Not sure where to start? Take a guided tour.
AI Site Grader
AI Query Scanner
llms.txt Generator
In short
Four findings, strongest first- 20.2%
of AI citations go to blog and editorial pages, from 14.41% of crawler visits.
Chapter 02 - 0.49×
citation efficiency for product pages: AI crawls them heavily and rarely cites them.
Chapter 03 - 1.84%
of citations point at the brand’s own site. The other 98.16% go to third parties.
Chapter 05 - 11.4M
crawler visits cross-referenced with 337K AI citations across 882 brands.
Method
AI crawlers read your product pages far more than anything else. AI answers cite your blog.
To measure that mismatch, we use a single metric: citation efficiency. It captures how well each page type turns crawler attention into AI citations. A score of 1.0× means a page type is cited in proportion to how often it gets crawled. Above 1.0×, the content outperforms. Below 1.0×, AI reads the pages but doesn’t recommend them.
AI reads one thing and cites another.
The mismatch
Product pages receive 13% of all crawler visits but earn just 6.3% of citations. AI reads your catalog obsessively. When it comes time to recommend, it points somewhere else.
Blog content shows the inverse: 14.4% of crawls, 20.2% of citations. Review sites punch 5x above their weight. FAQ pages get read but never recommended. About pages - rarely crawled - get cited far more than expected.
The pages AI reads most are not the pages it recommends. This gap defines where content investment creates real visibility.
Thirteen page types, ranked by efficiency.
Citation efficiency for all 13 classifiable page types. Scores above 1.0x mean the content earns more citations than its crawl traffic would predict. Below 1.0x, AI reads the pages but rarely recommends them.
The five overperformers share a trait: they provide context, opinion, or authority rather than product information. AI engines use them as evidence for recommendations, not as the recommendations themselves.
One workhorse, one laggard, one ghost.
Three page types tell the full story of citation efficiency. One converts crawler attention into recommendations at 1.4x. One wastes half its traffic. And one is nearly invisible.
One in five AI citations lands on a blog post. No other page type comes close. Blog content is also the stickiest - once AI cites a post, it tends to cite it again across an average of 8 different queries.
AI crawlers are obsessed with your catalog - 13% of all visits go to product pages. But only 6.3% of citations point there. Product pages are read twice as often as they're recommended.
The lowest citation efficiency of any page type in the study. FAQ pages receive 1.9% of crawl traffic but earn just 0.4% of citations. AI reads your FAQ for context but never surfaces it as an answer.
Four crawlers, four reading styles.
Each AI has a reading style
GPTBot obsessively scans product catalogs - 84% of its visits land on product pages. Meta's crawler is even more extreme at 96%. OpenAI Search is lighter but still product-focused at 77%.
ClaudeBot is the outlier. It's the only crawler that reads broadly across page types - products, homepages, blogs, and categories all receive meaningful attention. "AI-optimized" isn't one strategy. It's four.
Optimizing for ChatGPT means product page structure. Optimizing for Claude means content diversity. There is no single "AI-optimized" strategy - only specific ones.
Your own pages get kinder treatment.
Your own content gets better treatment
Only 1.8% of AI citations point to your own domain. The rest cite third-party sources about your category. But that small slice tells an important story.
When AI cites your own content, sentiment averages 76/100 - a +18 point lift versus third-party citations (58/100). Owned citations don't just increase visibility - they improve how AI portrays you.
Wikipedia alone captures 9.6% of all third-party citations. For most brands, the citation competition isn't competitors - it's the encyclopedia.
Your homepage is the stickiest citation. Once AI starts citing your homepage, it averages 35x repeat appearances. Invest in making your homepage AI-readable.
Sleepers and dead zones.
Beyond the big three, the matrix reveals outliers - page types with surprisingly high or low citation efficiency that most content strategies overlook.
The quiet overperformer. Solutions pages that frame your product around what it solves - not what it does - are cited nearly 2x more than they're crawled.
The highest citation efficiency in the entire study. G2, Capterra, and Trustpilot are almost never crawled but frequently cited as authority signals by AI.
AI cites "who you are" far more than expected. Your about page isn't just for humans - it's a key AI trust signal.
Whitepapers, templates, and research reports earn citations far beyond their crawl share. Authority content punches well above its weight.
Surprisingly low despite high SEO value. AI engines may avoid brand-created comparisons, preferring third-party sources for impartiality.
Dead weight for citations. AI reads them but won't recommend them. Pure compliance, zero visibility value.
Six moves the data supports.
The citation efficiency data points to six investment priorities, each grounded in specific findings from above.
Invest in blog content
1.4x20.2% of all AI citations point to blog content, with the highest repeat rate (8.0x) of any page type. Authoritative, data-rich editorial earns both first citations and repeat appearances across queries.
Enrich product pages
0.5xProduct pages receive 13% of crawls but earn only 6.3% of citations. Add expert context, use-case framing, and structured data. AI wants to recommend products - give it reasons to.
Build use-case pages
1.9xService and use-case pages are the quiet overperformer. Pages that explain what you solve - not what you sell - are cited nearly 2x more than they're crawled.
Don't count on FAQ
0.2xThe lowest citation efficiency of any page type. FAQ pages receive 1.9% of crawl visits but just 0.4% of citations. Use FAQ for user experience and support cost reduction, not AI visibility.
Get on review sites
4.8xG2, Capterra, and Trustpilot earn AI citations far beyond their crawl share. Third-party credibility signals matter more to AI than first-party claims.
Know your AI audience
GPTBot reads 84% product pages. ClaudeBot reads broadly across types. Meta only sees products. Your optimization strategy depends on which AI engines your audience actually uses.
How we measured it.
Data sources
337,362 unique URL-brand citation pairs extracted from AI-generated responses across 882 brands. Collected via Trakkr's citation snapshot pipeline from multiple AI providers.
Domain-aware URL classifier (v3) using path pattern matching combined with domain override sets. 58% of URLs classified into 13 page types; 42% remain "other." Citation efficiency scores are brand-normalized so each brand contributes equally regardless of volume.
Limitations
Many corporate pages have generic URL structures that resist URL-based classification. True distribution of classified types may be higher if we could resolve "other."
Data reflects a point-in-time snapshot. AI citation patterns shift as models update - results indicate current trends, not permanent rankings.
