The directory tax is a myth: G2 and Capterra are #1 in 0 of 835 categories
The common generative-engine-optimization advice says one directory listing can control whether AI recommends you. In Parse's category evidence, the directory chokepoint mostly is not there.
By Dimitry Apollonsky · July 8, 2026 · 8 min read
Contents
- There is no single source chokepoint in the median category
- The major directories are almost never the #1 source
- Three software brands in four are recommended with no directory citation
- Directory citations are a tiny slice of recommendation evidence
- When a category does lean on one source, it is usually community or video
- The category leaders are social, editorial, and owned sites
- Mainstream software categories are spread out too
- Both engines spread evidence across many sources
- True chokepoints exist, but they are outliers
- The practical playbook is a spread of sources, not a directory tax
- Get the data
- Sources
- Related research
We analyzed 5.42M evidence links, the cited pages that sit behind AI brand recommendations, across 835 monitored categories, 49,638 brands, 7,611 prompts, and 101,677 source domains.
There is no single source chokepoint in the median category
Across 835 monitored categories, the median category's single most-cited source domain accounts for only 7.4% of recommendation evidence. The median category draws on 260 distinct domains.
That is the opposite of a tax. In most categories, AI recommendations are assembled from a wide range of sources, not unlocked by one mandatory listing.
Takeaway
The major directories are almost never the #1 source
G2, Capterra, GetApp, Software Advice, SourceForge, TrustRadius, and Trustpilot are each the #1 source in zero of the 835 categories. Clutch is #1 in exactly one. By source type, a directory is #1 in only 8 categories.
Where the major directories do appear, their average shares are small: G2 appears in 333 categories at 0.3%, Capterra in 186 at 0.2%, GetApp in 174 at 0.4%, and Clutch in 35 at 0.9%.
| 1 | |
| 0 | |
| 0 | |
| 0 | |
| 0 | |
| 0 | |
| 0 | |
| 0 |
Three software brands in four are recommended with no directory citation
The charitable test is whether a directory citation is present anywhere behind recommended brands. Even there, the directory story is weak. Among all brands with at least 20 evidence links, 14.2% carry any directory citation. In software categories, where the directory premise is strongest, 24.0% do.
Takeaway
Directory citations are a tiny slice of recommendation evidence
Directory citations make up 0.74% of the recommendation evidence for brands in software categories, and 0.51% across all brands. That is not enough to explain recommendation visibility as a directory-led system.
When a category does lean on one source, it is usually community or video
The domains that do become #1 sources are more often 
youtube.com169
reddit.com142
The category leaders are social, editorial, and owned sites
By source type, social sources lead 321 categories, general editorial leads 230, and a brand's own site leads 129. Directories lead only 8. That mix says your source work should start with the pages, communities, and owned assets AI already uses in the category.
| Social | 321 |
| General/editorial | 230 |
| Brand or vendor site | 129 |
| News and finance | 34 |
| Ecommerce | 30 |
| Directory | 8 |
Mainstream software categories are spread out too
The finding is not only a side effect of small, obscure categories. CRM and Sales Pipeline Software draws on 852 domains, with 
| Payroll and HRIS | 853 | 8.8 | |
| CRM and Sales Pipeline Software | 852 | 8.3 | |
| Small-business accounting | 806 | 9.9 | |
| Email marketing | 715 | 6.6 | |
| Project management | 517 | 6.2 |
Both engines spread evidence across many sources
True chokepoints exist, but they are outliers
Across all 835 categories, the top-source share exceeds 14% only in the top 10% of categories. The single most concentrated category reached 53.2%, an accessories category dominated by
The practical playbook is a spread of sources, not a directory tax
Directory work is not useless. It is just too narrow to be the default explanation for AI recommendations. In many categories, the pages that matter most will be
Start by mapping the actual top domains in your category, then prioritize the small set of sources AI repeatedly uses. Treat directories as one kind of source among many, not the toll booth for the whole market.