The features that lift citations
Median citations-per-page lift, pages with the feature vs without: specific statistics 1.92x (6 audits); named author 1.91x (6 audits); comparison table 1.44x (6 audits); structured data 1.19x (5 audits); visible pricing 1.05x (6 audits); visible publish date 1.02x (5 audits).
Data table
| Series | median lift |
|---|---|
| Specific statistics | 1.9 |
| Named author | 1.9 |
| Comparison table | 1.4 |
| Structured data | 1.2 |
| Visible pricing | 1.1 |
| Visible publish date | 1 |
| Specifications | 1 |
| Direct answer in first 100 words | 0.8 |
What the citable page already looks like
Share of AI-cited pages carrying each trait (median across audits): direct answer in first 100 words 35%; named author 34%; visible pricing 31%; visible publish date 28%; comparison table 27%; updated in last 12 months 25%.
Data table
| Series | share of cited pages |
|---|---|
| Direct answer in first 100 words | 34.9% |
| Named author | 33.5% |
| Visible pricing | 30.6% |
| Visible publish date | 28% |
| Comparison table | 27.4% |
| Updated in last 12 months | 25.4% |
| Specific statistics | 24.9% |
| Specifications | 19.7% |
| Structured data | 18.6% |
| Cites its own sources | 3.2% |
| Article schema | 1.9% |
| Original data or research | 1.6% |
| FAQ schema | 0.4% |
The formats that win
Median citations per profiled page by content format: mixed 4.8/page; prose 4.5/page; bulleted 3.9/page.
Data table
| Series | citations/page |
|---|---|
| mixed | 4.8 |
| prose | 4.5 |
| bulleted | 3.9 |
How long cited content runs
Median citations per page by word count: under 500 words 2.4/page; 500-1,000 words 3.2/page; 1,000-2,000 words 6.3/page; over 2,000 words 7.1/page.
Data table
| Series | citations/page |
|---|---|
| under 500 words | 2.4 |
| 500-1,000 words | 3.2 |
| 1,000-2,000 words | 6.3 |
| over 2,000 words | 7.1 |
Which content types earn citations
Median citations per profiled page by content type: listicle 8.7/page; review 7.0/page; comparison 6.7/page; category-page 6.5/page; editorial 5.3/page; pricing 5.3/page.
Data table
| Series | citations/page |
|---|---|
| listicle | 8.7 |
| review | 7 |
| comparison | 6.7 |
| category-page | 6.5 |
| editorial | 5.3 |
| pricing | 5.3 |
| news | 4.7 |
| how-to | 4.6 |
| product-page | 3.1 |
| press-release | 2.5 |
| faq | 2.5 |
The words in AI-cited titles
Terms most common in third-party cited page titles: "best" (35.2% of titles); "2026" (33.8% of titles); "kitchen" (23.6% of titles); "shoes" (21.8% of titles); "running" (19.0% of titles); "running shoes" (17.6% of titles); "ireland" (13.4% of titles); "guide" (12.0% of titles).
Data table
| Series | share of titles |
|---|---|
| best | 35.2% |
| 2026 | 33.8% |
| kitchen | 23.6% |
| shoes | 21.8% |
| running | 19% |
| running shoes | 17.6% |
| ireland | 13.4% |
| guide | 12% |
| cruise | 12% |
| de | 9.7% |
| the best | 8.8% |
| shoes for | 8.3% |
| deals | 7.4% |
| in 2026 | 6.9% |
| kitchens | 6.9% |
Schema on cited pages
Structured-data types on AI-cited pages (median share of pages): BreadcrumbList 13%; Organization 7%; WebSite 6%; ItemList 4%.
Data table
| Series | share of cited pages |
|---|---|
| BreadcrumbList | 12.9% |
| Organization | 7.4% |
| WebSite | 5.8% |
| ItemList | 4.5% |
How models actually quote
The passages citations trace to run a median 131 words; 76% contain at least one number, and 19% were matched to the AI answer verbatim (across 5 attribution-depth audits).
The median cited page answers 3.0 questions outright, and 90% of cited pages answer at least one question directly.
Do the models agree?
A median 44% of profiled cited pages are cited by more than one model, and those consensus pages earn 2.8x the citations of single-model pages.
At domain level: 112,624 domains cited by one model only carry 15.8% of citation volume; 45,070 domains cited by 2-4 models carry 28.9% of citation volume; 7,237 domains cited by 5+ models carry 55.3% of citation volume.
Data table
| Series | share of citations |
|---|---|
| cited by one model only | 15.8% |
| cited by 2-4 models | 28.9% |
| cited by 5+ models | 55.3% |
How much brands actually control
Across 6 audited projects, the median brand holds 14% of the citations it could control, while third parties earn 83% of all citations — the earned layer, not the owned one, decides most AI visibility.
Method & limitations
- Benchmarks are medians over 16 completed content audits (6 projects) — cumulative evidence to date, not a monthly movement.
- Feature lifts and prevalence figures are correlational; the audits observe pages as they are, they do not run controlled experiments.
- Title terms and schema types come exclusively from public third-party pages, and only where they recur across many pages and domains.
- Every audit ran the same profiler, feature detection and denominators, so comparisons are like for like.
How every figure on this page is counted, floored and reviewed: our methodology.