To monitor brand references in ChatGPT live, first settle what "live" can mean. None of the seven ChatGPT monitoring tools we read claims to see what ChatGPT tells other people. Every one of them works by asking ChatGPT itself: you choose the prompts, the tool runs them on a schedule, and it records whether your brand is named and which pages are cited. The fastest schedule any of them states is daily. So real-time monitoring of ChatGPT is really three signals used together, and each one has a blind spot:
- Scheduled prompt checks: what ChatGPT answered a monitoring tool, for prompts you chose, at best once a day.
- ChatGPT-User visits in your server logs: the moment a user's question makes ChatGPT open one of your pages. Of the three, this is the only one that happens while ChatGPT is answering.
- Referrals tagged
utm_source=chatgpt.com: a person clicked a link to your site in a ChatGPT answer.
Every vendor statement below was read on that vendor's own website, and every OpenAI statement in OpenAI's own documentation, on 3 November 2026. Each is linked where it is used. We have not run these tools side by side, so this compares what each vendor says about itself; it is not a test of accuracy. Plans and features change often, so we have left prices out. Check each vendor's pricing page before you buy.
TopCited is a GEO platform with SEO included. Its own monitoring also runs on a schedule, none of the seven trackers here is ours, and several of them compete with us. We report what each says about itself and set out where TopCited fits at the end. For the wider guide to tracking ChatGPT mentions over time, see monitoring brand mentions in ChatGPT. For a comparison of ChatGPT trackers, see rank tracking tools for ChatGPT.
Key definitions
| Term | What it means here |
|---|---|
| Brand reference | ChatGPT naming your brand in an answer, or citing one of your pages as a source. Most tools report the two separately, as mentions and citations. |
| Mention and citation | Peec AI's definitions are typical: "A brand mention happens when an AI answer directly names your brand. A source citation happens when your website or content is used to generate an answer - even if your brand isn't named." |
| Prompt check | One run of one prompt against ChatGPT by a monitoring tool, with the answer recorded. |
| Cadence | How often a tool repeats each prompt check. |
| ChatGPT-User | The user agent OpenAI says ChatGPT uses when, while answering a user's question, it visits a web page. |
| OAI-SearchBot | OpenAI's search crawler. OpenAI says it "is used to surface websites in search results in ChatGPT's search features." |
utm_source=chatgpt.com | The tag carried by links in ChatGPT answers, which lets analytics tools label a visit as coming from ChatGPT. |
The mention and citation definitions are from Peec AI's FAQ. The ChatGPT-User and OAI-SearchBot definitions are from OpenAI's overview of OpenAI crawlers.
The three signals compared
| Signal | What it shows | How fast | Where it goes blind |
|---|---|---|---|
| Scheduled prompt checks | What ChatGPT answered the tool, for the prompts you chose, from the location the tool used | Daily at best in the seven trackers we read | It is the tool's answer, not any real user's, and each check is one answer |
| ChatGPT-User visits in server logs | A user's question made ChatGPT open one of your pages | As it happens | Silent when ChatGPT names you on the strength of another site's page, or answers without opening a page at all |
| Referrals tagged utm_source=chatgpt.com | A person clicked through to you from a ChatGPT answer | When the click happens | Counts clicks only, and for several days in October 2026 the tag went missing and visits were filed as Direct |
No single signal answers the question on its own. A prompt tracker tells you what ChatGPT says, but not to whom. Your logs and analytics tell you that real conversations reached you, but not what was said. Use all three, and trust a change when they move together.
Signal 1: scheduled prompt checks
This is what most ChatGPT monitoring tools sell. Each sends prompts to ChatGPT on a schedule and records the answer. When a vendor says "live" or "real-time", check what the word is attached to:
- Rankability's "live" is its query method. It says its tracker "sends your keywords to ChatGPT with live web search enabled — the exact same way a real user would search" (Rankability). The scans themselves run on a schedule: "Pick daily, weekly, or a custom cadence."
- Peec AI's "real time" is its data feed. Its Looker Studio connector, REST API and MCP integration "stream live visibility, source tracking, and multi-project metrics in real time". The checks behind them run once a day: "Peec AI executes each prompt once every 24 hours on every AI model you've selected" (Peec AI).
