The AI Citation Scorecard: How Leaders Measure Whether AI Search Is Sending Revenue

If you cannot tell whether AI search is sending revenue, you do not have an AI visibility program. You have a content habit. Leaders need a scorecard that is small enough to review monthly and hard enough that a vendor cannot decorate it.

Why the usual dashboard lies

Rank trackers assume a stable list of blue links. Answer engines do not work that way. The same prompt can cite you on Tuesday and a publisher on Thursday. A click may never happen because the answer was complete. A mention can also be wrong: the model names you and describes a service you do not sell, or places you in a city you left. Celebrating “we showed up in ChatGPT” without checking accuracy trains the company to chase logos.

The Farber citation scorecard exists so a CEO can ask one question in the monthly meeting: are we more likely to be named, correctly, on the prompts that create revenue, than we were 30 days ago?

Build the prompt set once

Twenty-five prompts is enough. More than forty becomes a research project nobody maintains. Pull them from sales calls and from Search Console queries, not from a keyword tool’s longest list. Cover five intents:

• Definition. “What is [category] and who is it for?”
• Selection. “What should I look for in a [provider]?”
• Comparison. “[Approach A] vs [approach B] for a [company type].”
• Commercial. “How much does [service] cost?” and “Who do you recommend for [service] in [market]?”
• Problem. The failure your buyers are already in: a suspended profile, a traffic drop after an AI overview, a chatbot that does not convert.

Freeze the wording. If a prompt is bad, replace it and mark the series break. Do not quietly edit prompts to manufacture a better month.

The five lines on the scorecard

1. Citation rate. Prompts where your brand is named, divided by prompts run. Report it per engine. An average hides a Perplexity win and a Google miss.
2. Accuracy. Of the citations, how many describe the offer, the market, and the proof correctly. An inaccurate citation is a liability. Log the sentence that is wrong and the page that should have prevented it.
3. Share of answer. How often you appear in the top three names versus the competitors you actually lose deals to. Being cited next to an irrelevant brand is not a win.
4. Source of the citation. Your site, a directory, a review profile, or a third-party article. If you are never the source, you do not control the next edit. That tells you whether the next sprint is on-site structure or earned mentions.
5. Assisted pipeline. Forms, calls, and booked consultations that named search, ChatGPT, Perplexity, or “I asked an AI.” Add one field to the form. Without it, citation rate floats free of money.

How to read a month

A useful month changes a decision. Use this reading order.

Accuracy before reach. If you are cited and misdescribed, fix the canonical page and the Google Business Profile before you publish anything new. You are scaling the error.

Commercial prompts before definitions. Ranking as the explainer of “what is GEO” feels good and may not move a deal. The “who should we hire” and “how do we recover” prompts are the ones sales feels.

One engine’s silence is a clue. If Google AI Overviews cite you and ChatGPT does not, your pages may be crawlable but not written as extractable answers. If Perplexity cites a trade article and never you, you have an authority gap, not a title-tag gap.

Flat citation rate with rising pipeline can still be a win. A correct citation on five commercial prompts is worth more than a mention on fifteen definitions. Do not let the rate become the only number anyone defends.

What the monthly review should decide

Keep the meeting under 30 minutes. Three decisions, no more.

• Which inaccuracy gets corrected this week, and on which URL.
• Which single commercial prompt you are not winning, and whether the gap is on your site or in third-party sources.
• Which tactic stops. If a page, a directory blast, or a prompt has not moved accuracy or pipeline in two cycles, stop funding it.

This is the same discipline we tell buyers to demand in what to look for in an AI marketing partner. If your current report cannot support those three decisions, it is a status deck. Pair the scorecard with the AI visibility audit once a quarter, and with AI Business Consulting if you want the cadence run for you. We will not invent a revenue number the scorecard did not earn.

Frequently Asked Questions