Home › Lab
Ultra SEO · AEO · GEOContinuous researchThe platform learns from every fix
The engines change behavior every week, warning nobody. Tracking that is not optional: it is what separates a number that holds from a number that aged.
Weekly retest
The same questions, in the same window, across the same six engines. Without that, one cycle and the next are not comparable.
Change lab
Every fix goes in measured: what it was before, what it became after, and across how many checks.
Entry criteria
A new page does not enter measurement the next day: Google starts at 7 days live and AI at 15. Measuring before that would give a number that means nothing.
The queue reorders itself
What proved to move citations rises in the fix queue. What did not move drops, even if the market agrees with it.
Why measuring alone teaches nothing
A tool that only measures watches the number change and never knows why: it did not make the change and does not know what it was. Here both sides are ours, so every fix becomes an experiment with an observed result.
Measures
The baseline, with a date and the total number of checks beside it.
Fixes
One change at a time, recorded with the layer it belongs to.
Remede
The same questions, the following week, by the same method.
Learns
What moved becomes a priority for every client, not just that one.
| Change applied | Layer | Citations | Before and after |
|---|---|---|---|
| Allow the AI reader through the CDN anti-bot | Access | +11 | from 4 to 15, in 30 checks |
| One page per location, with the address marked up | Programmatic | +7 | from 15 to 22, in 30 checks |
| Direct answer in the page first sentence | Content | +4 | from 22 to 26, in 30 checks |
| Change the body font and the spacing | Aesthetics | 0 | 26 before, 26 after, across 30 checks |
It is not a list of supported models
It is the behavior of each surface where the answer appears, retested every cycle.
- When an engine changes, the problem is ours. The agency does not rebuild its method with every update: the platform adjusts and the history stays comparable.
- When it changes enough to break the comparison, we say so. A curve that moved because of the engine, and not because of the work, is stated on screen instead of hidden.
- A new engine goes in when it returns data that holds up to comparison. One of them left after zero results in 327 queries.
What the lab is not
We have no privileged access
Nobody has it. The engines are closed boxes and do not publish how they choose what to cite. What we have is repeated measurement, from the outside, with a fixed method, which lets you infer rather than guess.
It is not a guarantee of results
Knowing what usually moves does not guarantee it will move in that case. That is why the number always ships with its total number of checks: so you can disagree with it using data, not opinion.