{"slug": "we-checked-whether-on-site-seo-predicts-ai-citations-the-data-says-mostly-no", "title": "We Checked Whether On-Site SEO Predicts AI Citations. The Data Says Mostly No.", "summary": "Causabi, a generative engine optimization tool, found that on-site SEO scores have almost no correlation with AI citation rates. In a study of 44 domains, the correlation between on-site readiness and citations was near zero, with 86% of domains receiving zero citations regardless of their score. Brand prominence was the primary driver of citations, not technical optimization.", "body_md": "Every GEO (\"generative engine optimization\") tool, including ours until\n\nrecently, sells some version of the same pitch: fix your robots.txt, add\n\nSchema.org markup, write FAQ schema, and AI engines will cite you more.\n\nWe build one of these tools — Causabi scans sites for AI-crawler readiness\n\nand generates fix files (robots.txt, llms.txt, JSON-LD, FAQ blocks). As part\n\nof validating our own scoring weights, we ran the numbers on whether the\n\nscore actually predicts getting cited. Short version: it mostly doesn't,\n\nonce brand prominence is in the picture.\n\n##\nWhat we measured\n\nWe scored 44 domains on a 6-category on-site readiness algorithm:\n\n- robots.txt (AI bots allowed or blocked)\n- Schema.org (Organization/LocalBusiness JSON-LD completeness)\n- FAQ schema (FAQPage markup, 3+ entries)\n- content depth/structure\n- brand/NAP signals\n- freshness (dateModified, recency)\n\nThen we checked how often each domain actually got cited by an AI engine\n\n(Claude, via its web-search tool, one measurement window, a fixed prompt set\n\nper domain).\n\n##\nWhat we found\n\n- On-site score vs. citation rate: Pearson r ≈ -0.08, Spearman ρ ≈ -0.03.\nFunctionally no correlation — if anything, a very slight negative one,\nwhich is more likely noise than a real inverse relationship at this\nsample size.\n- 86% of the 44 domains got zero citations in the window, independent of\ntheir score.\n- The domains that\n*did* get cited clustered almost entirely by brand\nprominence — well-known domains got cited at a noticeably higher rate\n(~0.16 of prompts) than everyone else (~0 for the rest of the sample),\nregardless of how well-optimized their markup was.\n\n##\nWhy I'm not overselling this\n\nn=44 is small. This is an internal validation exercise for our own product,\n\nnot a peer-reviewed study, and I don't want it read as one. Specific caveats:\n\n- Single engine (Claude) this round. Citation behavior differs meaningfully\nacross ChatGPT, Gemini, Grok, and Perplexity — we haven't run the same\ncheck across all four yet.\n- One time window, no longitudinal before/after. We didn't take a domain,\nimprove its score, and watch citations change over months. That's the\nactually convincing experiment and we haven't run it yet.\n- Prompt-domain matching wasn't blind. Some prompts were picked because a\ndomain plausibly related to that topic, which likely biases toward\ndomains that would get mentioned anyway.\n- \"Brand prominence\" is a fuzzy variable that probably absorbs some real\ncontent-quality signal we're not capturing separately. We can't fully rule\nout that what looks like \"brand wins\" is partly \"genuinely better/more\nauthoritative content wins,\" which on-site markup scoring doesn't measure.\n\n##\nWhat we still think is true, with more confidence\n\nSome things aren't correlational guesses — they're closer to mechanical\n\nfacts:\n\n-\n**robots.txt blocking is binary.** If `GPTBot`\n\n, `ClaudeBot`\n\n, or similar are\ndisallowed, that engine cites you zero times, by construction. About 89%\nof sites we've scanned block at least one AI crawler by default, usually\nby accident (a blanket `Disallow: /`\n\nthat predates AI bots existing).\n-\n**FAQ schema changes extraction, not inclusion.** For content that's\nalready in an engine's consideration set, structuring it as self-contained\nQ&A chunks seems to affect whether it gets pulled into a RAG-style\ncitation — this lines up with published research on chunking behavior. But\nthat's a \"how you're cited\" lever, not a \"whether you're cited\" lever.\n\n##\nWhere that leaves the product\n\nWe're rewriting our own copy to say what the score actually measures:\n\nAI-crawler readiness and machine-readability, not citation probability. No\n\ntool — ours included — can promise the second one. If your on-site work is\n\nmostly aimed at \"getting cited more,\" the more binding constraint for most\n\nsites is probably brand/mentions elsewhere, not another Schema.org type.\n\nThe scoring engine and fix generator are open source (MIT) if you want to\n\nsee the logic or run it on your own site without touching our SaaS:\n\nRepo: [https://github.com/SHADRINMMM/causabi-geo](https://github.com/SHADRINMMM/causabi-geo)\n\nSite (hosted version + monitoring): [https://causabi.com](https://causabi.com)", "url": "https://wpnews.pro/news/we-checked-whether-on-site-seo-predicts-ai-citations-the-data-says-mostly-no", "canonical_source": "https://dev.to/mikhail_shadrin_dev/we-checked-whether-on-site-seo-predicts-ai-citations-the-data-says-mostly-no-1j8f", "published_at": "2026-07-15 21:07:09+00:00", "updated_at": "2026-07-15 21:41:07.679015+00:00", "lang": "en", "topics": ["ai-tools", "ai-research", "developer-tools"], "entities": ["Causabi", "Claude", "GPTBot", "ClaudeBot"], "alternates": {"html": "https://wpnews.pro/news/we-checked-whether-on-site-seo-predicts-ai-citations-the-data-says-mostly-no", "markdown": "https://wpnews.pro/news/we-checked-whether-on-site-seo-predicts-ai-citations-the-data-says-mostly-no.md", "text": "https://wpnews.pro/news/we-checked-whether-on-site-seo-predicts-ai-citations-the-data-says-mostly-no.txt", "jsonld": "https://wpnews.pro/news/we-checked-whether-on-site-seo-predicts-ai-citations-the-data-says-mostly-no.jsonld"}}