Abstract holographic AI analyzing a network of floating web pages and citation links, visualizing ranking signals such as domain authority badges, brand tokens, layered content depth, and a page-speed gauge.

TL;DR: ChatGPT cites sites that combine real authority, visible brand signals, deep up-to-date content, and decent page speed. Prioritize content and brand visibility first—then layer in link-building and technical fixes for durable AI-search visibility.

ChatGPT is no longer a curiosity—it’s a major search channel. SE Ranking‘s large-scale analysis of 200,000 pages and 100,000 prompts surfaced clear, repeatable patterns about which websites ChatGPT cites. I watched the study closely and here’s how I interpret the findings, what worked, what didn’t, and how I’d run this as a growth campaign for a client.

Table of Contents

What the study tested

The experiment mapped citations across 200K source pages and parsed 100K prompts to find the signals ChatGPT relies on. Four factor groups emerged as primary drivers:

  • Domain authority and trust — backlink diversity, domain trust, and page trust.
  • Brand visibility — overall traffic, homepage traction, presence on forums and review platforms, and Google rankings.
  • Content quality and freshness — depth, structure, length, FAQs, quotes, and recent updates.
  • Technical SEO — load times and Core Web Vitals thresholds.

What went right — clear, actionable signals

The strongest, most reliable findings line up with what I expect from any modern discovery engine: trust and usefulness win.

  • Authority scales citations quickly. Once domains pass noticeable link and trust thresholds, citations jump. Example thresholds from the study: sites with ~32,000 referring domains see a major lift; domain trust ≥90 and page trust ≥28 correlate with far more citations.
  • Brand presence matters. Domains with steady traffic and active profiles on Reddit, Quora, and review sites get cited more. The homepage is a signal—domains with ~7.9K homepage visitors saw nearly double citation rates.
  • Long, structured, fresh content performs. Pages >2,900 words, organized into readable sections (120–180 words per section), with FAQs, stats, and expert quotes, significantly increase citation likelihood. Updating within three months nearly doubles your chance.
  • Speed thresholds, not perfection. Pages with FCP under ~0.4s and Speed Index under ~2.2s avoid citation drop-offs. Oddly, an INP in the 0.8–1.0s range correlated with more citations than an ultra-perfect (<0.4s) INP—suggesting content that’s a bit richer (not overly barebones) fares better.

What went wrong — myths and misfires

The study also disproved several popular “AI SEO” shortcuts that people have started to treat as gospel:

  • LLMs.txt doesn’t help. The research found no positive impact from LLMs.txt; in their analysis, including it didn’t improve citation rates and removing it improved model predictability.
  • FAQ schema isn’t a magic bullet. FAQ content helps when it clarifies and expands coverage, but the structured markup itself didn’t increase citations—pages without FAQ schema actually fared slightly better.
  • Over-optimized URLs and titles hurt. Exact-match and keyword-stuffed titles correlated with fewer citations. ChatGPT prefers clear, topic-aligned language rather than SEO-heavy constructions.

What I would do differently — my strategic plan

Here’s how I’d run a campaign to win citations in ChatGPT, organized by timeline and impact.

Immediate (30–90 days)

  • Audit and expand priority pages: target 1–3 cornerstone articles per product/topic and push them toward 2,500–4,000 words with clear sections (aim for 120–180 words per section).
  • Add high-impact evidence: stats, charts, expert quotes, and short FAQs that answer the likely prompts ChatGPT handles.
  • Refresh content quarterly—update numbers and references now so the page is within the 3-month “freshness” window.
  • Fix obvious speed problems: reduce FCP and LCP if they exceed 1.1s. Prioritize server, image, and render-blocking fixes.

Mid-term (3–9 months)

  • Build branded visibility: contribute helpful answers on Reddit and Quora, claim profiles on G2, Trustpilot, Capterra, and push for genuine reviews.
  • Strengthen internal linking: make Page Trust flow to cornerstone content (use contextual links from high-traffic pages).
  • Start targeted link-building that favors diversity and topical relevance over pure volume—data studies, deep guides, comparisons, and tools are your best link magnets.

Long-term (9–18 months)

  • Invest in domain trust via PR, partnerships, and an increasingly diverse backlink profile. Trust is cumulative—it compounds over time.
  • Run recurring content audits to remove or upgrade low-trust pages (improve citations, expert inputs, and topical depth).
  • Experiment with creator partnerships and original research to create linkable assets that accelerate authority.

“If I were running this, I would’ve layered in creator partnerships upfront.”

Attention doesn’t always mean action—and this test proved that clarity and credibility are what convert attention into citations.

Fast tactical checklist

  1. Pick 3 cornerstone pages per product/topic and expand them to 2.5–4K words with 120–180 word sections.
  2. Include FAQs, original data, and expert quotes; update every 3 months.
  3. Improve internal linking to funnel Page Trust to those cornerstone pages.
  4. Actively build diverse, high-quality backlinks—focus on topical relevance and variety.
  5. Claim and optimize review and marketplace profiles (G2, Trustpilot, Capterra, Yelp).
  6. Fix speed issues that push FCP >0.4s or Speed Index >2.2s; avoid pages slower than 1.1s.
  7. Avoid gimmicks: skip LLMs.txt prioritization, don’t over-stuff titles/URLs, and use schema only when it improves clarity for readers.

FAQs

How quickly will I see ChatGPT citations after making updates?

There’s no instant switch—content and brand signals compound. Expect measurable improvement in citation likelihood over months as you boost content depth, traffic, and backlinks. Freshness updates can produce faster gains if your pages meet authority thresholds.

Is LLMs.txt worth implementing?

Based on the study, no. LLMs.txt showed no positive impact on citation rates—in fact, it didn’t move the needle. Focus on signals the model actually uses: authority, visibility, content quality, and page performance.

How important is Google ranking for ChatGPT visibility?

Very. ChatGPT leans on signals similar to Google’s: overall domain traffic and rankings matter. If your pages rank in the top ~45 positions for relevant queries, citation likelihood increases significantly.

What content length and structure should I aim for?

Aim for long-form coverage—pages above ~2,900 words performed best. Break content into digestible sections of ~120–180 words, include FAQs and evidence, and refresh regularly.

Which technical metrics should I prioritize?

Focus on avoiding slowness: get FCP under ~0.4s, keep Speed Index under ~2.2s, and avoid LCP or FCP exceeding ~1.1s. A reasonable INP (0.8–1.0s) correlates with more citations than an ultra-minimal page.

Final thought

Getting cited by ChatGPT is predictable if you treat it like a modern quality-assessment system. Build real authority, show up where people talk about you, make your content complete and fresh, and don’t let technical sloppiness undercut otherwise great work. Follow the checklist above, prioritize quick content wins, and let trust compound.

Like this content? Consider subscribing to get more. No spam, I promise.