GEO INSIGHT
30-Item GEO Checklist — From Question Design to Remeasurement
We organized the 30-item GEO analysis checklist into seven groups: goals, question design, measurement conditions, classification, competitors, site, and remeasurement. It includes the reason for each item, the recording format, and the next action for each result combination.
Quick answer
Practical GEO analysis is not merely a task of verifying whether a brand appeared in the AI response. You must first establish business goals and target questions, and then record whether brand mentions, source citations, and recommendations are included.
Review raw data by question and platform first, rather than the overall average, and compare recurring competitor brands and source types with the technical and content gaps on your own site.
Determine the question set before publishing content, analyze the site structure, content, and external sources together, and reinforce pages connected to important questions first. After the work is completed, re-measure using the same questions and conditions.
The purpose of the checklist is not to increase the number of checks, but to determine which questions and pages to reinforce first.
Fixing Questions and Conditions
Record the full question, platform, search mode, and measurement time to allow comparison between the initial measurement and the re-measurement.
Separating Judgment Metrics
Record brand mentions, official source citations, and recommendation inclusions separately, rather than combining them into a single impression metric.
From Analysis to Execution
Connect competitor and source observations to actual reinforcement work on website structure, original content, and internal links.
Recording based on non-exposure
We do not collect only positive results; we also record unmentioned and uncited responses, erroneous responses, and platform differences as grounds for future work.
Quick Conclusion
What this checklist covers and does not cover
GEO Checklist 30 Items
Three judgments to fix before starting the check
Measurement Record Form
Determining the next action based on result combinations
Remeasurement: What to Preserve and What Not to Claim
How to Document E-E-A-T Evidence
Conclusion — A good checklist makes the next steps clear
GEO Checklist for Practitioners
- ✓ [Analysis Objective] Have you defined the business objectives you intend to verify in this analysis?
- ✓ [Analysis Objective] Have you identified the key conversions among Inquiry, Comparison, Recommendation, and Brand Understanding?
- ✓ [Question Design] Have you distinguished between defining, comparing, and recommending questions?
- ✓ [Question Design] Have you fixed the question set and versions?
- ✓ [Question Design] Have you included non-branded questions that do not directly involve the brand?
- ✓ [Measurement Criteria] Did you record the platform and measurement date?
- ✓ [Measurement Criteria] Did you record whether web search or search mode was used?
- ✓ [Measurement Criteria] Was it verified in an independent conversation or session?
- ✓ [Measurement Criteria] Did you not count error responses as non-exposures?
- ✓ [Decision] Did you distinguish between brand mentions and source citations?
- ✓ [Decision] Did you distinguish between recommendation inclusions and simple mentions?
- ✓ [Decision] Did you save the actual cited URL?
- ✓ [Decision] Did you leave a screenshot of the evidence for positive results?
- ✓ [Decision] Did you also record unmentioned and uncited results?
- ✓ [Competitors/Sources] Have you identified competitors that appear repeatedly?
- ✓ [Competitors/Sources] Have you classified the domains and source types of repeated citations?
- ✓ [Competitors/Sources] Have you compared the content gaps between your company and your competitors?
- ✓ [Site/Content] Is the official brand name consistent?
- ✓ [Site/Content] Does an official page exist that connects to the core question?
- ✓ [Site/Content] Are the title, description, and canonical clear?
- ✓ [Site/Content] Is necessary structured data, such as Organization and Article, available?
- ✓ [Site/Content] Are the author, publication date, modification date, and source visible?
- ✓ [Site·Content] Are related posts linked via internal links?
- ✓ [Site·Content] Is similar content not duplicated across multiple domains?
- ✓ [Site·Content] Do the actual screen content match the structured data?
- ✓ [Run·Remeasure] Have you prioritized based on impact and difficulty?
- ✓ [Run·Remeasure] Have you recorded the person responsible for the fix and the completion date?
- ✓ [Run·Remeasure] Have you remeasured using the same questions and conditions?
- ✓ [Run·Remeasure] Have you distinguished between newly cited URLs and weak questions?
- ✓ [Run·Remeasure] Have you documented the next reinforcement plan?
