October 2026 Benchmark
Korean SaaS AI Visibility Benchmark
A new baseline: the same 12 Korean B2B SaaS domains re-measured after a measurement-engine revision (measured 2026-10-16)
Key findings
- • Across 12 domains the median score (p50) is 72 out of 100 and the top 10% reach 76 — 10 at grade A and 2 never mentioned (grade D), a clear split between the top and the bottom
- • All three platforms mentioned the brands at the same 83% rate, but own-domain citation rates differed widely — ChatGPT 83%, Perplexity 67%, Gemini 33%
- • Across 3 same-day rounds, individual scores differed by at most 6 points (median 2) and no domain changed grade
A new baseline from this issue
We revised the measurement engine after the August issue. This issue re-measures the same 12 domains with the revised engine and is not directly comparable with issues through August. From this issue the benchmark is published quarterly.
Sample and measurement conditions
The 12 Korean B2B SaaS domains selected for the July issue were kept unchanged, along with the inputs. Individual brands and domains are not disclosed (anonymized aggregate).
Inclusion: B2B SaaS operated by Korean companies · public website · spread across 8 sub-verticals · public recognition (news/funding history). Exclusion: consumer-first services, Korean branches of global vendors, and RanketAI itself (conflict of interest). The sample and inputs are held fixed from the 2026-07 issue, and for repeated runs per domain only the latest valid run enters the distribution.
Major LLMs (ChatGPT · Gemini · Perplexity) · diverse purchase/comparison/discovery prompt combinations · multi-signal axes — measurement engine revised after the 2026-08 issue · See the full methodology
Based on live answers from ChatGPT · Gemini · Perplexity · measured 2026-10-16 · 3 rounds on the same day
Score distribution
Category score percentiles (0-100) and grade distribution. Grades run on four levels (A/B/C/D).
| Median (p50) | Top 25% (p75) | Top 10% (p90) |
|---|---|---|
| 72 | 74 | 76 |
Grade distribution
| A | B | C | D |
|---|---|---|---|
| 10 | 0 | 0 | 2 |
Platform comparison
Even for the same sample, mentions and citations differ sharply by platform. The denominator is the 12 valid measurements per platform.
| Platform | Brand mention rate | Own-domain citation rate |
|---|---|---|
| ChatGPT | 83% | 83% |
| Perplexity | 83% | 67% |
| Gemini | 83% | 33% |
All three platforms mentioned the brands at the same rate, yet the share of brands whose own domain was cited as evidence ranged from 83% on ChatGPT to 33% on Gemini. Being mentioned and being cited move independently.
Citation source patterns
Top third-party sources cited in AI answers (sample domains excluded). Alongside blog platforms, wiki-style pages, and comparison sites, official product pages of competing and alternative products rose into the top tier.
| Host | Citations |
|---|---|
| impactflow.kr | 12 |
| daouoffice.com | 8 |
| blog.naver.com | 7 |
| naver.worksmobile.com | 7 |
| tistory.com | 7 |
| global.it-trend.jp | 6 |
| microsoft.com | 6 |
| namu.wiki | 6 |
| support.microsoft.com | 5 |
| workspace.google.com | 5 |
Interpretation and priorities
- • Observation — well-known brands were mentioned by all three platforms in most answers. The difference came less from whether a brand was mentioned than from which documents were cited as evidence.
- • Observation — official pages of competing and alternative products rose into the top cited sources. When building comparison answers, the platforms appear to consult each product’s official pages directly.
- • Priority — organize comparison facts such as features, pricing, and target customers on your official pages in a citable form, and check whether your information exists on the leading source types (blog platforms, wiki-style pages, comparison sites). This is not a guarantee of causation.
Frequently asked questions
Where does Korean B2B SaaS AI visibility stand now?
For this sample of 12 domains the median score is 72, the top 10% reach 76, 10 domains sit at grade A, and 2 were never mentioned (grade D). This is an aggregate of a public web sample, so do not generalize it to the whole industry.
Scores move between runs — how reliable are they?
Across 3 same-day rounds individual scores differed by at most 6 points (median 2) and no domain changed grade. Because this sample is weighted toward well-known brands, brands scoring near a grade boundary can move more. Reading grades and repeated-measurement trends is safer than reading a single score.
Why are scores so much higher than in the August issue?
Because we revised the measurement engine after the August issue. It does not mean the brands became that much more visible, so read this issue as a new baseline rather than comparing it with August.
Why do results differ by platform?
Each platform draws on different sources and cites them differently. In this measurement the three platforms mentioned brands at the same rate, but own-domain citation rates differed widely — ChatGPT 83%, Perplexity 67%, Gemini 33%.
Can we see our own ranking?
No. This report is a fully anonymized aggregate and does not disclose individual brand rankings. To find your position, measure your own domain and compare its grade (A-D) against the distribution.
What kinds of content get cited in AI answers?
In this measurement blog platforms (Naver Blog, Tistory), wiki-style pages, comparison sites, and official pages of competing and alternative products were the leading cited sources. Third-party surfaces outweigh own domains; the full list is in the citation source table above.
Limitations
- • Aggregated from 12 domains sampled from the public web — not product telemetry and not a representative sample of the whole industry.
- • The measurement engine was revised after the 2026-08 issue. This issue is a new baseline and is not directly comparable with issues through 2026-08.
- • Across 3 same-day rounds, individual scores differed by at most 6 points (median 2) and no domain changed grade. Because the sample is weighted toward well-known brands (10 at grade A), brands scoring near a grade boundary can move more than this on re-measurement.
- • Based on a single measurement date (2026-10-16); with this sample size one domain equals 8.3 percentage points. Sub-vertical breakdowns and fine-grained percentile reads are limited.
Data download
Where would your domain land in its category?
Measure how AI answers mention and cite your brand under the same conditions, and see where your grade (A-D) sits in the category distribution.
Measure brand visibility →