> ## Documentation Index
> Fetch the complete documentation index at: https://docs.visibility.so/llms.txt
> Use this file to discover all available pages before exploring further.
# AI Visibility
> Track how often ChatGPT, Claude, Gemini and Perplexity name your brand when buyers ask about your topics, set up the prompts that get checked, and read the score with the right expectations.
AI Visibility shows how often four AI assistants — **ChatGPT**, **Claude**, **Gemini** and **Perplexity** — name your brand or link to your site when asked the questions buyers actually ask about your topics. You decide which questions are asked; Visibility runs the checks on a schedule, keeps every answer, and turns the results into a score you can follow over time.
Open it from the project sidebar under **Reports → AI Visibility**.
## What it measures
Each check asks one **prompt** (a natural-language question such as "What are the best store locator apps for Shopify?") of one assistant and records the full answer together with the links it cites. An answer counts as featuring you when:
* one of your **brand names** or your project domain appears in the answer text, or
* your site appears among the answer's citations.
Prompts are grouped under **topics**, so you can see where you show up — and where you don't — subject by subject. Across a set of answers, Visibility also records which competitors were named, which brands led each topic, and which sites the assistants cited most.
AI Visibility measures presence, not tone. It tells you whether and how prominently you were named — not whether the assistant recommended you.
## The page at a glance
| Tab | What it shows |
| ------------ | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Overview** | The visibility score ring, a score per assistant, the trend over time, a brand-presence grid for every prompt, competitors mentioned and top cited domains. A date-range filter (**7d**, **30d**, **90d**, **365d** or **Custom**; 30 days by default) sits in the header. |
| **Topics** | One row per topic with its visibility percentage and prompt count. Click a topic for its brand leaderboard, the actual answers and the cited sources. |
| **Setup** | Your topics and prompts, which assistants each prompt is asked of, and a live cost estimate. |
A banner above the tabs shows when the next automatic run is due and when the last one finished. Until a project has at least one prompt, the tabs are hidden and only the Setup card (**Prompts by topic**) is shown.
If the page is missing from a project's sidebar, a company admin can switch it on under the project's **Settings → Features → Pages → AI visibility**. New projects have it on already.
## Set up tracking
### 1. Tell Visibility what to look for
Brand names, competitors and the market live in the project's general settings. Company admins can edit them; other members see them read-only.
In the project, go to **Settings → General** and scroll to **SEO + GEO**.
Under **Brand names**, type a name and press Enter. One or more names are pre-filled from your domain; add product names and any spellings the assistants might use. A name that is very short or an everyday word (store, app, cloud, design) shows an amber warning because it will match unrelated text and overstate your visibility — prefer distinctive names.
Under **Competitor domains**, type a domain such as `competitor.com` and press Enter or a comma. You can paste a list separated by commas, semicolons or new lines. Once answers start coming in, a **Suggested competitors** box lists domains that already appear in them — click a pill to add one. Directory sites are listed separately; adding a directory as a competitor removes it from the AI Citations tab in [Opportunities](/opportunities).
**Location** is the market the assistants are asked from (**United States** by default). It is the only locale setting for AI Visibility.
Click **Save changes**.
The project **Domain** is your brand's domain in AI answers and is locked once any data has been collected. To track a different site, create a new project.
The **Topics** pills in this section are descriptive project context. They are not the topics AI Visibility asks about — those are created on the Setup tab, next.
### 2. Add topics
Go to **Reports → AI Visibility** and open **Setup**. A project with no prompts yet opens on this card directly.
Type a subject in **Add a topic…** and click **Add topic**. If the project tracks keywords, you can click one of the chips under **Suggestions from your tracked keywords** instead.
Each topic appears as its own group, empty until you add prompts. Topic names are unique within a project; adding a name that already exists simply reopens that topic.
### 3. Generate or write prompts
In an empty topic, click **Generate 5 prompts**. In a topic that already has prompts, pick a number from 1 to 5 in the dropdown and click **Generate**. The prompts are written for this project from its brand names, competitors, domain, description and location, and are phrased to make an assistant name vendors — best, top, alternatives, comparisons. Prompts identical to ones already under the topic are skipped. A toast confirms **Generated N prompts**.
