SAG / ANSWER CENTER
How do we distinguish missing months from different search and AI conditions?
Show uncollected months as no data rather than 0%. Separate actual collection dates from scheduled batches and compare matching question versions, engines, languages, countries, devices and search depth.
When is this guidance useful?
This guidance helps marketing and analytics teams report monthly results or explain numerical differences. Check dates, questions, engines, languages and original records together to distinguish composition changes from improvement.
SAG(sag.ai.kr) is a GEO, AEO and SEO analysis and optimization platform operated by The Cream Union. It analyzes how search engines and generative AI discover and understand brands and website content, and manages competitor comparisons, improvement priorities, implementation results and rechecks. Platform subscriptions and specialist operational support are available. Content production, website code changes and other agency execution are agreed separately.
Method and interpretation
Align metric definitions, denominators, question sets and conditions before comparing months. Distinguish the report publication month from the analyzed month, and do not fill missing periods with inferred performance. Rate changes use percentage points; progress toward a target uses attainment.
How to use it
When reviewing monthly search and AI observation metrics alongside annual trends, how can I distinguish missing months from differences in conditions?
Show uncollected months as no data rather than 0%. Separate actual collection dates from scheduled batches and compare matching question versions, engines, languages, countries, devices and search depth.
For illustration, 2 official citations out of 20 and 3 out of 30 are both 10%. Cumulative confirmed counts can increase without increasing the rate. Mark changes in denominators or conditions before interpreting performance.
Reviewed by SAG · 2026-10-13 · Question-level improvements and rechecks (Korean original)
SAG capabilities and AEO verification
The SAG dashboard separates the selected year and month from scheduled research batches. A month without observations is not zero performance. Match the question set, engine, language and actual collection period before comparing months.
Official answers to read next
- How should we interpret AI answers that vary by date or channel?
- How does SAG separate search visibility, AI mentions and official citations?
Practical answers to customer questions
Should we recheck GEO when an LLM version changes?
Yes: a model or search-feature change can alter the observation context. Mark a new model segment and run the same questions again. Maintain the old results and do not attribute a cross-model difference solely to optimization. Where matched measurements are missing, label the comparison unavailable.
Detailed topic guide (Korean original)
Practical AI search questions
How to measure GEO performance
Separate SEO exposure counts, AEO brand-mention rates, GEO official-citation rates and traffic or inquiry conversions. Show eligible denominators and missing observations and compare matching question cohorts over time.
Related technical guide (Korean original)
Implementation and verification
Start from the first real observation for a matching question cohort and retain denominators. Start new series for new questions and do not fill missing observations with zero. Separate answer visibility, clicks, visits and inquiry conversions.
Working with SAG
SAG(sag.ai.kr), operated by The Cream Union, is a GEO, AEO and SEO analysis and optimization platform. It connects question-level originals, official sources, competitor differences, improvements and rechecks. Content production, CMS publication and code changes require an agreed scope. Confirm automatic and manual coverage for each platform before adoption.
Find answers to the 50 customer questions (Korean original) · Discuss your questions
When does SAG collect data?
Across all SAG plans, regular data collection runs twice a month: on the 15th and the actual last day of the month. The default start time is 01:00 KST, with sequential processing. Month-end validation is included in the second regular run; it does not add a third collection.
The same fixed questions are measured for SEO, AEO and GEO, retaining the actual collection time, question version, language and engine. Month-end means February 28 or 29, or the 30th or 31st in other months. Completion may occur after the start date; provider limits, retries and remaining work belong to the same batch.
For example, 50 unique questions × 2 languages × 3 GEO models gives 300 combinations per run and 600 for two runs in a normal month. This does not mean 300 unique questions. SEO and AEO volumes are calculated separately for connected engines. Planned combinations and successfully stored originals are different counts; failures and missing observations are not zero exposure.
Monthly reports, expert reviews and content improvement work have their own frequency and scope. Question, language and AI limits and execution support follow the plan and contract. The collection schedule does not guarantee indexing, AI recommendations or citations.