AI visibility measures how often your brand and content are mentioned in AI-powered search results across platforms such as ChatGPT, Google AI Mode, and Perplexity. It helps you understand your visibility in the evolving AI search landscape and compare it to competitors.

Traditional SEO focuses on rankings on standard search engine results pages (SERPs) and on measuring clicks, impressions, and rankings. AI Visibility measures the visibility of AI-generated answers, enabling users to find information without clicking through to websites, and requires new metrics such as brand mentions and citation frequency.

The initial release includes ChatGPT, Google AI Mode, and Perplexity, which collectively account for the majority of the AI search market share. Additional platforms may be added in future releases based on market adoption and customer feedback.

Not currently, but possibly in the future.

We use vetted third-party web scraping partners to collect response data in ChatGPT and Google AI modes. This approach yields more accurate results than API-only solutions because it captures the full context, including system prompts and user memory that affect responses. For Perplexity, we currently use only the Perplexity API.

Data is updated weekly, which should be sufficient for meaningful trend analysis.

For ChatGPT, there are differences between the scraped and API content. We use vetted third-party scraped content for ChatGPT and Google AI Mode because it’s more accurate: it captures the full context, including system prompts and user memory, which influence responses.

This is not currently planned, though we may revisit this in the future.

We use LLM knowledge to:

Your confidential data is not used to train LLMs nor shared with other customers.

Scrapers cannot operate while logged in, as it would be too easy for Google to detect/block them. Google AI Mode is only available in the US, UK, and India when logged out. We are unable to run Google AI Mode campaigns in other countries because of this limitation.

No, not currently.

AI platforms do not provide first-party data; therefore, AI visibility is the best proxy available. Building prompts using your GSC data improves the relevance of the AI visibility results. See “Do we see any differences between the scraped content and API content?”

Strategic pages, as shown in the AI Discovery report, are those that generate 500 or more GSC impressions. You should focus on having these pages retrieved by LLMs.

Retrieval is when the ChatGPT-User bot visits a site during a user query.

Not currently, but it may be available in a future release. Currently, the data includes ChatGPT, Perplexity, and Google AI Mode (not AIO).

The data on the AI Discovery page shows a comparison of LLM activity vs. pages with impressions on Google SERPs from GSC. Strategic pages (i.e., those receiving 500+ impressions in GSC) should also see a high volume of activity from LLMs. When drill-down becomes available in these charts, you should first investigate the strategic pages that LLMs are not retrieving, followed by pages that received between 100 and 500 impressions.

The percentage of answers mentioning your brand aggregated across LLMs (e.g., when the brand is visible in a single ChatGPT prompt but not Perplexity, visibility = 50%).

The number of visits comes from your integrated logs, identified by referral data and ?utm_source=chatgpt.com entries (only available for ChatGPT).

By exact URL

An LLM performs sentiment analysis.

Not currently, but planned for a release in the near future.

Yes, we ask an LLM to check prompt relevancy. We are adding a moderation step for PAA questions after we saw ChatGPT flags that have caused workflow failures.

They should not capture or store prompts for future model training, as we are using licensed models under our license terms. Botify cannot warrant actions of third parties we do not control, but those entities are contractually required not to use the data. Refer to our Terms of Use for Botify GenAI Features for more information.

An answer is the combination of a prompt, a platform, date, and country, since a prompt can be run multiple times across each. The “Answers” page displays the list of prompts in the first column. For each run, we retrieve the overall response from ChatGPT, Perplexity, and Google AI Mode, which may vary. This means there may be duplicate prompts, but when considered with the platform and country, each is unique. You can click on a row for more details and see the platform’s full response.

Prompts are typically run within a few hours of campaign creation and after the first campaign edit in a week. Refer to Managing Campaign Updates for more information.

Prompts are scheduled to be run weekly on the day they were created. The next scheduled date is shown at the top of the Campaigns page. Refer to Managing Campaign Updates for more information on data updates.

We use the top keywords to generate prompts with People Also Ask questions; we do not get prompts directly from the keywords. We use the most recent GSC keywords for every new campaign and update them every three months to ensure the campaign remains accurate.

Since we do not have first-party data from AI search engines, we need to find effective directional prompts. "People also ask" shows popular queries from Google search, which are different from what you would ask ChatGPT, but still a good indicator of your visibility (i.e., the "directional" part).

It depends on how much the top keywords are changing over time:

We use an LLM to generate topics from the prompts when each campaign batch runs. We generate three topics per prompt. Some campaigns may have more topics than others if there is little similarity across prompts. Broader topics not explicitly mentioned may be used if the prompts strongly imply a general subject (e.g., “best budget laptops” → electronics).

These numbers are the links we retrieve from the response body to cited domains. We do not retrieve the links used as sources in the AI response.

The competitive analysis shows how your brand compares to competitors across AI platforms. You can:

An LLM analyzes the answers to detect brands. You can select your main competitors for SoV computation.

They are treated as a single brand to ensure an accurate AI Visibility report. Note that misspellings of competitor brand names are not currently consolidated.

Mentions refer to your brand appearing in the AI Search Result; citations refer to the client domain appearing in the AI Search Result.

An LLM determines if your brand is mentioned in the answers.

Not currently.

No, brand names are not case-sensitive in brand mentions.

PAA is used for top keywords, but not for categories. We use “People Also Search For” (PASF) first, then an LLM to generate category prompts.

We use PASF to identify related keywords, then an LLM to generate prompts based on those keywords.

An LLM evaluates intent as follows:

There is no average campaign size. Prompts are distributed equally across all campaigns by default. You can adjust this distribution using sliders to increase or decrease the percentage allocation for each campaign.

They are based on the last 60 days.

The number of keywords is dynamic and adaptive. We fetch as many keywords as necessary to generate the configured number of PAA questions.

60 days.

The country flag is the country where we asked the question to the LLM. A prompt in the same language but in different countries will not produce the same answer, depending on your location. Since we do not specify the language when we submit the question, the LLM may answer in the same language as the question or in the language of the original keyword or category. This can produce unexpected results by returning answers in a language not native to a country (e.g., Italian in France).

It is a snapshot of the top-performing keywords by clicks over the last 60 days.

You can use segmentation-style groupings with AI Visibility, but it’s done through Campaigns rather than the “segments” filters you use in Botify Analytics reports. In AI Visibility, you track and filter results by campaign, and each campaign can be built from different prompt/keyword sources, including your website categories.