Profound vs Peec vs Standing: how the three actually measure

Most comparisons in this category are feature tables written by whoever is selling. This one is about method: how each tool collects a number, and what that number can support once you have it. Every figure below comes from the vendor's own pricing page, documentation or published research.

What all three do

You give the tool a set of questions your buyers ask AI. It runs them against several engines on a schedule, records whether you were named and what got cited, and charts it over time. Any of the three will tell you roughly whether you show up.

The differences start with how many times a question gets asked before a number is reported.

Breadth or depth: the trade every tool makes

Every tool here spends the same budget. The only real decision is whether to spend it on more questions or more runs of each question.

Profound and Peec spend it on breadth. Profound's Starter plan is $99 a month for 50 prompts; Growth is $399 for 100. Peec starts at $95 for 50 and reaches 350 on Advanced. Both run each prompt once per engine per day.

For Profound this is confirmed by arithmetic on their own pricing page: 50 prompts on one engine across 30 days is exactly the 1,500 monthly responses they list, leaving no room for repeats. Their help centre says prompts "refresh once per day." Peec never states a run count anywhere in its documentation, but its credit formula settles it: "1 prompt × 1 model × 1 day = 1 credit."

We spend the budget the other way. Ten questions, plus rival-specific ones, asked five times each on five engines. Fewer questions, and every one of them answered enough times to see how much the answer wobbles.

That is the whole trade. A wide basket asked once tells you about more of your market. A narrower basket asked five times tells you how much of what you are seeing is real.

What one run cannot give you

Ask a question once and you get an answer. Ask it five times and you get a range. The range is the thing, because these engines are not deterministic, and both competitors say so.

Peec's documentation puts it plainly: "AI models have inherent randomness in source selection. A source might appear one day and be completely absent the next." Their advice is to watch trends over weeks. No number is attached to how many weeks, or how large a change has to be.

That is the gap. Search either product's documentation for a confidence interval, an error bar, a variance figure or a significance threshold and there is nothing. The score moves, and you are left to judge by eye whether the movement meant anything.

Profound is the more striking case, because they have the data. In June 2026 they published a genuine variance study: 5,271 configurations, 753 prompts across seven platforms, run once a day against ten times a day for a fortnight. It is the most serious methodological work anyone in this category has published. It also computes precisely what a confidence interval needs, and none of it reaches the product. No error bar, no minimum detectable change, no significance test in the dashboard.

Their conclusion was that "running 10× or 100× wouldn't meaningfully change what the numbers tell you," because most of the residual movement is the platform drifting rather than sampling noise. We think that finding is sound, and that its scope is narrower than it reads. The precision claim is explicitly about "a portfolio of thousands of prompts" aggregated over two weeks. A customer on the Starter plan has 50 prompts on a single engine and opens the dashboard every morning. That is a different statistical situation from the one the study validates.

It also sits awkwardly beside their marketing. Their feature page says you can "tie visibility changes directly to content updates or campaigns you've launched." If most observed movement is platform drift outside your control, then most of what the dashboard shows cannot be attributed to your work, and their own research is the evidence.

Three things we publish that neither does

A confidence band on the score. Computed across the five runs, shown in the product, so you can see how precise the number is before you act on it.

Alerts that stay quiet below the noise. Because there is a spread, a change can be held until it is larger than what the score does on its own. Neither competitor has this, because neither has the sampling to compute it.

What happened when an engine failed. If an engine does not answer, we report that rather than recording a zero. Neither vendor publishes a failure policy. It matters more than it sounds: a failed request and a genuine no-mention are different events, and a score that treats them alike drifts downward for reasons that have nothing to do with you.

We also name the model version behind each engine. Peec deliberately does the opposite, abstracting them behind "model channels" that "match results across all underlying model versions on that channel." The measuring instrument can then change under a stable label, which moves your trend line with nothing to explain why.

Where each one fits

Profound and Peec are built for coverage. They track more questions than we do, refresh every 24 hours, and Profound reaches nine engines on Enterprise. If your priority is watching the widest possible basket and seeing a fresh number each morning, that is what they are designed to do, and they are mature products.

We are built for a different question: not "what is the number today" but "did it actually move." That is why the basket is narrower and each question is asked five times, why the band is published rather than kept in a blog post, and why an alert waits until a change clears the noise.

If nobody ever asks you how you know a number moved, breadth is the better buy. If someone does, and increasingly someone does, you need the spread.

Last checked July 2026 against each vendor's published pricing, documentation and research. If anything here has gone out of date or is wrong, tell us and we will correct it.