Our board asks how much revenue comes from AI search and our analytics can't answer that: what are we missing
When the board asks for AI search revenue and analytics has no answer, here is what we find missing in client setups and how we close each gap.
This question reaches us in almost the same words every few weeks. A board member read something about AI search, asked what share of revenue it drives, and the marketing team could not say. The usual reaction is to blame GA4. The real problem is that four separate things are missing, and each one has a different fix.
Missing piece one: a clean AI traffic segment
Out of the box, AI referrals are scattered across referral sources, and some land in direct. The first job is to build a channel group that collects the named AI sources, ChatGPT, Perplexity, Gemini, Claude, Copilot and others, and attach conversions to it.
We do this with Promptwatch's visitor analytics, which counts AI-referred visits and conversions through a lightweight script, the GTM template or the WordPress plugin. Having it outside GA4 is useful because the definition stays stable when someone edits channel rules. Promptwatch shows the result next to your other data, so finance can see one set of numbers.
Expect the number to be modest for many sites. Say that before anyone else does.
Missing piece two: the Google blind spot
Google's AI Overviews and AI Mode traffic sits inside Search data. You cannot split it exactly. We tell clients plainly that this share is estimated, and we use Search Insights, which overlays Google Search Console and Bing Webmaster traffic (including Copilot) on AI crawls and citations, to show direction rather than a precise figure.
Boards handle uncertainty fine when it is labelled. They handle surprises badly.
Missing piece three: the influence that never clicks
A buyer asks an AI engine for options, sees the brand named, and goes to search for it later. That conversion looks like branded search. No tool captures that perfectly, and we do not pretend otherwise.
What you can do is track visibility on a fixed prompt panel and compare it to branded search and direct trends over the same weeks. Promptwatch's Visibility Score, Share of Voice and A/B period comparison make that comparison simple to draw. It is correlation, and we present it as such, as a separate line from the measured floor.
Missing piece four: the link back to causes
Even with clean traffic, the board's next question is "what do we spend to get more?" To answer that you need to know which pages the bots read, which pages get cited, and where you are absent.
Agent Analytics gives IP-verified crawler logs for GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot and the other real AI bots, with errors explained and a citation rate per page. Citation analytics shows the pages and outside domains the engines rely on. The 360 Insights view puts crawls, citation clicks and AI-referred visits side by side per page. In our experience this is the view that turns "AI search might matter" into a prioritized fix list.
Crawler logs begin on the Professional plan at $245 a month. Essential at $95 covers visitor analytics and site audits, so some clients start there and move up once the crawler question becomes real.
What we report to the board
Three lines, always in this order:
- Measured: AI-referred sessions and conversions from tagged sources, with the trend.
- Estimated: Google AI surfaces and assisted influence, labelled as estimates.
- In progress: crawler fixes, tracking gaps closed, content gaps being filled.
Under the lines goes one chart, not twelve.
A published example of what the measured end can look like is the TreeHouse case study on Promptwatch's site, where AI search became about a quarter of organic conversions, mostly from ChatGPT, after crawler access was fixed. That is their result, and we use it only to show it can be measured.
What we will not promise
We will not give a board a total AI revenue figure that includes everything AI touched. It cannot be known. We also will not promise that any engine will recommend a brand. Answers shift between runs and models, and no tracker changes that.
A starting checklist
Within a week most teams can do the following: define the AI channel group, add conversion events to it, set up a fixed prompt panel of buyer questions, connect crawler logs through the CDN or WordPress plugin, and schedule a monthly PDF report. After one full month you have a baseline, and the board question has a real answer instead of a shrug.