Reflexa records a baseline the moment a change is applied, then checks back fourteen days later and tells you plainly whether it moved, stayed flat, or made things worse. Most tools stop at "applied". This is the step after that, and almost nobody else in this category does it.
This is the follow-through on everything else the Reflexa agent does on your site: not just whether a change can be made and approved, but whether it actually did anything once it landed. Below is what gets recorded, why fourteen days, and why a dashboard score cannot answer this on its own.
Because landing a change and that change moving anything are two different facts, and most tools in this category only report the first one. A title gets rewritten, a switch gets flipped, a redirect goes live, and the tool marks the job done. Whether an AI engine answered differently afterward, whether Google showed the page any differently, whether the underlying check started passing, is a separate question that "applied" never actually answers.
Why it matters: a log of changes made is not the same thing as a log of changes that helped, and only the second one tells you what to do next.
Three things, tied to the one page the change touched, not to the site as a whole. At the moment a change is applied, the agent records what AI did with that page over the previous fourteen days, what Google showed for it, and how the specific check that change answers stood. Fourteen days later it reads the same three again and compares.
| What is recorded | Read again after |
|---|---|
| What AI did with that page over the previous 14 days | 14 days |
| What Google showed for that page | 14 days |
| How the check the change answers stood | 14 days |
Do this: for what the agent can change in the first place, a switch, page text, or anything else you ask for, see Can an AI agent actually change my website?
Fourteen days is long enough for the page to be recrawled by the engines and search surfaces that matter, and short enough that the result is still about this change, not everything else that has happened on the site since. Checking the next day would catch almost nothing, since a page has to be fetched again before anything can shift. Waiting a quarter would mean other edits, other pages, and ordinary drift are all mixed into the same reading, and the verdict would no longer be about the one change it was meant to judge.
Why it matters: a window chosen to be fast enough to feel useful and slow enough to be honest is the whole point, a shorter one would be noise and a longer one would be guessing.
Because past a certain point, too much else has changed for the result to still belong to that one edit. A page gets crawled again, other fixes land, competitors change their own pages, and the site itself moves on. A change older than ninety days is not scored moved, flat, or worse, because by then a verdict would be attributing a result to a cause the agent can no longer isolate with any honesty.
Bottom line: a measurement window has to end somewhere, and ending it rather than stretching a guess indefinitely is the more honest choice.
A plain sentence with the figures behind it, not a single score. "Moved" means the figures the change was meant to affect are measurably better after fourteen days than the baseline recorded when it was applied. "Flat" means they are effectively unchanged. "Worse" means they went the wrong way. Any figure that cannot be read yet, because a counter is not connected or a check has not run again, is reported as not measurable, never folded in as a zero.
Why it matters: a missing number and a bad number look identical on a dashboard that only shows totals; they are never the same thing, and treating them the same leads to the wrong next step.
It isn't, not without saying plainly what is different this time. A change the agent has already seen do nothing is on record as having done nothing, so proposing the identical fix again, with no new reason to expect a different result, is treated as a waste of a click rather than a fresh recommendation.
No, not on its own. A readiness or visibility score is a snapshot of the whole site at one moment. It has no baseline tied to the one page a specific change touched, so it has nothing to compare that change against, and it cannot tell you whether the thing you just approved is the reason anything moved. This is the part a score cannot do: attribute a result to a cause.
Why it matters: if the tool you are paying for can only tell you the site's current number, not whether your last change helped, it is answering a different question than the one you asked.
None of this requires buying anything, and each step takes a few minutes:
Reflexa records a baseline the moment the change is applied, what AI did with that page over the previous fourteen days, what Google showed for it, and how the check it answers stood, then reads the same three again fourteen days later and returns a verdict: moved, flat, or worse.
Fourteen days gives engines and Google enough time to recrawl the page and for day-to-day noise to settle, while the result is still close enough to the change to be attributed to it rather than to everything else that happened on the site since.
It is reported as not measurable, never as a zero. An absent counter or a check that has not run again yet is a gap in the data, not evidence that nothing happened.
No. A change already seen to do nothing is not proposed again without saying plainly what is different this time, rather than repeating a request that is already on record as not having worked.
Not on its own. A score is a snapshot of the whole site at one moment, with no baseline tied to one specific edit, so it has nothing to compare a change against and cannot tell you what moved because of it.
The free check reads your site the way ChatGPT, Claude and Perplexity read it, the same starting point the agent records before any change. 3 minutes, evidence included.
The product: Reflexa Agent · how it works · på svenska