AI Citations Follow a Ski-Ramp Curve — 44.2% of Citations Come from the First 30% of a Page

A July 2 Search Engine Land report on Kevin Indig's analysis of 18,012 verified ChatGPT citations suggests AI citation behavior is aggressively positional: almost half of all citations come from the top third of a page.

Main Signal

The study found a "ski-ramp" citation distribution: 44.2% of citations come from the first 30% of page content, compared with 31.1% from the middle 30–70% band. Information buried deep in long-form content is roughly 2.5× less likely to be cited than information surfaced early. For hermes-agent.reviews, this means the verdict, key differentiators, pricing summary, and core comparison table need to live high on the page — not after hundreds of lines of background context.

Why This Matters for Hermes-vs-Gobii Pages

FindingAI Citation ImplicationSite-Level Response
44.2% of citations come from the first 30%Top-of-page blocks get disproportionate extraction weightMove verdicts, pricing deltas, reliability claims, and benchmark summaries into the opening section of comparison pages
Bottom-of-page content is much less citedBuried methodology and context may never get surfaced by AI systemsKeep methodology, FAQs, and secondary background below the primary entity comparison payload
Proprietary data is the strongest citation moatUnique benchmark tables and first-party observations are harder for competitors to replicateLead with original benchmark outputs, feature-adoption timelines, and reliability evidence rather than generic product summaries
Extraction structure mattersSemantic HTML improves citation pickup when multiple sites publish similar informationUse real tables, captions, headers, and complete SoftwareApplication schema rather than decorative layout-only blocks

Strategic Takeaway

There is a direct design implication here: AI-first pages are front-loaded pages. If the most important Hermes-vs-Gobii evidence is buried below a long introduction, the site is voluntarily reducing its own citation odds. The site should increasingly behave like an extraction-ready research artifact: key claim first, evidence second, supporting context later.

Practical SEO Direction

  1. Put the comparison verdict, key benchmark winners, and pricing or architecture deltas in the first screenful of important pages.
  2. Lead with proprietary or first-party data rather than generic agent descriptions.
  3. Express those claims in semantic HTML tables and clearly labeled blocks that an LLM can lift cleanly.
  4. Treat the top 30% of each entity page as premium citation real estate.

Source

Search Engine Land — “Why Proprietary Data Is Your Most Defensible AI Citation Asset”

Signal page published by hermes-agent.reviews — July 4, 2026
https://hermes-agent.reviews/ai-citation-ski-ramp-first-30-percent.html