Cloudflare launched its AI visibility dashboa

On August 6th, Cloudflare added a new piece to its AEO Suite: the AEO Visibility Dashboard, available only in early access and requested from the Overview tab of the Cloudflare dashboard. There is no general availability date yet and no public pricing. The suite's existing tool, Agent Readiness, answers a more basic question: whether an AI agent can find and read the site at all. The new dashboard aims one step higher, at whether that site shows up, gets named, and comes off well when someone asks an assistant something.

Five days ago we wrote that the IAB had set the first standard for measuring brand visibility in AI, and that the standard certifies no one: it establishes no certification and evaluates no individual provider, and the burden of proof sits with whoever is buying the tool. The question worth asking any vendor is about method, not result. Cloudflare, five days later, is the case with a name attached to that question, and it is worth saying up front that it has something genuinely different from most tools in this category. The issue is not that the difference is fake. It is that the release attaches it to numbers that do not actually come from there.

The gap between the release and the tab

The dashboard measures four things: Citation Rate (which AI platforms pull from the site), Mention Rate (whether the assistant names the brand), Prominence (how much of an answer gets attributed to that brand), and Share of Voice (how the brand compares against competitors). According to technical coverage of the launch, the models included are Anthropic's Claude and OpenAI's GPT, and nothing else. Cloudflare's press release does not name any model, so that coverage has to be pieced together from outside reporting rather than the primary source.

The release is explicit about something else: why its method is supposedly better than everyone else's. It says most tools "attempt to fill that gap by only sending test prompts to AI chatbots and sampling the responses, a method limited in scale and prone to inconsistency if not paired with other data signals." And it offers the alternative: "deeper, more transparent insights than sampling test prompts alone," possible "because it operates at the network layer between AI platforms and the websites they access, observing actual crawl and referral activity directly, across millions of sites." That is a strong claim, and, for the rest of the product, an accurate one.

The problem shows up when you look at what the visibility tab specifically does. ppc.land's technical read is pointed: "the blog post describes the AEO tab doing precisely what the release characterises as the limited method: probing assistants with likely prompts and reading the answers." Citation Rate, Mention Rate, Prominence, and Share of Voice do not come from observed traffic. They come from prompts Cloudflare wrote, run against the models, with the answers read afterward. It is the same method the release uses as the example of what needs to be surpassed.

A panel per category, not a snapshot of your account

There is a second technical detail that compounds the first, and it matters most for anyone using Share of Voice to compare against competitors. According to ppc.land, Cloudflare "builds a benchmark panel once per industry category and reuses that baseline across every account" in that category. The figures a site owner sees "are drawn from a pre-computed dataset for that category, not from queries issued at the moment of the scan."

That changes what the number means. Share of Voice does not compare the site against its competitors today; it compares the site against a category sample taken at some point, at a refresh cadence the release does not specify. For a brand that launched a campaign last week, or for a category where assistants shift their favorite sources quickly, that gap between the data and the moment can be the difference between a useful reading and a late one. Neither of the two issues, synthetic prompts or a reused panel, makes the number useless. It makes them a directional snapshot, not the real-time measurement of your own account that the release's language suggests.

Where the network-layer data actually lives

This is where Cloudflare earns the credit the release claims, just in a different part of the product. According to ppc.land, the data genuinely built at the network layer does not live in the AEO Visibility Dashboard. It lives in a separate panel, AI Operator Activity, which tracks actual crawl and referral traffic per AI operator. That is exactly what almost no other AI-visibility vendor can offer, because almost none of them sit on the network infrastructure that bot visits and referrals actually pass through. Cloudflare does, because it operates CDN and DNS for a large share of the web.

The practical distinction is this: Agent Readiness answers whether the site is accessible to AI agents. AI Operator Activity answers what real traffic from those agents is already arriving, measured on the network. The AEO Visibility Dashboard answers how the brand shows up in simulated answers, with a sampling method comparable to any other tool in the category, running on a panel shared across an industry. Those are three different questions, with three different levels of reliability, packaged under the same release and described with the same "network layer" language.

The question that separates vendors

It is worth reading this with the same standard for any AI-visibility vendor, not only Cloudflare: when a tool claims a data advantage, the question is which specific panel each metric you are looking at actually comes from, rather than accepting the general advantage as if it applied to the whole product. For a team already on Cloudflare, the concrete move is to treat Citation Rate, Mention Rate, Prominence, and Share of Voice for what they are, a directional reading, useful for spotting a trend and not for budget decisions, and to check Agent Readiness and AI Operator Activity separately whenever the question is about technical access or traffic already underway.

What we still do not know is considerable: there is no general availability date or price, it is unclear how often the per-category benchmark panel refreshes, and the release does not say whether model coverage will expand beyond Claude and GPT or whether Gemini, Perplexity, Copilot, and Grok are absent by product decision or technical limit. There is also no public statement from Cloudflare on how many of the IAB standard's nine criteria the AEO Visibility Dashboard would meet if someone asked. That question, published five days ago, is still unanswered here too.

Cloudflare launched its AI visibility dashboard, and its four metrics do not come from the network layer that makes it different