GEO tutorial

AI Visibility Experiment Readout: connect changes to evidence and outcomes

A deployment is an event, not a result. An experiment readout connects the hypothesis and release to comparable evidence so the team can keep, revert or extend the change.

4 min read

Illustration for AI Visibility Experiment Readout: connect changes to evidence and outcomes, with no text.

Key takeaways

  • How to evaluate SEO and GEO experiments with transparent evidence.
  • A readout says whether to keep, revert, iterate or expand a change. It starts with a hypothesis and a window, and it lists what else moved at the same time. A case study is a story after the fact. An update log records that something changed. This page is the decision.
  • State the hypothesis, target query cohort, changed pages, expected mechanism and primary outcome before release.
  • If a title rewrite raises impressions but not qualified sessions, keep the visibility result separate from the conversion result and test the page promise next.
  • SEO and AI data have lag and noise.
Contents

What this workflow decides

How to evaluate SEO and GEO experiments with transparent evidence.

A deployment is an event, not a result. An experiment readout connects the hypothesis and release to comparable evidence so the team can keep, revert or extend the change.

A readout says whether to keep, revert, iterate or expand a change. It starts with a hypothesis and a window, and it lists what else moved at the same time. A case study is a story after the fact. An update log records that something changed. This page is the decision.

How to do it

State the hypothesis, target query cohort, changed pages, expected mechanism and primary outcome before release.

Write the hypothesis, the query cohort, the changed URLs and the primary outcome before release.

Record baseline impressions, clicks, position, AI mentions, citations, engagement and conversions with a fixed observation window.

Store the baseline for impressions, clicks, mentions, citations and conversions inside a fixed window.

Separate the test change from concurrent releases, seasonality, consent loss, outages and external algorithm changes.

List concurrent releases, seasonality, consent loss and algorithm news next to the result.

Compare results by page, query, language and device and report uncertainty instead of declaring causation from one movement.

Compare by page, query, language and device. Report uncertainty. Do not pick only the queries that rose.

Decide keep, revert, iterate or expand, then store the evidence and next owner in the release log.

Record keep, revert, iterate or expand, and name the next owner.

How to evaluate SEO and GEO experiments with transparent evidence.
A readout says whether to keep, revert, iterate or expand a change. It starts with a hypothesis and a window, and it lists what else moved at the same time. A case study is a story after the fact. An update log records that something changed. This page is the decision.

What a finished record looks like

If a title rewrite raises impressions but not qualified sessions, keep the visibility result separate from the conversion result and test the page promise next.

What to keep in the record

  • Hypothesis, cohort and primary outcome
  • Baseline and the fixed window
  • Confounders listed beside the result
  • Decision and the next owner
Hypothesis, cohort and primary outcome
Write the hypothesis, the query cohort, the changed URLs and the primary outcome before release.

What changes the conclusion

  • SEO and AI data have lag and noise.
  • Without a control or stable baseline, attribution is weaker.
  • Do not cherry-pick only positive queries.
SEO and AI data have lag and noise.
If a title rewrite raises impressions but not qualified sessions, keep the visibility result separate from the conversion result and test the page promise next.

Frequently asked questions

Is a deploy a result?

No. A deploy is the event. The readout is the comparison.

Can one movement prove cause?

No. Without a stable baseline the attribution is weaker, and you should say so.

May the readout show only the queries that improved?

No.

Sources

These are the official sources for the part of the workflow they govern. A documented workflow is not a ranking or a citation outcome.

  1. Google: Search Console performance report

Start a free AI visibility audit

Open ViziGEO and review a documented sample of questions, mentions, recommendations and citations. The audit records what was observed. It does not guarantee a mention, a citation or a recommendation.