Type the question in **Add a prompt for "topic"** (3–500 characters) and click **Add**.
Every new prompt — generated or hand-written — starts enabled and is asked of all four assistants.
Prompt text cannot be edited in place. Remove the prompt with its trash icon and add the corrected version.
### 4. Choose which prompts run and which assistants are asked
Each prompt row on the Setup tab has an on/off switch, four assistant pills and a trash icon.
* **Switch** a prompt off to keep it without checking it; switch it back on whenever you like.
* **Click an assistant pill** (**ChatGPT**, **Claude**, **Gemini**, **Perplexity**) to add or drop that assistant for the prompt. A filled pill is asked; a faded one is skipped. Each prompt keeps at least one assistant.
* **Trash icon** on a prompt removes it; the trash icon on a topic header removes the topic and all its prompts. Past results stay in the history charts.
The topic header shows how many of its prompts are enabled, and a counter at the top of the card shows the number of prompts across your company.
### 5. Check the cost before anything runs
The blue notice at the top of **Prompts by topic** shows what your current prompts will cost:
* With a pinned schedule it reads **About N Credits/month for AI-visibility checks**, with the arithmetic underneath — for example "12 prompts across 4 assistants = 48 checks × 4 runs/month".
* Without one it reads **N Credits per check of these prompts**.
The estimate adds up each prompt's own assistant selection, so dropping an assistant from a prompt lowers it immediately. Fewer prompts, fewer assistants per prompt, or a slower schedule are the three levers. Very large prompt sets are trimmed per run, with prompts rotating so every one is covered across successive runs; the notice tells you when this applies.
## How checks run
**Weekly by default.** Once a project has at least one enabled prompt, Visibility schedules the first run shortly afterwards and then repeats weekly. You don't need to configure anything; the banner on the AI Visibility page always shows the concrete next and last run times.
**Run now.** To check immediately, or to retry a failed run:
In the project, go to **Settings → Schedules**. Find the **AI Visibility** card — it is **Active** when the project has at least one enabled prompt and shows **\~N Credits est. / run**.
Confirm in the dialog — **Run AI Visibility now? This runs immediately and will cost \~N Credits, debited from this company's balance right away.** The button shows **Running…** while the check runs, usually under a minute, and a toast reports whether it succeeded.
A red bar on the card shows the error with a **Retry** button.
A manual run is the same job as the scheduled one, so it also moves the next automatic run a week out. Only one AI Visibility run per project can be in flight at a time, and manual runs are rate-limited. There is no run button on the report page itself.
**Pinning a day and time.** Click **Edit** on the same card to choose when checks run, in UTC — for example **Every Monday at 09:00**. Weekly is the most frequent option; AI Visibility cannot run more often than once a week.
**History.** The **History** tab under **Settings → Schedules** lists every run with **Succeeded** or **Failed**, a **Scheduled** or **Manual** badge, its start time and the credits it cost.
### What a check costs
| Detail | How it works |
| --------------------- | --------------------------------------------------------------------------------------------- |
| Rate | **10 credits per prompt per assistant** — the current rate, shown in the app before you spend |
| When it is charged | Debited from the company's credit balance each time a run happens |
| Where to see it first | The Setup tab notice, and **\~N Credits est. / run** on the Schedules card |
| Not enough credits | The run is refused up front rather than partly charged |
Anyone with access to the project can start a run, and every run spends the company's credits. AI Visibility checks always run on Visibility's own data access and are billed in credits — a company's own DataForSEO key does not cover them. See [Credits and BYOK](/credits-and-byok) and [Billing and subscriptions](/billing-and-subscriptions).
Your agents can read this report while they work on tasks — for example to plan content around the questions you are missing from. Only the schedule and **Run now** start checks.