- Ahrefs ties "real-time" to a schedule too. Its Brand Radar page says: "Track prompts daily, weekly, or monthly to stay in the loop with real-time AI changes" (Ahrefs).
None of this is a criticism of the tools. It is how every tool we read works, and it means a "live" ChatGPT tracker gives you a daily sample at best.
The seven prompt trackers compared
| Tracker | How it checks ChatGPT | Fastest cadence it states | What it says it records |
|---|---|---|---|
| Semrush Prompt Tracking | Prompts you add, with a location and language per campaign | Daily | AI Visibility, Mentions, Owned Sources, Average Position, and a snapshot of the actual ChatGPT Search result |
| Ahrefs Brand Radar | Custom Prompts you define; also a pre-built AI Visibility Index | Daily for Custom Prompts; the Index is re-tested monthly | Mentions, citations, AI share of voice and sentiment |
| OtterlyAI | Search prompts on ChatGPT and three other AI search engines | Daily | Brand visibility and sentiment, and citation sources |
| Peec AI | Each prompt on every AI model you select | Once every 24 hours | Brand mentions and source citations, filterable by model, country and prompt tag |
| Nightwatch | Queries about your industry on ChatGPT, Claude, Gemini and Perplexity | Daily | Position, visibility score, growth trends and the URLs cited |
| Rankability | Your keywords, sent to ChatGPT with web search enabled | Daily, weekly or a custom cadence | Mentions, citations, your position among the cited sources, up to 25 brands per response, and a snippet of the response |
| Keyword.com | Scrapes ChatGPT's web interface rather than using the API | Not stated on its ChatGPT page, which says it queries ChatGPT regularly | Brand mentions, sentiment scores, cited links, and the full scraped response |
Semrush Prompt Tracking
Semrush's knowledge base says Prompt Tracking "lets you monitor your brand's daily visibility for a custom set of prompts across AI search platforms, including ChatGPT, Google AI Mode, and Gemini". It needs a subscription to the AI Visibility Toolkit (Semrush Knowledge Base). You set a location and language for each campaign. In ChatGPT Search, it "ranks each domain in the citation area from top to bottom". A View SERP pop-up shows "the actual ChatGPT Search, Google AI Mode, or Gemini result for your prompt", so you can read the answer itself.
Ahrefs Brand Radar
Brand Radar sells two things (Ahrefs). Custom Prompts are the questions you define, and Ahrefs says it "runs them on demand as often as daily". You pick the cadence per AI platform. The AI Visibility Index is built from prompts that Ahrefs derives from its keyword database, runs through each AI platform and stores, and its "question sets are re-tested monthly on a 90-day reporting window". For live monitoring the Custom Prompts are the relevant part. The Index is a benchmark.
OtterlyAI
OtterlyAI's pricing page lists "Daily tracking frequency" on every plan it prices, with tracking of ChatGPT, Google AI Overviews, Perplexity and MS Copilot, and Claude, Google AI Mode and Gemini as add-ons. Each plan also lists brand visibility and sentiment analysis and citation source tracking (OtterlyAI pricing). Its Standard and Premium plans also list Agent Analytics (see signal 2).
Peec AI
Peec AI runs each prompt once every 24 hours on every model you choose. It says it "tracks both “used” (your content informed the AI's answer) and “cited” (your URL is explicitly mentioned)", and you can filter by model, country and prompt tag (Peec AI).
Nightwatch
Nightwatch says its AI tracking monitors "how ChatGPT, Claude, Gemini, and Perplexity respond to queries about your industry", and that "AI visibility data is updated daily, alongside your regular rank tracking data" (Nightwatch). It puts AI visibility next to your Google rankings in the same dashboard.
Rankability
Rankability's ChatGPT tracker records "your exact position among all cited sources for every keyword you track", and says "every scan captures up to 25 brands mentioned in the same ChatGPT response" (Rankability). Each scan keeps "a snippet of ChatGPT's actual response". It also accounts for location: "A plumber in Austin gets a different ChatGPT answer than a plumber in Chicago."
Keyword.com
Keyword.com says it tracks ChatGPT "using the scraping method", to capture "the exact, real-world experience and responses that actual users see when interacting with the web interface, rather than relying on the API which can sometimes output different results" (Keyword.com). You can "access the full scraped result of every tracked ChatGPT response". It is also open about which model it sees: "ChatGPT-5 is used most of the time, but not always — the specific model depends on availability and system conditions at the time the prompt is tracked."
