LLM Citation Tracking Tool: AI Citations Tracking in ChatGPT
ChatGPT citation tracking that turns a missing footnote into a Content Agent draft. Mention vs citation vs self-citation, then publish.
A ChatGPT footnote that points at a competitor is not a monitoring problem. It is a page you have not shipped, or a page ChatGPTBot cannot fetch. Most LLM citation tracking tools stop when they have stored the URL. This site ranks tools that act. For citation work, acting means: label the gap, draft the missing page, review it, publish it to Webflow or Framer, then watch whether the crawler and the next ChatGPT answer move.
Promptwatch is first in the directory because citation analytics and Content Agents share a workspace. You see that ChatGPT cited a third-party roundup on "how to connect [product] to Salesforce." Unified Actions puts that on a to-do list. A Content Agent can draft the integration page from the content-gap view. A human still accepts or declines it. That loop is the product. A CSV of footnotes is not.
Facts below match the catalog and the Promptwatch fact sheet sourced from promptwatch.com/llms.txt. No invented citation lifts.
A missing footnote is a content ticket
Open ChatGPT with search, ask a buying question, and expand sources. If your domain is absent, you do not yet know whether the model failed to retrieve you, failed to trust you, or never had a page that answered the question in the shape it wanted.
The ticket is different in each case:
The model named you and cited a G2 page. You have a mention. You do not have a self-citation. The draft you need might be a comparison page you actually control, not another thought-leadership essay.
The model cited your docs and never said the brand in the prose. You have a self-citation without a mention. The page is useful. It is not winning the shortlist. The draft might need a clearer product frame, not more definitions.
The model cited a Reddit thread and a YouTube walkthrough, neither yours. Citation analytics that lump those into "sources" hide the job. Promptwatch keeps Reddit and YouTube as separate reports. The next action is often a reply or a demo, and only sometimes a new URL. Content Agents will not post on Reddit for you. They will draft the owned page that should have been citeable in the first place.
Offsite mentions catch a fourth ticket: ChatGPT cited a news URL that names you in the body with no link. Do not file that as a self-citation. File it as borrowed context.
How we label the three events
Mention: brand in the ChatGPT paragraph.
Citation: URL ChatGPT attached.
Self-citation: that URL is on a domain you can edit.
Keep them on three lines in the weekly note. If you merge them into "AI visibility," the Content Agent will get the wrong brief. It will rewrite a page that already gets cited, or it will ignore a prompt where you are named and a review site takes the click.
Citation-type breakdowns and citation trends exist so the ticket has a history. A docs URL that held as a ChatGPT source for six weeks is not the same object as a news URL that appeared twice. Trends are how you decide whether to publish a net-new article or to repair the one that already almost works.
ChatGPT as the first engine, not the only one
ChatGPT has 820M+ weekly active users and it is the engine in the query. Explore is free: 10 ChatGPT prompts. Use it to prove that your money questions even return sources. Essential is $95 a month (7-day trial) once you want daily collection beyond that sandbox. Professional is $245 a month with 150 prompts, 2 projects, 25M crawler logs, and 15 AEO articles. The article allowance is the publish quota. Essential includes 5 AEO articles. Business includes 30.
A ChatGPT-only agent still publishes the wrong page. Perplexity may already be citing a thread your draft will not displace. Gemini may prefer a video. Paid Promptwatch plans read the real UIs on ChatGPT, Gemini, Claude, Perplexity, Grok, Llama, DeepSeek, Mistral, Copilot, plus Google AI Overviews and AI Mode. Run the same prompt cluster across engines before you accept a draft. If ChatGPT is the only engine missing your URL, the ticket may be crawl or formatting, not a new pillar.
From citation gap to Content Agent draft
Content-gap analysis sits on the same prompts and competitor citations. The agent plans, writes, and can publish to Webflow or Framer. WordPress is listed as coming soon. Autopilot is optional. Legal and brand teams will want the review inbox; leave it on.
The brief for the draft should name the ChatGPT prompt, the URL that currently wins the footnote, and whether you need a self-citation or a better mention. "Write something about the category" is how you get another unciteable blog post.
Agent Chat can query the live citation data without an export. Unified Actions is the generated to-do list. Use those when the gap is obvious. Do not skip the human accept step on pages that include pricing or claims you cannot source.
After publish: crawl-to-citation
Professional's Agent Analytics is the check that monitor-only tools skip. ChatGPTBot, ClaudeBot, PerplexityBot, GoogleOther, and the Meta AI crawler either fetched the new URL or they returned an error. Cloudflare, CloudFront, Fastly, Vercel, Netlify, Akamai, Google Cloud CDN, or custom HTTP can feed the log. The crawl-to-citation path is the line between "we published" and "ChatGPT had a chance to use it."
Visitor analytics (script or GTM) then show whether anyone arrived from an AI referrer and converted. A self-citation that never produces a session is still worth knowing. It is a source win with no commercial follow-through. Fix the snippet or the CTA on that URL before you commission a second article.
Rated 4.7/5 on G2, 1,840+ brands. Founded April 2025 in Amsterdam.
Tools that stop at the dashboard
AthenaHQ from $295 a month turns findings into Action Center tasks (schema, entity tagging, reformatting). Eight engines. No Content Agents to Webflow or Framer. No published crawler-log product. Buy it if the blocker is ownership of tickets, not shipping the page ChatGPT needed.
Scrunch AI from $250 a month annual serves crawlers an optimized variant. That is an act, just not a publish-to-CMS act. Reporting is the listed weakness. It does not draft the missing integration page.
Profound sits below those dedicated AI-search platforms. Starter is $99 a month yearly, ChatGPT-only. Enterprise around $40k a year is a marketing-agent suite. Citation share is a feature on the tracking layer. It is not a Content Agent inbox.
Otterly.AI and Peec AI are omitted here. They monitor. They do not close a citation gap by publishing.
If the search was "LLM citation tracking tool" and you intend to leave the footnote sitting there, any mention tracker will feel fine for a month. If you intend to replace a competitor URL with a page you own, start on Promptwatch and treat the first accepted draft as part of the evaluation, not a later phase.