Measure, never manufacture.
Most tools estimate a number and present it as fact. We built SEOAIO on the opposite promise. These 6 rules govern every score we show, and we hold ourselves to them in public, on our own site included.
- 1
Measure, never manufacture.
Every number we show is observed, not modeled to look good. If we can’t measure something, we do not invent a figure for it, we say so.
- 2
Show the receipt.
Every score opens to the evidence behind it: the observations, the confidence band, the sample size, and the exact source, the engine, page, or vendor it came from.
- 3
Hold the score until it’s real.
Your overall Visibility Score stays held until enough of its seven measurements are genuinely complete. A partial average is not an overall score, so we won’t print one.
- 4
Label what we haven’t measured.
Coverage is disclosed, always. An engine is “measured” only when it genuinely is. The rest are labeled rolling-out or planned. We never imply coverage we don’t have.
- 5
A comparison needs enough to compare.
We publish the floor: no benchmark renders below five of your sites, and fifteen for any comparison against other companies. Below it you get nothing, not a shaky number, and every figure prints its sample size beside it. We show the median and the middle range, never an average, because one outlier should not speak for a group.
- 6
We don’t sell the remedy.
We measure; we do not sell you the links or the retainer whose success we’re grading. The audit isn’t a funnel for a service, so the number has no reason to flatter.
Scoring edition vscore-2.0.0. Every score we store is stamped with the edition that produced it, so a number measured under an older edition is never silently compared against a newer one.
The engines we measure: and the ones we're still building
- PerplexityMeasured live
- Google AI Mode and AI OverviewsMeasured live
- ChatGPTMeasured live
- ClaudeMeasured live
- GeminiRolling out
We won't claim an engine is measured until it genuinely is. Today we measure Perplexity, Google AI Mode and AI Overviews, ChatGPT and Claude live; Gemini stays labeled honestly until it comes online.
Why we store your runs
Some tools sell “zero data storage” as a feature. It has a structural cost they do not mention: a tool that keeps nothing can never show you last month, never prove a change happened, and never put an interval on a trend, because there is nothing to compare against. Every claim it makes expires the moment the tab closes.
We store your measured runs, dated, with the engine, sample size and instrument version on every row. That is what makes a before-and-after a measurement instead of a memory, and it is why the answer to “did the fix work” on SEOAIO is a number with an interval rather than a feeling. What we store is the measurements; what we refuse to blend, gate, or resell is on the refusal page.
Why we never touch your site
Several tools in this market now measure your site, draft the fix, and publish it into your CMS. The better ones put a governance gate in front of that: a named human has to approve anything before it ships. That is a genuine safeguard and it is worth more than the tools that skip it.
It does not solve the problem it is aimed at. An approval step governs who signs off on a change. It cannot make the party that made the change an independent judge of whether the change worked, because the next measurement is now partly a measurement of their own work. Whoever approved it, the marker and the exam-sitter are still the same organisation.
So we do not write to your site. No plugin, no CMS connection, no publish step, nothing of ours in your repository. We read what is publicly there, sample the engines, and store what came back. When your next run moves, the movement is evidence about your change rather than about ours, and that is the whole reason the number is worth anything. It is the harder promise to keep and the only one that makes the measurement independent.
Run a measured audit on your site.
Free, no sign-up. Every number comes with its receipt.