Why a daily check is not what your customers saw
The vendors themselves give three reasons not to read one check as the answer everyone gets:
- Location. Rankability's plumber example above, and Semrush's per-campaign location and language setting.
- Model. Keyword.com says the model behind an answer depends on "availability and system conditions at the time the prompt is tracked."
- Interface. Keyword.com scrapes the web interface because the API "can sometimes output different results."
So treat each check as one sample. Read trends across many prompts and many days, not a single answer on a single morning.
Signal 2: ChatGPT-User and OAI-SearchBot in your server logs
OpenAI's overview of its crawlers describes three of the user agents you will see in your logs:
- ChatGPT-User: "When users ask ChatGPT or a CustomGPT a question, it may visit a web page with a ChatGPT-User agent." OpenAI adds that "ChatGPT-User is not used for crawling the web in an automatic fashion" and that "because these actions are initiated by a user, robots.txt rules may not apply."
- OAI-SearchBot: "OAI-SearchBot is for search." OpenAI says "Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though can still appear as navigational links."
- GPTBot: it crawls "content that may be used in training our generative AI foundation models." OpenAI says each setting "is independent of the others", so you can allow OAI-SearchBot and disallow GPTBot.
A ChatGPT-User request is the nearest thing to a live signal you can get. It means that, at that moment, a conversation led ChatGPT to open that page. Read it with three limits in mind:
- It is not every answer. OpenAI says ChatGPT may visit a page. When an answer names you on the strength of another site's page, nothing reaches your server.
- It shows that your page was opened, not what was said. The visit does not tell you whether the answer named you, or how.
- A user-agent string is only text the client sends. OpenAI publishes the IP addresses for each agent (ChatGPT-User, OAI-SearchBot, GPTBot). Match the IP against those lists before you count a visit.
An OAI-SearchBot visit means something else: your page is being crawled for ChatGPT's search features. It tells you that you are eligible to appear, not that you were mentioned.
You do not have to read raw logs yourself. Ahrefs offers Bot Analytics, which it describes as "Watch AI crawlers in real time: when they read your pages and which ones they visit most", free while in beta (Ahrefs). OtterlyAI lists Agent Analytics on its Standard and Premium plans (OtterlyAI pricing). Profound describes its Agent Analytics as a way to "Understand how bots view, cite, and refer to your pages" (Profound).
Signal 3: referrals tagged utm_source=chatgpt.com
A visit from a link in a ChatGPT answer normally arrives tagged utm_source=chatgpt.com.
- The tag has been widened before. On 16 June 2025, Search Engine Roundtable reported that "OpenAI has updated the ChatGPT interface to add additional UTM parameters to help with tracking more clicks from ChatGPT to your website". The tag now covered the links under "more" sources as well as the citations (Search Engine Roundtable).
- It went missing in October 2026. On 14 October 2026, Search Engine Roundtable reported that the links had stopped carrying it. It quoted Cedric Huesler of Adobe: "There's no tag and no referrer, so your analytics files the visit under direct." The report was updated on 15 October: Huesler had revised his post "to say this was a bug", and wrote that "ChatGPT is tagging its links with utm_source=chatgpt(.)com again" (Search Engine Roundtable).
A tagged referral is real-time evidence that a person acted on an answer. It counts clicks only, though, and a sudden drop in tagged visits can be a labelling change rather than fewer mentions. Before you conclude anything from a drop, check whether Direct visits to the same pages rose in the same days.
How to set up live ChatGPT brand monitoring, step by step
- Make sure ChatGPT's search can reach you. Allow OAI-SearchBot in robots.txt. OpenAI says that after a robots.txt update "it can take ~24 hours" for its systems to adjust.
- Separate OpenAI's agents in your server logs. Filter on ChatGPT-User and OAI-SearchBot, and check each IP against OpenAI's published lists.
- Alert on ChatGPT-User visits to the pages you care about, using your own log tooling or one of the agent analytics products above.
- Make a ChatGPT segment in your analytics for visits tagged
utm_source=chatgpt.com. - Write the prompts your buyers ask. Include your brand name, product comparisons and "best" questions in your category, not only the queries you already rank for.
