Benchmarking Turns AEO Into an Actual Channel
For the last 18 months...AI search has been the only performance channel in marketing running without a price discovery mechanism.
Almost every marketing leader I have talked to in the last six months has asked some version of the same question…
“How are we actually doing in AI search compared to our competitors?”
Most arguments about AEO measurement get framed as a data problem but they’re really not. They are a market structure problem, and until this week the channel was missing the one piece of infrastructure every other legitimate marketing channel has.
Consider what makes a performance channel legit inside an enterprise. Not volume. Not even attribution, which most channels have only partially solved. The thing that matters is price discovery, a mechanism by which the market tells you what your performance is worth relative to everyone else trying to do the same thing.
AEO will never have a literal price the way paid does. That is not the point. Price discovery in market structure terms is about comparative valuation, not dollar amounts.
SEO has it. The entire channel is built on it. You cannot rank in position three without two URLs sitting above you, and Google’s results page is a daily, public, brutally honest auction of relative quality. Position is comparative by definition. There is no such thing as a good ranking in a vacuum.
Paid search has it. Auction data, CPC benchmarks, impression share, quality scores. You know what a click is worth because the market sets the price every time you bid.
Brand has it (even if imperfectly). Share of voice, unaided recall, brand tracking studies. Signals that are noisy but real. A marketing leader knows roughly where the company sits.
Even email, the oldest channel in the stack, has open and click rate benchmarks granular enough to be useful.
AEO has had none of this. For nearly two years the channel has run on absolute counts. I have sat in rooms where one site’s monthly citation count was treated as a triumph, and rooms where another site’s count, many times larger, was treated as a disappointment. In neither case did anyone actually know which read was correct. This is not because the people working in the channel are unsophisticated. It is because the data to answer the question correctly did not exist outside any one platform’s walls.
What looked like a measurement problem was actually a structural one. You cannot have a legitimate channel without comparative valuation. You can have activity, you can have spend, you can have dashboards that look the part. But until participants can see where their performance sits in a real distribution, the channel is running on faith, and budgets allocated on faith get cut first when the cycle turns. I am already watching that happen, unfortunately.
What benchmarking infrastructure in AEO actually looks like
This week Profound shipped Benchmarking in Agent Analytics, and it is the first credible answer to this problem the category has produced. They built the missing layer, the one that lets the channel calibrate itself against itself.
The mechanics are straightforward. Profound compares your AI citation performance against more than 100,000 pages tracked across the Profound Network, filterable by industry and company size, refreshed weekly. I have been a Profound customer long enough to remember when a cross-network view like this was technically impossible. That it is now table stakes tells you how fast this category is moving. Every page lands in one of four tiers, Poor, Fair, Good, or Great, based on where it falls in the percentile distribution. You see your site’s overall tier at a glance, and you can click into any page to see exactly where it sits in the network.
That is the description. Here is what it actually does.
It turns “we got cited 32,000 times” into “we are in the 73rd percentile for our industry and company size, with seventeen pages in the bottom quartile dragging the site average down.” Those two sentences are not the same. Not in category, not in usefulness, not in what they let you do on Monday morning.
The page-level view is the part I did not expect to matter most but I think it does. Site-level benchmarks are useful for executive conversations. Page-level benchmarks are what let content teams act. When every cited page on your site is ranked against peers, the bottom-tier pages stop being a guess about where to focus. They become a worklist.
There is a quiet design choice worth flagging. Pages need at least five AI citations to be included, and any tier segment with fewer than twenty sites is suppressed. The team understood that benchmarks at small sample sizes are worse than no benchmarks, because they give false precision. The thresholds are conservative enough that what you see can be trusted.
The part most people will get wrong
The first-order effect of this shift is obvious. Marketing leaders can finally answer the board question with a percentile instead of a screenshot. That is real and it matters, but it is the least interesting consequence.
Here are three things I expect over the next twelve to eighteen months that most teams have not priced in yet.
Budgets will get defensible. A percentile is something a CFO can argue with. A citation count is not. When AEO investment can be defended against a distribution (”we are in the 40th percentile, here is what it costs to move to the 70th”), the channel graduates from experimental line item to forecastable spend. That is the transition every emerging channel has to make, and most stall there for years. AEO might compress that timeline considerably. This is timely given I’m already companies who invested early in AEO start to look back and see if the return was worth it with skepticism on the investment.
Enterprise content teams will shrink, not grow. This is the prediction I am most confident in and the one most teams will resist. Most content orgs have been hiring against a “more content” thesis for the AEO era. The data does not support that thesis once you can see the distribution.
A fintech I advise pulled their AI citation data by page. Their top 14 pages were driving 61% of their AI citations. The bottom 180 pages were driving 9%. I want to say that out loud one more time, because I do not think most people have internalized what a Pareto distribution looks like when it shows up in their own content org. Fourteen pages. Sixty one percent. One hundred and eighty pages. Nine percent.
They did not have benchmarking when they built that long tail, so they had been investing in it on instinct for nearly 18 months, assuming volume would eventually compound. The distribution said the opposite. Kill half the bottom tier, take the saved budget, and triple down on the fourteen pages already doing the work. Most content orgs are sitting on a version of this same distribution and do not know it yet.
Once page-level percentiles exist, the math is unavoidable. The shrinking is what everyone will fixate on. It is the least interesting part. The real shift is in what the job becomes. The skill that compounds in the AEO era is not the ability to produce more content. It is the ability to architect workflows that account for performance across every discovery channel at once. That is the Marketing Engineer, the role I wrote about recently. The companies that see this first will rebuild the content org around a few people who think in systems instead of output. The companies that don’t will keep hiring against a volume thesis the data has already killed. I have watched this exact pattern run twice in SEO over the last decade. The people who read it early built careers on it. The people who read it late are still writing the brief they wrote in 2019.
The information asymmetry between participants and non-participants will widen fast. Benchmarks compound. The more sites in the network, the sharper every comparison gets, and the harder it becomes for anyone outside the network to credibly claim they know their position. Six months from now, the gap between teams who can answer “where do we rank” with a percentile and teams who can only answer with a screenshot will be the difference between teams that get budget and teams that lose it.
What this means for the category
The category is too young for any one platform to have won anything yet. The next eighteen months will be a real fight, and right now Profound is out in front of it.
The company that ships the benchmarking layer first usually holds a structural advantage in the period that follows, because comparative valuation is itself a network effect. Every additional site in the Profound Network sharpens every benchmark in it. Every sharpened benchmark makes the platform more valuable to be measured by. Every increment of value makes the next customer easier to win. It is not a moat that closes quickly. It is one that compounds, and the compounding starts now. Props to Profound for shaping this and finding a way to make the data they’re collecting valuable to the entire industry.
We have been calling AEO a channel for 18 months. Some of us have been calling it that slightly longer, and feeling slightly stupid about it every time the data part of the conversation came up. A channel without price discovery is a behavior. A channel with price discovery is an operating discipline.