AI Search KPIs: The 4 Numbers to Show Your Board

The best KPIs for AEO and GEO are four numbers a board can follow: how often AI answers mention you, your share of those answers against named competitors, visits referred by AI assistants, and the leads those visits bring. Search Console can't show AI clicks yet, so these four are the proof you can count today.

GEO (generative engine optimisation) is the work of getting your brand named and cited in AI answers from tools such as ChatGPT, Gemini, Perplexity and Google AI Overviews. AEO (answer engine optimisation) is the older name for much the same job. If you lead marketing at a B2B SaaS company, the work is often the easier part. The harder part is the meeting where the CEO asks whether any of it is paying off.

We've heard that question on two sales calls and in one prospect email this autumn, in almost the same words: is AI search traffic actually worth it, because leadership wants proof. Here are the four numbers we'd put on one board slide, and how to collect each one with tools you probably already have. If you're still working out why traffic fell in the first place, start with our guide to traffic that's down while rankings look fine.

Why can't Search Console show AI clicks yet?

Search Console can't show AI clicks because Google's new AI report counts impressions only, and the clicks it does record are mixed into the normal Performance report. So the report most SEO teams open first can show that you appear in AI answers, but it can't show whether anyone clicked through.

What the generative AI report shows

Google launched the Search generative AI performance reports on 3 June 2026 and says that, as of 31 August 2026, they have rolled out to all websites worldwide. The reports list five things: impressions, pages, countries, devices and dates. There is no clicks column. Google says it may add metrics over time, but it hasn't said which or when.

That makes the report useful for one board question (are we showing up in Google's AI features at all?) and no use for the next one (did that send us anyone?).

Where AI Overview and AI Mode clicks actually go

Clicks from AI Overviews and AI Mode do get counted, just not anywhere you can separate them. Google's documentation on AI features and your website says they're reported in the Performance report within the Web search type, alongside every other organic click. There is no filter that pulls them out.

Meanwhile the clicks that used to come from the top result are thinner. Ahrefs' study of 300,000 keywords, published on 4 February 2026, found that the presence of an AI Overview correlates with a 58% lower average clickthrough rate for the top-ranking page, comparing December 2025 with December 2023. That's one reason boards ask the question at all: rankings can look stable while traffic falls.

Our view: don't wait for Google to add a clicks column. Report what you can count today, and say plainly what each number misses.

What are the 4 numbers to show your board?

The four numbers are mentions, share of answers, AI-referred visits and leads from AI. Read in that order, they take the board from "are we there?" to "is it making money?", and each one covers a gap left by the one before.

1. Mentions: are we in the answers at all?

A mention is a response from an AI assistant that names your brand. Ahrefs counts a single mention when a brand appears at least once in an AI generated response, however many times it's named. We use the same rule. Measured as: of the responses to your fixed set of buyer prompts this month, how many name you.

Collect it with a tracker such as Ahrefs Brand Radar, or by hand: run the same prompts monthly in ChatGPT, Gemini, Perplexity and Google AI Mode and log the answers in a sheet.

2. Share of answers: are we winning against competitors?

Share of answers is your mentions as a share of all mentions of the brands you track. Ahrefs calls it AI share of voice. Ahrefs' guide to AI share of voice gives the basic formula: responses that mention your brand, divided by responses that mention any tracked brand, times 100. We call it share of answers on a board slide because nobody outside marketing needs to learn what share of voice means.

3. AI-referred visits: are people clicking through?

AI-referred visits are sessions that arrive from AI assistants. In GA4 they sit in the AI Assistant default channel, which Google describes as users arriving from sources like ChatGPT, Gemini, Deepseek, Copilot or Grok. Measured as: GA4 sessions in the AI Assistant channel this month, from the Traffic acquisition report.

Treat this number as a floor. The same GA4 page says the channel excludes Google's AI Overviews and AI Mode, which are counted as Organic Search.

4. Leads from AI: is it making money?

Leads from AI are new leads that came through, or were sent by, an AI assistant. Measured as: new leads this month whose self-reported source in your CRM is "AI assistant", shown next to GA4 key events from the AI Assistant channel.

You need both, because the two can disagree by a lot. In Graphite's analysis of n8n (June 2026), GA4 last-touch credited AEO with roughly 1% of conversions, while n8n's post-conversion survey credited it with roughly 9%.

The board slide

One row per number, plus a column for what each one misses, so nobody in the room over-reads it.

How do you collect each number?

You can collect all four with a fixed prompt list, GA4 and one new form field. None of it needs a new analytics stack. We cover the full method in our guide to measuring AI search visibility, so this section sticks to what the board slide needs.

