You have been running AI search work for a quarter, and now you have to put it in front of the board. So you open the same dashboard you use for everything else, the one with sessions and clicks and attributed pipeline, and you go looking for the line that proves the program worked. It is not there. Either it reads flat, or it reads down, on the exact initiative you are supposed to be defending. And you are standing in front of the people who approve your budget with a slide that makes your best work look like your worst.
This is the reporting trap, and almost everyone walks into it, because the instinct is right and the instrument is wrong. The instinct is to translate AI search into the numbers the board already trusts. The problem is that those numbers are precisely the ones that cannot see AI search, so the translation does not just undersell the work, it actively argues against it.
Why the dashboard lies about this specific program
The return from AI search arrives off-click, and a click-based report is built to miss exactly that. We have made this case at length about whether the work is worth it at all, but in the boardroom it becomes a sharper problem, because now the blindness is not just yours, it is on a screen everyone is looking at.
Two things break the number. First, most of what AI search does never becomes a click. The engine answers in the chat, the buyer reads it, forms a view, and arrives at your form already holding a verdict, so the influence happened in a conversation your analytics never saw start. Second, the fraction that does click mostly lands in Direct, because AI referrals strip their referrer, so the traffic you did earn is filed under the one channel that proves nothing. The board is reading a report where the wins are invisible and the leftovers are misattributed. No wonder the line looks flat.
So the reflex is to fix the number. Build an attribution model, assign AI search a share of pipeline, put a confident percentage on the slide. Do not. That is the move that gets the program killed, and I will come back to why.
Report exposure first, because a board understands absence
The single most useful thing you can put in front of a board is not a metric. It is the answer itself. Run your real buyer questions, the category question, the comparison question, the “best tool for X” question, through the engines live, and show the board what comes back. Who the engine names. Where you sit in that list, or whether you appear at all. Which competitor it reaches for first.
This reframes the entire conversation away from “did we get traffic” and toward “is the buying decision for our category already happening in a room we are absent from.” A board does not need a tutorial to understand competitive absence. Showing them that the engine recommends three competitors and never mentions you, on the exact question your buyers ask, lands harder than any conversion chart, because it is not a marketing metric, it is the market choosing without you in the room. And it sets up the only honest framing of the spend: the exposure exists whether or not you invest, so doing nothing is not the zero-cost option, it is just the one whose cost never appears on a report. That is a risk a board can act on.
Report leading indicators, not lagging traffic
The second thing to show is movement, and early on the movement does not live in traffic. It lives one layer up, in the indicators that shift before any click does. Are your pages getting crawled by the retrieval engines. Are you starting to appear on Perplexity, on Google’s AI answers, on Copilot, for the buyer questions that matter. On how many of those questions are you now cited, versus zero last quarter. Where do you sit in share of voice against your named competitors when the engine builds its shortlist.
These are the honest early scoreboard, and they have a second job on a board slide: they let you name the flat stretch before it is used against you. AI search compounds on a curve that looks like nothing is working for weeks before it bends. If you have not told the board that in advance, the flat quarter reads as failure and someone moves to cut the line. If you have, and you can point to citations climbing on the buyer questions while headline traffic is still flat, the same quarter reads as a program loading. Same data. The difference is whether you set the expectation before the number did.
Report pipeline influence as evidence, not as a percentage
The third thing the board actually cares about is revenue, and here you have real signal, it is just not the kind that fits in a cell. It is the demo that opened with the buyer already informed, quoting a framing you recognize from the engine’s answer. It is the sales call where the rep hears “I read that you do not have X” about a feature you shipped a year ago. It is the deal where the prospect arrived comparing you to the exact competitor set the engine names.
Collect those. A handful of real, specific, sourced moments from sales is more persuasive to a board than a made-up attribution percentage, because each one is checkable and none of them insults the room’s intelligence. Label them for what they are, directional evidence that the engine’s answer is reaching your pipeline, not a measured contribution. A board can hold “we cannot yet put a clean number on this, and here are six concrete instances of it working” far better than you expect. What it cannot hold is a precise-looking number that falls apart under one question.
What I got wrong
The first time I had to put an AI search program in front of a client’s board, I did the thing I just told you not to do. I built a beautiful attribution dashboard, modeled a Direct-leak recovery, and put a confident number on the slide: this much pipeline, this quarter, from AI search. It looked authoritative. Then a board member, a former CFO, asked one question. How did you model the leak. And the honest answer was that I had estimated it, because the real number is unknowable by design. The percentage had no floor under it. In the space of that one question, the whole program went from “promising” to “the thing the marketing team is inflating,” and it nearly got cut, not because it was failing, it was working, but because the slide meant to defend it had made a claim it could not back.
I rebuilt the next quarter’s report from scratch. No invented percentage. Instead: the live exposure check with the actual engine answers, the leading indicators moving quarter over quarter, share of voice against the named competitors, and five real sales moments where the engine’s answer showed up in a deal. It was a smaller, humbler story. It survived the board, and the budget held. The confident big number nearly killed the program. The honest smaller one saved it. I do not build the attribution slide anymore.
The report that survives the room
Reporting AI search to a board is not a measurement problem, it is a framing problem, and the failure is trying to force a new kind of return through the old dashboard’s slot. The traffic number will make working work look like failure, and the attribution number you build to fix it will make credible work look inflated. Both roads end with someone reaching for the budget.
The report that holds does three things instead. It shows the board the exposure, the actual answer the engine gives your buyers, so absence becomes a risk they can see. It shows leading indicators moving, with the flat stretch named in advance so nobody misreads the loading for a stall. And it shows pipeline influence as concrete, sourced evidence rather than a percentage that cannot defend itself. Smaller, slower, honest. It is the version that is still funded a year from now.
If you want your AI search program reported the way a board can actually act on, the exposure it is missing, the indicators that are moving, and the story the traffic dashboard is hiding, that is the engagement, and it runs on a documented method.