- Run them daily in a prompt tracker, with the location and language set to match your customers.
- Add the competitors that appear in the same answers. The sources ChatGPT cites in your place are the pages to study.
- Read the stored answer before you react. Semrush keeps a snapshot, Keyword.com the full scraped response, and Rankability a snippet.
- Review weekly, across all three signals. A change in the tracker that also shows in ChatGPT-User visits and in tagged referrals is real. A change in one signal alone may be noise.
- When tagged referrals fall suddenly, check Direct first. In October 2026 a tagging bug moved ChatGPT visits into Direct for several days (signal 3).
Mistakes to avoid
- Reading one answer as the answer. A check is one sample from one location and one model.
- Reading "live" in vendor copy as continuous monitoring. Check whether it describes the query method, the data feed or the schedule.
- Blocking the wrong bot. OAI-SearchBot decides whether you can appear in ChatGPT search answers, and GPTBot is for training. They are set separately.
- Counting unverified user agents. Anyone can send a request that calls itself ChatGPT-User. Check the IP.
- Reading a referral drop as a mention drop. It can be a tag that went missing.
Where TopCited fits
TopCited is a GEO platform with SEO included, built to produce the work rather than only the report. It is not a live feed of ChatGPT either.
- Scheduled AI answer monitoring. TopCited's content monitoring runs your query sets on a schedule you set (cron, with timezone support). It records the brand mentions in the answers, and reports gap analysis and share of voice.
- Rewrites, not issue lists. The CORE optimizer rewrites your existing content and scores it before and after, including keyword coverage. So the page you want ChatGPT to open is the one that gets improved.
- A starting point. To see where you stand, request a GEO audit through the form on our site.
Keep your server logs and your analytics as the first-party record of what actually reached you, and check every tracker against them.
Frequently asked questions
Only partly. None of the seven ChatGPT monitoring tools we read on 3 November 2026 claims to see what ChatGPT tells other people; each asks ChatGPT itself on a schedule, and the fastest cadence any of them states is daily. What does happen live is a ChatGPT-User visit in your server logs, when a user's question makes ChatGPT open one of your pages, and a referral tagged utm_source=chatgpt.com, when someone clicks through. Use all three together.
Of the trackers we read, Semrush Prompt Tracking, Ahrefs Brand Radar Custom Prompts, OtterlyAI and Nightwatch state daily checks, Peec AI runs each prompt once every 24 hours, Rankability offers daily, weekly or a custom cadence, and Keyword.com's ChatGPT page does not state a frequency. None states a cadence faster than daily.
Look for the ChatGPT-User agent in your server logs. OpenAI says that when users ask ChatGPT or a custom GPT a question, it may visit a web page with that agent, and that it is not used for automatic crawling. Check the IP against OpenAI's published list before counting the visit, because anyone can send that user-agent string.
ChatGPT-User visits a page during a user's conversation. OAI-SearchBot crawls for ChatGPT's search features, and OpenAI says sites opted out of it will not be shown in ChatGPT search answers. GPTBot crawls content that may be used to train OpenAI's foundation models. OpenAI says each robots.txt setting is independent of the others.
Segment the visits tagged utm_source=chatgpt.com in your analytics. Watch Direct too: Search Engine Roundtable reported on 14 October 2026 that links in ChatGPT answers had stopped carrying the tag, so visits were filed as Direct, and updated its report on 15 October to say the cause was a bug and the tagging had returned.
Answers depend on where and how ChatGPT is asked. Rankability says a plumber in Austin gets a different answer than a plumber in Chicago, Keyword.com says the model behind an answer depends on availability and system conditions, and it scrapes the web interface because the API can output different results. Treat every check as one sample and read the trend.
Not according to OpenAI. GPTBot is for training, and OAI-SearchBot is the one used to surface websites in ChatGPT's search features. OpenAI's example is a site that allows OAI-SearchBot to appear in search results while disallowing GPTBot. It says a robots.txt change can take about 24 hours to apply to search results.
No. TopCited is a GEO platform with SEO included. Its content monitoring runs your query sets on a cron schedule with timezone support and reports brand mentions, gap analysis and share of voice, and its CORE optimizer rewrites existing content with before and after scores. Use your server logs and analytics for the live signals.