Choosing a service
How to Read GEO Case Studies
Assess GEO case studies by baseline, scope, dates, measurement, and attribution limits. Distinguish documented observations from unsupported result claims.
TL;DR
- Check the baseline, sample, dates and exact outcome before reading the headline.
- Distinguish percentage points, relative change and a claim that the work caused the change.
- Use a case to judge method and fit; it cannot guarantee the same result for your business.
A GEO case study is useful when you can trace its claim back to the work and the data. Check what changed, when it changed and how it was counted. A strong headline is not enough to show that the same work will help your business.
Recalculate the headline before accepting the story
Ahrefs’ 2026 schema analysis estimated citation changes of −4.6% for AI Overviews, +2.4% for AI Mode and +2.2% for ChatGPT. Only the first was a clear change from zero in the statistical test. Its already-cited page sample and matched design do not establish cause. Pew’s AI-summary study is another example where observed behavior should not be turned into a causal claim.
Worked case-study reading
A made-up case says “AI visibility grew 50%.” The supplied material shows 20 of 100 usable answers before and 30 of 100 after, under the same stated sample rule. It also lists page edits made between the two observations.
| Calculation or evidence | Supported reading | What is still missing |
|---|---|---|
| 20% before; 30% after | Increase of 10 percentage points. | Whether sample conditions and answer coding stayed comparable. |
| (30 − 20) / 20 = 50% | Relative increase of 50%. | Whether random variation or other changes explain some of it. |
| Page edits between observations | A documented intervention occurred. | A design that isolates its causal effect. |
| No visit or sales records supplied | The case concerns sampled answer visibility. | Evidence for referral, qualified-lead or revenue claims. |
The gap is a headline that can be mathematically correct but easy to overread. The output is a qualified restatement: “In this sample, mentions rose from 20 to 30 out of 100 usable answers.” Accept the case as evidence of that observation only after checking raw records and conditions. Ask separately whether the method and page scope resemble your situation.
Start with the baseline and scope
Find the starting counts, dates, market, platforms and page set. Check which questions were asked and how the team chose them. If the sample changed, ask how the author made the two periods comparable.
Scope matters too. A result from one brand, market or page type may not fit yours. Keep that limit beside the headline. The AI visibility tracking guide shows which sample rules to check.
Separate the action from the observation
List what the team changed. Then list what the tools or sales records showed. These are two sets of facts.
A page edit may happen before a rise in mentions without being its sole cause. Other releases, demand shifts or changes in the platform may matter. A sound case explains what it can and cannot show.
Check the evidence and authorization
Look for source records, metric rules and dates. A public case should also have the right to name the client and show its work. Do not treat a stock chart or a made-up example as a client result.
Sementa’s delivery program, sample audit and analysis of a public report show different kinds of material. They do not all make the same kind of claim.
Use the case to ask better questions
Ask which parts of the method fit your offer, pages and team. Who would supply facts, approve changes, publish them and check the work? What would a fair baseline look like for you?
Use those answers in the brief. The agency selection guide helps turn them into questions for suppliers.
Frequently asked questions
Does Sementa present client results in this guide?
This guide explains how to assess evidence. The linked resources cover our delivery process, a sample audit, and an independent analysis of a third-party public report.
Can a strong case guarantee our result?
No. Use it to understand the process and context, then assess the scope and evidence relevant to your business.
Is a large percentage increase always impressive?
Check the starting count. A small baseline can produce a large relative change. Report the counts and percentage-point change beside it.
Does a before-and-after comparison prove cause?
No. It shows sequence. Other changes, sample variation or wider demand can contribute unless the design addresses them.
Bring the case-study claim and its supporting records to scope an evidence review.
