Research shows that 90% of pages cited by ChatGPT rank position 21 or lower in traditional Google search.
That is not a quirk. It is a signal. It tells you that the disciplines of traditional SEO and Generative Engine Optimisation (GEO) are fundamentally different — and that a business with modest Google rankings can compete for AI citations on equal footing with established market leaders.
For local businesses that have been told they can't compete with larger competitors in organic search, this is significant. The citation economy doesn't inherit Google's existing pecking order. It has its own rules — and those rules favour clarity over authority, depth over volume, and structure over seniority.
What Google ranks for vs. what AI cites for
Google's ranking algorithm is built on decades of refinement around a core principle: links from trusted sources signal that a page is worth reading. Domain authority, backlink profiles, dwell time, click-through signals — these are the inputs that determine where a page sits in the results list.
Language models don't follow PageRank. They were trained on text. When they decide what to cite, they are making a judgement about whether a page clearly and reliably answers the question being asked — not about whether that page has accumulated trust signals from other websites over years of operation.
- Domain authority (backlink volume and quality)
- Page-level authority from external links
- Keyword relevance and density
- User behaviour signals (dwell time, CTR)
- Core Web Vitals and technical SEO
- Content freshness and update frequency
- Entity clarity — can the AI identify what you are?
- Answer-first structure — does the page lead with the answer?
- Topical depth — is this comprehensive on a narrow subject?
- Structured data quality (JSON-LD, FAQ, schema)
- Third-party corroboration — are you cited elsewhere?
- Machine-readable content architecture
The practical implication: you can win categories Google told you to give up on
Think about the high-competition local search terms that you've accepted as unwinnable — "marketing agency Buckinghamshire," "SEO consultant Milton Keynes," "accountant High Wycombe." These terms are expensive, competitive, and dominated by businesses that have been building domain authority for years.
Now ask a different question: when someone asks ChatGPT "who is the best marketing consultant in Buckinghamshire for a small business?", does the answer inherit Google's existing rankings? No. It pulls from what the model has learnt about businesses in that space — which is heavily influenced by how clearly and authoritatively those businesses have described themselves and their expertise online.
A smaller business with a well-structured website, clear entity definitions, rich schema, and two or three deeply authoritative articles on its specific area of expertise can appear in that AI response — even if it doesn't crack the top 20 on Google for those keywords.
The three content failures that make pages invisible to AI
Most business websites fail to be cited by AI for the same reasons, regardless of their Google ranking:
Failure 1: No clear definitional statement
Most websites start with a tagline or an aspirational headline rather than a direct, factual statement about what the business is. AI systems need a clear, extractable description: "Marghella Marketing is a growth consultancy based in Buckinghamshire, providing AI-enhanced local marketing systems for established SMEs." That sentence — near the top of the page — is what gets cited. A headline like "Grow Faster. Work Smarter." provides nothing for a language model to extract.
Failure 2: Content written for the scroll, not for the question
SEO content is often designed to keep people on the page — gradual reveals, teased information, engagement-first writing. AI citation rewards the opposite: the answer immediately, then the explanation. A blog post that takes four paragraphs to arrive at its point will lose the citation to a shorter, more direct article that answers the same question in the first sentence. This is uncomfortable for most content writers but essential for GEO.
Failure 3: Wide and shallow instead of narrow and deep
A website with fifteen service pages and five-hundred-word descriptions of each is harder for an AI to trust as authoritative than a website with three services and two thousand words of genuinely useful expertise on each. Topical authority — the quality of being the most thorough, reliable source on a specific subject — is a stronger citation signal than breadth. Narrowing the scope and deepening the content frequently improves Share of Model significantly.
What this means for your SEO investment
None of this argues against traditional SEO. The two disciplines work together — being indexed, having a clean technical foundation, and earning backlinks all contribute to the credibility signals that AI models use alongside their content-based assessments. A business that ranks well on Google and has optimised for GEO will outperform one that has done only one or the other.
The argument is more specific: if you've been told that your local business can't compete for organic visibility because you can't match competitors' domain authority, the citation economy offers a different path. You don't need to win the backlink race to earn AI citations. You need to be the clearest, most structured, most directly useful source on your specific area of expertise in your specific local market.
That is achievable. And for most local businesses in Buckinghamshire and Milton Keynes, the competition for those AI citations is currently near zero.
For a practical guide to measuring where you currently stand, see Share of Model: the metric replacing page rank. If you want a full diagnostic of your site's machine legibility and citation gaps, that's what the AI Search Domination audit covers.