How to Track AI Citations and Connect Them to Revenue

Why citation counts alone cannot answer the revenue question

A dashboard that says 'your brand was cited hundreds of times this month' tells you very little. Cited where, on which question, linking which page, and did anyone buy? Revenue questions need a chain: a buyer asks a question, an engine answers, your page is linked or your brand is named, someone clicks or remembers, and a deal closes. Tracking AI citations is only the first link in that chain.

Keep the two answer-level signals apart from the start. A mention is your brand named in the answer. A citation is the answer linking one of your pages as a source. Mentions influence recall and shortlists; citations are what can produce a click. Blending them into one 'visibility score' hides the part you can actually connect to revenue.

The three layers of AI attribution

Think of the setup as three layers, each with its own data source and blind spot. The answer layer is collected by running the same prompts on ChatGPT, Perplexity, Gemini, Copilot and Google AI Overviews on a schedule and recording mentions, citations, cited URLs and competitors. Independent AI-search tracking can automate this; a spreadsheet works for a first baseline.

The click layer lives in your analytics. The revenue layer lives in your CRM or back office. The mistake most teams make is building one layer well and assuming the others. A strong citation share with no referral segment is a guess. A referral segment that never reaches the CRM stops at sessions.

• Answer layer: prompt set, engines, mentions, citations, cited URL.
• Click layer: AI referral sessions, landing pages, on-site conversions.
• Revenue layer: deals, funded accounts or subscriptions with an AI source attached.

Fix the prompt set before you start and change it rarely. If you swap half the prompts every month, a rise in citation share may only mean you picked easier questions. Add new prompts as a separate group, keep the core list stable for at least two quarters, and record which engine, country and language each run used, because answers differ across all three.

Setting up the click layer in analytics

ChatGPT search links carry utm_source=chatgpt.com, which makes them easy to isolate. Perplexity, Gemini and Copilot appear partly as referrers: perplexity.ai, gemini.google.com and copilot.microsoft.com. Create a custom channel group called AI Assistants that matches those sources before the default rules sort them into Referral or Organic.

Google is the exception. Clicks from AI Overviews and AI Mode are counted inside Search Console's normal Web performance data, not in a separate report, so you cannot split them out cleanly. Track them indirectly: watch the queries and pages where AI features appear and compare click-through over time.

Then make sure the cited pages fire the same conversion events as the rest of the site. The landing page report for the AI Assistants channel is where you find out whether citations land on pages that convert or on blog posts with no next step.

Carrying AI source data into the CRM

Sessions are not revenue. Pass the first-touch source and landing page into a hidden form field or your sign-up API so every lead record keeps AI referral source data. Then add a self-reported question to the form: how did you first hear about us, with 'ChatGPT or another AI assistant' as an option. Much AI influence is zero-click; the buyer reads an answer, remembers your name and later arrives through branded search or direct. Self-reported answers catch what analytics cannot.

For a forex broker the chain looks like this: a citation on a 'lowest spread for small accounts' prompt, a session landing on the fee page, a demo account, a live account application, and a funded account. Each step has its own system. Joining them by user ID is what lets you say which prompts and which cited pages produced funded clients, rather than which produced traffic.

What a useful monthly report should show

A good report answers a decision, not just a question. For each priority prompt it shows mention rate and citation rate per engine, the cited URL, and the competitor holding the source slot when it is not you. Below that, it shows AI referral sessions by landing page, their conversion rate, and the deals or accounts they fed.

Look for honest caveats. Engine answers vary run to run, so a fixed prompt set sampled several times per month is more reliable than a single snapshot. Attribution will always be partial. Studies report that AI referrals convert several times better than classic organic visits, but your own funnel data should set the budget. A report that only ever goes up is a reason to ask harder questions.

What Suggesting.ai sets up for you

Suggesting.ai begins with a free 48-hour brand and AI presence audit that builds your first answer-layer baseline: your buyer prompts, mentions and citations per engine, and the pages competitors are cited for. We then help your team set up the AI Assistants channel, the CRM fields and the self-reported question, so the monthly report runs from citation to revenue.

The organic GEO work and, where the category is eligible, ChatGPT Ads management are reported in the same view. For financial services, OpenAI's ad policy generally prohibits ads outside the US, so we confirm eligibility per market in the audit. Retainers start from €12,000 per month plus 15% of ad spend for one or two markets, with exact scope set after the audit.