Frequently asked questions
How many items are sufficient for the GEO checklist?
The order is more important than the number of items. Although this document organizes 30 items, the key is not to mix the 9 items (Goals, Questions, Conditions) that must be completed before measurement, the 8 items (Assessments, Competitors) filled out during measurement, and the 13 items (Sites, Executions) processed after measurement. If you skip the pre-measurement items, you cannot compare with the previous month no matter how many of the remaining 21 items you fill out.
Why must brand mentions and source citations be counted separately?
Because the subsequent tasks are different. If there is a mention but no citation, the name is known but there is no document to use as evidence, so work is required to create original information. If there is a citation but no mention, the document is being used but is not connected to the brand, so work is required to organize the attribution of the publisher and author. You cannot determine which side it is based solely on the 'Impression Rate,' which combines the two values.
How do I handle error responses?
Do not count them as non-exposures; leave them as separate statuses. If you include error, discontinuation, or rejection responses in the denominator, the rate will be lower than the actual figure. You must also distinguish 'unidentified valid responses' from 'uncited.' This is because the former represents a state that cannot be verified, while the latter represents the result of verification.
What should I do if I want to change a question set?
Do not delete existing questions; instead, create a new version. If you delete questions because the results are low, you lose the basis for comparing the before and after of improvements. Keep the version number, added or modified questions, the reason for the change, and the application date, and do not connect a trend line at the point where the version changed.
If the measurement results are low, should I fix the content first?
Please check the access conditions first. Access paths vary by platform. OpenAI separates OAI-SearchBot for search display and GPTBot for training, and states that robots.txt may not apply to queries initiated by users. Perplexity also operates PerplexityBot for indexing and Perplexity-User for user request queries separately. Even if you increase content while access is blocked, the results will not change.
Does applying JSON-LD increase AI citations?
There is no dedicated structured data for exposing AI features. Google states in its documentation that "there are no additional requirements to appear in AI Overviews or AI Mode, and no special optimization is needed." Structured data is a format that conveys the document type and publisher; if included, it must match the content visible on the screen. Marking up invisible content may result in the loss of Rich Results eligibility through manual intervention.
Does increasing E-E-A-T improve rankings?
Google explicitly states that E-E-A-T itself is not a specific ranking factor. It explains that Trust is the most important of the four elements, and the others contribute to it, advising that these criteria be used as questions to self-evaluate your content. Therefore, what you should leave behind is a record, not a mark. It is sufficient if the screen shows who wrote what, when, on what grounds, and what could not be verified.
How often do you remeasure?
It is more practical to use events as a standard rather than a fixed cycle as an absolute standard. These include when a major page is modified, when new content is published and indexed, when a question set version is changed, and during the monthly operational check. Remeasuring before confirming indexing results in measuring time lag rather than the effectiveness of the work.
What Does a GEO Practitioner Need to Do?
First, you must fix the goals, questions, platforms, and evaluation criteria, and record brand mentions, official URL citations, and recommendations separately. Identifying unexposed questions and insufficient official pages from the results by question and platform is the starting point for reinforcement.
After reinforcing content to fill the identified gaps and applying website structuring, you must measure again using the same questions and conditions. By repeating this process, GEO becomes an operation of continuous measurement and reinforcement rather than a one-off content publication.
Continue reading
Sources and references
- 1. Creating helpful, reliable, people-first content — Google Search Central · Check 2026-08-23
- 2. Structured data general guidelines — Google Search Central · Check 2026-08-23
- 3. AI Features and Your Website — Google Search Central · Check 2026-08-23
- 4. Google Search Document Change History — FAQ End of support for rich results — Google Search Central · Check 2026-08-23
- 5. Generative AI performance report (Search) — Google Search Console Help Center · Check 2026-08-23
- 6. Bots / Crawlers — OpenAI Developers · Check 2026-08-23
- 7. Perplexity Crawlers — Perplexity Docs · Check 2026-08-23
AI search answers and citation sources may vary depending on the question, time, search results, and platform status. Record the question, platform, and time, and measure impressions, citations, and recommendations repeatedly using the same criteria. All official documents cited in this document were verified on 2026, 8, 23.