## Read the report
### Overview tab
Pick a window with the date filter first; every section below respects it.
| Section | What it means |
| ---------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Visibility** score ring | Of all answers in the window, the share in which an assistant named your brand or linked your site, from 0 to 100. The ring is red when the score is low, amber in the middle and green when it is high. |
| **By assistant** | A 0–100 score per assistant that blends how often your brand appears in that assistant's answer text with how often your site is among its citations. A dash means that assistant produced no answers in the window. |
| **Visibility over time** | One point per day that had a run; higher is more visible. It reads **Not enough data yet** until two or more days have runs. |
| **Brand presence by prompt** | A grid of every prompt against the four assistants: a green check where you appeared, a grey cross where you did not, and a dash where that assistant is not asked that prompt. This is the fastest way to see which questions you are missing from. |
| **Competitors mentioned** | How many times each competitor domain from your settings was named or cited. |
| **Top cited domains** | The sites the assistants link to most across all answers in the window — the places to earn a mention or a listing. |
Before the first run, the tab reads **No visibility data yet**.
### Topics tab
The Topics tab lists every topic with a visibility bar, its percentage and its prompt count. Click a topic to drill in; **All topics** takes you back.
The **Brand leaderboard** ranks every brand the assistants named for this topic. Your row carries a **You** badge.
* **Share of voice** — the percentage of the topic's answers that named the brand.
* **Avg position** — how early in the answer the brand tends to appear (1 = named first).
* **Citations** — distinct links to that brand's site.
Below the table, **By assistant** shows the per-assistant scores for this topic only.
One accordion per prompt, with a pill such as **3/4 mentioned** (or **Awaiting run** before the first check). Expand a prompt to see, per assistant, **Mentioned** or **Not mentioned** and how many sources it cited. Expand an assistant to read its full answer and its **Citations** list. Prompts are read-only here — add or generate them on the Setup tab.
The pages the assistants cited for this topic, ranked by **Mention rate**, with how many **Brands** each source carried. Sources that keep appearing are the pages worth being listed on.
The leaderboard, answers and sources reflect the latest run for that topic.
### The same numbers elsewhere
* The project **Overview** page has an **AI visibility** section with a stat strip (Visibility, Best assistant, Answers checked, Competitors seen), the ring, per-assistant bars, the trend, the top five competitors and cited domains, and an **AI visibility report** link.
* Each card on the [Projects](/projects) list shows an **AI vis.** stat with the latest run's percentage.
* Custom dashboards offer five widgets under the **AI Visibility** group in **Add widget**: **Visibility score**, **Visibility trend**, **Brand presence by prompt**, **Top cited domains** and **Competitors mentioned**. See [Dashboards](/dashboards).
* The **AI Citations** tab in [Opportunities](/opportunities) lists sources that cite a competitor but not you, built from your stored answers.
## How to read a score honestly
An answer counts when one of your brand names or your domain appears as a whole word in the text, or your site is among the citations. Matching is exact and reproducible. That makes the number trustworthy, but it also means the score is only as good as your brand-name list: a missing spelling undercounts you, and a generic name (store, app, design) overcounts you. Keep the list distinctive and complete.
Scores change only when a run happens — weekly by default — so the trend is a step chart, not a daily line. Judge a window that contains several runs rather than reacting to a single one.
With five prompts across four assistants, one changed answer moves the score by several points. Broader prompt sets give a steadier signal; the cost estimate on the Setup tab tells you what that costs.
A brand named tenth in a list counts as present just like a brand named first. Use **Avg position** and **Share of voice** on the Topics tab to see how prominently you feature, and read the actual answers under **Prompts**. The report does not measure sentiment.
An assistant can mention you without linking to you, or link to you without naming you. The per-assistant score blends both; **Citations** on the leaderboard and **Top cited domains** show the linking side on its own.
Adding, removing or disabling prompts changes what is being measured, so scores before and after are not directly comparable. Settle the prompt set first, then track. Removing a prompt keeps its past results in the history.
In the presence grid a dash means that assistant was not asked that prompt; under **By assistant** it means the assistant produced no answers in the window. Neither counts against you.
## Related
AI citation gaps — sources that cite competitors but not you.
Put AI Visibility widgets on a client-facing dashboard.
How credits are spent and what your own keys cover.
Tracked keywords feed the topic suggestions on the Setup tab.