Best platform to monitor citations in ChatGPT and Perplexity: how we do it for clients

Clients tend to say "are we mentioned in ChatGPT?" when they mean "are we being cited?". The difference decides the work. A mention is the engine naming a brand. A citation is the engine pointing to a page. For ChatGPT search and Perplexity, citations are the thing that sends traffic, and they tell you which sources the engine trusts for a topic.

This is how we monitor citations for clients, and why the platform we run it on is Promptwatch.

Why a mention tracker is not enough

A tool that only reports whether your brand name appeared leaves you guessing about three things:

Which of your pages get cited, and for which prompts.
Which other domains the engine cites in the same answers.
Whether the engine can even fetch your page.

Each of those leads to a different piece of work. Page-level citations point to content. Third-party domains point to PR and community work. Fetch problems point to engineering. Without the data, all three look like "we need more content".

What we monitor, in order

1. Citations by page and domain. We start with Promptwatch's citation analytics: page- and domain-level citations with URL positions for the prompt panel. This tells us how many answers cite the client at all, and how many cite competitors and third parties instead.

2. Third-party sources. The Reddit, YouTube and LinkedIn citation reports show the community and video sources behind answers. Offsite mentions show where the brand is named on third-party pages without a link, which are often the best outreach targets.

3. Specific URLs. Page Tracker follows individual pages, ours or a competitor's. If a client updates a guide, we watch whether citations follow.

4. Citation patterns over time. Citation Trends classifies each citation by content type and source type, and reports per page when it was first cited, how long it took to ramp up, its peak citation rate and decay half-life. That tells us which pages hold their position and which fade, and shapes the refresh schedule.

5. Crawl access. Agent Analytics shows real-time, IP-verified AI crawler requests, including ChatGPT-User, OAI-SearchBot, GPTBot, PerplexityBot and Perplexity-User, with errors and failure explanations. Crawl-to-citation paths show a citation rate per page. A page that is crawled often and never cited is a content question. A page that is never crawled is a technical one.

6. Traffic. Visitor analytics attribute AI-referred visits and conversions by engine. This is the number the client's finance team cares about, and it is what keeps a retainer alive.

ChatGPT and Perplexity are not the same

We keep the two engines separate in reports. They draw on different sources and cite at different rates, so a combined number hides both. Promptwatch monitors ChatGPT (Search) and Perplexity through real-interface monitoring, and runs panels daily per model, country and language. We also use query fan-outs per model, which show the sub-queries an engine runs before answering. When a client is missing from an answer, the fan-out often shows the angle their pages do not cover.

Setting it up for a new client

Build 30 to 50 buyer prompts across organic, brand and competitor-comparison types.
Track both engines from day one and let two weeks pass before drawing conclusions.
Connect crawler logs early. Promptwatch supports Cloudflare, AWS CloudFront, Fastly, Vercel, Netlify and others, plus a WordPress plugin. Crawler logs begin on the Professional plan, so check the plan before promising this in a scope.
Install the visitor script, GTM template or WordPress plugin so AI-referred traffic is recorded from the start.
Review the Action Board weekly. Its Discovery agent suggests evidence-backed actions each week.

Other tools we have looked at

Several platforms report citations, and some do it well for a lower price. Mentionable records cited domains and the fan-out queries ChatGPT and Perplexity run. Ceyo maps sources by type. Allmond gives full response and citation access from $29 a month with a weekly or monthly cadence. Ahrefs Brand Radar reports on Reddit and YouTube sources, with prompts modeled from keyword data. For a quick view of who gets cited, those work. We stay on Promptwatch because the follow-up questions, whether the page was crawled and whether the citation brought visits, are answered in the same workspace.

What we tell clients

Citations move. A page can be cited for weeks and then disappear with no change on the client's side, and the reverse is also true. We report trends against a fixed panel, never a single reading, and we do not promise that a page will be cited. What we can promise is a clear picture of which sources win, whether the engines can read the client's pages, and what the next fix is.