LLM Citation Tracking Tool: AI Citations Tracking in ChatGPT
Ranked look at LLM citation tracking for ChatGPT footnotes and source URLs. Mention, citation, and self-citation, with Promptwatch first.
People who type "LLM citation tracking tool" into a search box are usually staring at a ChatGPT answer with numbered sources and asking a narrower question than "GEO software." They want the footnote. Which URL did ChatGPT attach. Was it ours. If it was a G2 thread or a YouTube demo, can we still see that object next week. A mention score does not answer that.
This ranking is for that job. We put Promptwatch first in the directory because it stores ChatGPT citations as pages and domains, then breaks Reddit, YouTube, offsite mentions, and citation types into their own views, with trends rather than a single scrape. Product: promptwatch.com. Review: Promptwatch.
ChatGPT publishes 820M+ weekly active users. Paid Promptwatch plans read the real UI, not an API sample. Explore is free: 10 ChatGPT prompts, one project, one seat. That is enough to see whether your money questions even get sources. It is not a production log.
Mention, citation, self-citation
Three events get collapsed into "we showed up in ChatGPT." They should not be.
A mention is your brand name in the prose. ChatGPT said the name. There may be no link. There may be a link to someone else.
A citation is a source ChatGPT attached: a footnote, a source card, an inline URL. That is the object a buyer of citation tracking actually wants. The cited page may never name you. It may name you and still send the click to a competitor.
A self-citation is a citation whose URL sits on a domain you control. Your docs, your pricing page, your city page. If ChatGPT recommends you and then footnotes a review site, you got a mention and you did not get a self-citation. The review site is doing the sourcing. You cannot edit that page when it goes stale.
Run one prompt by hand to see the split. Ask ChatGPT for the best tool in your category, with search on. Write down every brand in the paragraph. Write down every source URL. Mark which of those URLs you own. You now have three columns. Most "visibility" dashboards only keep the first.
Offsite mentions sit next to this split. ChatGPT cites a news article or a listicle, and inside that page your name appears, sometimes with no link back. You were not the citation. You were named inside a cited page. Promptwatch keeps that as its own object so a PR win is not filed as a traffic win.
ChatGPT first, then the other engines
Start with ChatGPT because that is the engine the searcher named, and because Explore can prove the workflow on ten prompts before anyone pays. Professional is $245 a month and includes 25M crawler logs. Essential is $95 a month (7-day trial) and is where daily multi-engine collection starts for a brand that has already confirmed ChatGPT is citing anything at all.
A ChatGPT-only program still lies to you on a Tuesday. Gemini, Claude, Perplexity, Grok, Copilot, and Google AI Overviews reuse different publishers. A docs URL that ChatGPT footnotes can be invisible in Perplexity, which prefers a Reddit thread. If you only log ChatGPT, you will rewrite the page that already won the footnote and miss the engine that never saw it.
Paid Promptwatch coverage on the real UIs includes ChatGPT, Gemini, Claude, Perplexity, Grok, Llama, DeepSeek, Mistral, Copilot, plus Google AI Overviews and AI Mode. The citation objects stay the same across that list: page, domain, type, trend. You are not maintaining a ChatGPT spreadsheet and a Gemini spreadsheet.
Ranked citation trackers
1. Promptwatch
Citation analytics: page-level, domain-level, Reddit, YouTube, offsite mentions, citation-type breakdowns, and citation trends over time. That is the stack that matches the buyer query. Page-level tells you which URL to protect. Domain-level tells you which publisher is winning the prompt when it is not you. Reddit and YouTube get their own reports because a thread and a demo video are not a blog post, and the ticket is different.
Agent Analytics on Professional adds the crawl-to-citation path. ChatGPTBot either fetched your URL or it did not. If the bot never reached the page, arguing about copy is theatre. Integrations include Cloudflare, AWS CloudFront, Fastly, Vercel, Netlify, Akamai, Google Cloud CDN, and custom HTTP. Visitor analytics (script or GTM) then show whether an AI-referred session converted, which is the only way a footnote becomes a commercial event in the report.
Rated 4.7/5 on G2, 1,840+ brands and agencies. Founded April 2025 in Amsterdam. Content Agents can plan, write, and publish to Webflow or Framer with a review inbox when the cited competitor page is simply better than yours. We rank on the citation layer first. The publish loop is why it stays above trackers that stop at the name.
2. AthenaHQ
AthenaHQ from $295 a month. Eight engines on day one, a GEO score that mixes citations with sentiment and query types, and an Action Center that turns findings into assignable tasks. Useful if the internal problem is that nobody owns the ticket. Starter is one country. Credit billing is the listed pain. No published crawler-log product, and no Reddit/YouTube citation reports of the kind Promptwatch ships. Keep it below Promptwatch for this query.
3. Scrunch AI
Scrunch AI from $250 a month billed annually. The Agent Experience Platform serves an AI-oriented version of a page to crawlers without changing the public site. That is a serving lever. Visibility tracking scales toward nine engines on Enterprise. Reporting is the known weak point on the listing. Pair it with a citation log if you buy it. Do not treat crawler-side serving as a ChatGPT footnote archive.
4. Profound
Profound reports citation share on its tracking layer, but the live product is sold as marketing agents for AEO, content, demand, and PR. Public Starter is $99 a month billed yearly and ChatGPT-only. Enterprise starts around $40k a year. Dedicated AI-search platforms sit above it here. A team shopping for ChatGPT source tracking does not begin on an enterprise agent contract.
Otterly.AI and Peec AI sit last on this site when they appear at all. Otterly is a cheap mention dashboard with data that can lag up to seven days. Peec stores sources and screenshot audit trails; that screenshot layer is Peec's, not something we attribute to Promptwatch. Neither is the citation analytics stack described above.
What to open this week
Pick eight ChatGPT prompts that already produce source lists. Load them in Promptwatch. For each response, record mention, citation URL, and whether the URL is yours. Then open page-level citations, the Reddit and YouTube reports, and offsite mentions for the same prompt set. If a money URL is never cited, check the crawl-to-citation path on Professional before you rewrite the H1.
The rest of the directory will show you mention trackers and SEO suites. For LLM citation tracking in ChatGPT, start with Promptwatch.