A baseline is the reference state of a page or piece of content captured before any experiment variant is generated or deployed — the unchanged “before” that every proposed change is measured against.
| Term | Baseline |
|---|---|
| Category | AI, Content Generation and Experiments |
| Also known as | Open Baseline |
| Where it appears | Experiments |
What it means in RankGear
Inside the Experiments workspace, RankGear records the current state of a page — its live copy, structure, and factor readings — the moment you open an experiment and before you render or publish any variant. That snapshot is the baseline. It fixes a single “starting point” so that when the AI drafts alternatives or you deploy a change, the comparison is like-for-like rather than against a target that has quietly shifted underneath you.
How to interpret it
Treat the baseline as a control, not a goal. Before you apply or publish anything, read the supplied context, the model you chose, the cost boundary, and the proposed output against it. Its usefulness is comparative: it tells you how far a variant has moved from the original and in which direction, not whether that move will earn a ranking. A larger gap from the baseline means a bigger change, not necessarily a better one.
Example
You open an experiment on a service page whose intro reads as generic. RankGear captures the baseline — the current copy and its factor profile. You ask the AI to critique the section and draft a tighter, entity-led opening. Before deploying, you set the drafted variant beside the baseline, see exactly which sentences and factor readings changed, revise the parts you disagree with, and only then publish.
Important considerations
- AI output can be incomplete or wrong. Verify factual claims, statistics, citations, and credentials in a variant before you let it replace the baseline.
- A baseline is only as clean as its isolation. Unrelated edits made to the page mid-experiment become confounds that blur the before-and-after comparison.
- Differences from a baseline are indicators of change, not proof of impact. Correlation is not causation, and any provider metrics shown are comparative indicators on their own scale — not Google scores, and not a promise that a variant will rank.
Related terms
Part of the RankGear glossary · how RankGear measures · the 870 factors.