Most business owners ask: how do I get into AI answers? The more useful question is: what is already in AI answers, and does my content look like that? The patterns are visible. The gap between what AI systems cite and what most small business websites publish is consistent and specific.
Semrush, 2025
Semrush, 2025
WPRiders, 2025
The six patterns that appear in AI answers
These are not hypotheses. They are the characteristics of content that consistently appears in AI-generated answers when the subject is a UK service business. Examining what ChatGPT, Perplexity, and Google AI Overviews actually return for service-related queries reveals the same patterns repeatedly.
Content that opens with a direct, specific answer to a question — not a service description, not a brand statement, a literal answer. "A standard boiler service in the UK typically costs between £80 and £120." "No, you do not need planning permission for a single-storey rear extension under four metres." AI systems extract the answer. They do not read your page sequentially and decide to quote paragraph seven. The answer needs to be at the top, clearly stated.
Pages that pair a specific question with a specific answer, and mark that pairing up with FAQPage schema. Semrush analysis of AI Mode citations found FAQ-format pages significantly overrepresented in AI answers relative to their share of the wider web. The structural logic is straightforward: AI systems are built to answer questions. Content already formatted as a question-and-answer match is far easier to extract and attribute. FAQPage schema makes the mapping explicit rather than inferred.
Content containing specific numbers is cited at significantly higher rates than content making general claims. "Three to five working days." "Average UK costs range from £400 to £850 depending on scope." "Most clients see initial results within six to eight weeks." Generic alternatives — "we offer competitive pricing", "quick turnaround", "results vary" — give AI systems nothing to quote. The AI is constructing an answer for someone who asked a real question. Specific data is a citable answer. Vague reassurance is not.
Content attributed to a named person with verifiable credentials performs better in AI citations than anonymous copy. WPRiders research found sites with proper author markup — specifically a Person entity linked via the author property in Article schema — are significantly more likely to be cited. This reflects the E-E-A-T signals Google has been building into its quality guidelines for years. An answer attributed to "Jane Clarke, Chartered Accountant, ICAEW member since 2009" carries more weight with AI systems than the same answer published under no name.
Content whose headings structurally resemble the question being asked. Not keyword-stuffed — structurally similar to how queries are formed. A heading that reads "What should I look for when choosing a solicitor?" followed by a direct answer is citable. A block of text that happens to mention similar topics across several paragraphs is not. AI systems match questions to content. Pages that think in questions — and structure content around specific questions — are naturally easier to match against the queries AI receives.
Content on pages where schema markup has explicitly declared what the business is, where it operates, and what the page is about. An answer on a page with FAQPage, LocalBusiness, and Article schema is verifiable content from a known entity in a declared location covering a stated topic. The same content on an unmarked page is floating prose. AI systems making real-time decisions about what to cite prefer the verifiable option. The content does not change. What changes is the confidence the AI can have in attributing it correctly.
What AI systems consistently do not cite
The exclusions are as informative as the inclusions. These are the content types that appear rarely or never in AI-generated answers for service business queries, regardless of how well they perform in traditional Google search.
Generic service descriptions
"We offer a comprehensive range of [service] solutions tailored to your needs." This type of copy describes nothing specific, answers no question, and gives AI systems nothing to quote. It is written for impressions, not for information extraction.
Marketing-language headlines
"The region's most trusted [service] provider." AI systems are constructing factual answers. Superlative claims with no supporting evidence are not citable. They do not help a user understand what the business does, where it operates, or what it costs.
Answers buried in prose
Content that reaches the point after three paragraphs of context-setting. AI systems pull answers efficiently. If the response to "how much does X cost?" appears in paragraph four after two paragraphs of explanation, it is significantly less likely to be extracted than if it appears in the first sentence.
Pages blocked to AI crawlers
Robots.txt rules that block GPTBot, ClaudeBot, or PerplexityBot mean the content cannot be indexed by those systems regardless of quality. This is more common than expected — many sites inherited crawler-blocking rules from previous SEO configurations without knowing it.
Unsigned, unattributed copy
Content published without author attribution carries no E-E-A-T weight. AI systems are increasingly distinguishing between content from identifiable, credentialed sources and content from anonymous pages. The same answer attributed to a named expert outperforms unsigned copy.
Homepage-only content
Businesses whose website is essentially a one-page brochure — service headline, brief description, contact form — have almost nothing for AI systems to cite. AI answers are constructed from specific content. A homepage that describes rather than explains is rarely the source.
Cited vs not cited: the content gap in practice
"AI systems are not reading your website looking for keywords. They are looking for specific, attributable answers to specific questions. Every page on your site either answers a question clearly or it does not. The ones that do not are invisible to AI search, regardless of how well they rank on Google."
Where does your content stand?
Use this checklist to assess your current content against the citation patterns above. These are observable characteristics, not strategic goals — either a page has them or it does not.
The pattern in one sentence
AI systems cite content that answers a specific question, contains verifiable data, comes from an identified source, and sits on a page where schema markup has confirmed the business context. Content that does none of these things does not appear in AI answers — regardless of Google ranking, website age, or how many visitors it receives.