name, so every answer repeats it. For those prompts we only count your domain or your other brand names as a mention. Otherwise every answer would look like a win.

What's the difference between mentioned and linked?

Mentioned means the answer names your brand in the text. Linked means your domain is among the answer's sources. You can be one without the other. The diagnosis page suggests fixes for both.

Can I change which engines a prompt runs on?

Engines are chosen when you add a prompt. To use different engines, delete the prompt and add it again with the new selection.