Mentions and share of answers

Pick your buyer prompts and your competitor list once, then keep both fixed. We usually start with 20 to 50 prompts that a buyer would type when building a shortlist, such as "best [category] software for [use case]", plus three to five competitors your sales team actually meets. Our guide to AI share of voice covers how to build the prompt set and run the sums.

Fixing the setup matters more than the tool. Ahrefs warns that the number depends on which prompts, platforms and competitors you track, and that answers can vary between runs, so it advises trusting trends that persist across several reporting periods.

AI-referred visits

Open GA4, go to the Traffic acquisition report, keep the default channel group and read the AI Assistant row. Before you show it, tell the board two things.

First, AI Overviews and AI Mode sit inside Organic Search, so this row will never include Google's own AI features. Second, AI tracking can change without notice. Search Engine Roundtable reported on 30 October 2026, based on tracking shared by Natzir on X, that Gemini stopped adding utm_source=gemini to cited links around 28 September. The same report says visits from the Gemini iOS app send no referrer, so they land in GA4 as Direct.

It's also worth checking where Perplexity visits land in your own GA4. Google's channel description doesn't name it, so look at the source before you assume it's in the AI Assistant row.

Leads from AI

Add a "How did you hear about us?" field to your demo or contact form, with "AI assistant (ChatGPT, Gemini, etc.)" as one of the options. Store the answer as a field in your CRM, so you can count AI-sourced leads and, later, AI-sourced pipeline. Report it next to the GA4 key events from the AI Assistant channel, never instead of them.

Self-reported data has its own gaps, because people skip questions and remember loosely. But it catches the buyer who asked ChatGPT for a shortlist and then typed your URL by hand, which no click-based model will ever see. We go further into this in the AI search attribution problem nobody has solved yet and in our guide to SaaS lead attribution.

What does good look like after 90 days?

After 90 days, good looks like direction, not a target: mentions and share of answers trending up on the same prompt set, AI-referred visits growing from a small base, and the first self-reported AI leads showing up in the CRM. We don't publish benchmark targets for these numbers, because they depend so heavily on your category and your prompt set.

The month 1, 2 and 3 slide

Keep the same four rows and add one column per month. Month 1 is your baseline, and it's fine if leads from AI is close to zero. Month 2 shows whether the setup is stable.

By month 3 the board can see a trend in mentions and share of answers, and you can start to show AI leads next to the pipeline numbers they already follow. To put a value on that pipeline, our guide to AI search ROI shows the sums.

Reading a drop: noise or a real loss

The four numbers also help you find the cause of a drop. If mentions fall, check first whether the prompts, competitors or platforms changed. If visits fall while mentions hold, look at tracking before strategy, because a lost UTM tag or referrer can cut the GA4 number overnight. If leads fall while the other three hold, look at the form and the sales process.

Put the four numbers on one slide this month

You don't need a perfect model to answer the board. Fix your prompts and competitors, read the AI Assistant row in GA4, add one form field, and show the four numbers with what each one misses. The answer gets stronger every month you keep the setup the same.

That's why the slide ends in leads. If you want help setting up the four numbers for your own category, our free AI search visibility audit starts with your buyer prompts and competitors, and our GEO service for B2B SaaS covers the work that moves them.

Frequently Asked Questions

What are the best KPIs for AEO and GEO?

The most useful KPIs for AEO and GEO are mentions, share of answers, AI-referred visits and leads from AI. Mentions and share of answers show whether AI assistants name you on your buyer prompts. Visits and leads show whether that visibility turns into traffic and pipeline. Together they cover presence, competition, clicks and revenue on one slide.

Can I see AI Overview clicks in Search Console?

Not separately. Google's generative AI report in Search Console shows impressions, pages, countries, devices and dates, with no clicks. Clicks from AI Overviews and AI Mode are counted in the main Performance report under the Web search type, mixed in with other organic clicks.

What should I tell the board if one of the numbers drops?

Tell them whether the setup changed, and whether the drop has lasted more than a month. A changed prompt set or competitor list makes the month incomparable, and AI answers vary from run to run. If the setup was the same and the drop holds for a second month, treat it as real and say what you'll do about it.

How often should I report AI search KPIs to leadership?

We recommend monthly, with a trend view each quarter. Weekly numbers move too much on small samples to be useful. Report on the same day each month, with the same prompts and competitors, so each month compares cleanly with the last.

Is share of voice the same as share of answers?

Yes, in this guide they mean the same thing. Share of answers is our board-friendly name for AI share of voice: the responses that mention your brand as a share of the responses that mention any brand you track. Ahrefs Brand Radar also offers a version weighted by AI-adjusted search volume, so note which one you report and keep it the same each month.