How do you monitor whether AI assistants cite your company?

Monitoring AI citations means running the same set of questions through AI assistants every month and logging whether the company appears, in what position, with which source and who appears in its place. The first round, run before any change to the site, is the baseline: without it there is no way to prove progress. This guide covers what to measure, how to set up the round and what to do with the result.

What is AI citation monitoring?

It is the periodic tracking of a brand's presence in AI-generated answers. The unit of measurement is the target question, the sentence the client would type. The main metric is the citation rate: the share of those questions in which the brand is cited.

The question set stays frozen from one round to the next, so the comparison is fair. The concepts are in the target prompt, citation rate and AI share of voice entries.

What should you measure in each round?

Six metrics cover the essentials: citation rate, AI share of voice, coverage by engine, accuracy and sentiment, AI referral traffic and crawler access. The first four come from the question round; the last two come from the site's own data.

Source: ReBo AI Visibility measurement protocol, published at /en/metrics/.

What each metric proves, and what it does not, is in how AI Visibility is measured.

How do you monitor, step by step?

In five steps: define the target questions and freeze them, log the baseline before changing the site, repeat the round every month with the same criterion, cross-check the result with the site's own data and turn the gaps into an editorial agenda.

Step 1 · Define and freeze the target questions

Write the questions in the client's vocabulary, not the company's. Four types cover the essentials: the problem question, the choice question ("who does this in my market?"), the selection-with-criteria question ("list three companies that...") and the brand question, the only one that includes the company name. Once defined, the questions are frozen and versioned.

Step 2 · Log the baseline before changing the site

Run the questions in ChatGPT, Gemini, Claude, Perplexity and Copilot, in a private session and with search turned on when the engine offers it. For each run, log the date, the engine, whether the company appears, in what position, which sources the answer shows and who appears in its place. The first-round protocol is in how to test, in 20 minutes, whether AI assistants cite your company.

Step 3 · Repeat the round every month with the same criterion

AI answers vary between runs, so a single round is indicative and the monthly series shows the trend. The citation criterion has to be fixed: on ReBo's ruler, the company name written in the answer or a link to its domain counts; a generic mention of the industry does not.

Step 4 · Cross-check the result with the site's own data

Three free sources complete the question round. On 2026-06-03 Google announced generative AI reports in Search Console, with impressions and pages in AI Overviews and AI Mode, no click data, and rolling out first to a subset of sites. The AI Performance report in Bing Webmaster Tools shows Copilot citations, the queries used to retrieve sources and the pages cited.

The third source is the site itself. Analytics shows visits arriving from chatgpt.com, perplexity.ai and other assistants, although some of them show up as direct traffic. Server logs show when GPTBot, OAI-SearchBot, ClaudeBot and PerplexityBot visited each page. The AI crawler entry explains each bot.

Step 5 · Turn the gaps into an editorial agenda

Every question where a competitor appears and the company does not is a citation gap, and becomes the brief for a new page or a rework of an existing one. Every wrong fact the AI repeats points to a source to fix, on the site or outside it. Without this step, monitoring becomes a report that only describes.

Spreadsheet, tool or service: how do you choose the way to monitor?

It depends on the size of the question set and on what you want to do with the result. A spreadsheet is enough for the first round; a tool saves time with dozens of questions across several engines; a service makes sense when measurement has to turn into action every month.

Source: ReBo AI Visibility measurement protocol, October 2026.

What the global SEO suites state about their own modules is in the guide to monitoring tools. At ReBo, continuous monitoring is Workstream 06 of the service: the same round every month, compared with the baseline, with the gaps turned into the next cycle's agenda.

How often should you monitor citations?

Once a month, with the same questions, on the same engines and with the same criterion. A higher frequency raises the cost without changing the reading, because a new page takes time to be crawled; a lower frequency misses the effect of each publication.

An extra round makes sense after a big change, such as unblocking a crawler or publishing a whole archive. The step-by-step plan to improve the result is in how to get your company to appear in ChatGPT and other AI assistants.

Keep reading

For the first round, use the 20-minute test. For the full protocol, see how AI Visibility is measured. To see what level the company is at, see the AI Visibility framework.