How AI Search Is Changing Who Finds Your Business — And What You Can Do About It

Something fundamental changed in how people find businesses in 2024 and 2025. Not gradually. Not theoretically. Right now, while you are reading this, a meaningful percentage of potential customers are getting answers from an AI — not a list of blue links.

ChatGPT now handles approximately 17% of global search queries. Google AI Overviews appear in at least 16% of all searches — and 40%+ in some categories. Perplexity AI has crossed 100 million monthly active users. These are not small numbers. And for most UK small businesses, they represent an invisible audience.

This article explains exactly what has changed, how each AI search platform works differently, what "being cited" means versus "ranking", and — most importantly — what the five components of AI visibility are and how our audit measures each one.

Key Takeaway: AI search engines now handle a substantial share of search queries. They work on a citation economy of 2–7 sources per answer — far fewer than Google's 10 blue links. Being cited requires different optimisation than ranking. 78% of companies are already investing in GEO, but only 23% are measuring it. That gap is where the opportunity sits.

---

The Shift from 10 Blue Links to AI-Generated Answers

For twenty-five years, search worked the same way: type a query, get a list of links, click one. The value your website generated from search was proportional to your position in that list. Rank first, get the click. Rank tenth, get almost nothing.

AI search breaks this model in two ways.

First, instead of listing links, AI search engines generate an answer. The AI reads dozens of sources, synthesises the information, and produces a direct, conversational response. The user gets what they need without clicking through to your site — or they click the citation sources that the AI has chosen.

Second, instead of returning ten results, AI engines typically cite two to seven sources per answer. The number is not fixed, but it is dramatically smaller than Google's results page. This is not a ranking problem — it is a citation problem. And citation works by different rules.

The implications are significant. For queries where AI Overviews appear, organic click-through rates on Google have dropped by approximately 30%. For some informational queries, click-through rates have fallen even further. The traffic is not disappearing — it is going to the cited sources or staying with the AI itself.

---

ChatGPT, Perplexity, and Google AI Overviews: How Each Works Differently

Understanding why GEO requires platform-specific thinking starts with understanding how these systems actually retrieve and cite information.

ChatGPT (OpenAI)

ChatGPT operates in two modes that affect how your business appears.

Training data mode: When ChatGPT answers questions without web search, it draws entirely on its training data — which has a knowledge cutoff date. Newer businesses or businesses that have changed significantly since the cutoff may be described inaccurately or not at all. For this mode, the only way to influence your presence is to ensure your business is described accurately and consistently across the web — because that consistency is what gets absorbed into training data over time.

Web search mode: When ChatGPT searches the web (which it does for most queries involving current information, local businesses, or recent events), it uses OpenAI's GPTBot crawler. If your robots.txt blocks GPTBot — even accidentally — your website cannot be cited in web search results regardless of content quality. This is the single most common GEO problem we find.

ChatGPT's web search mode prioritises content that is:

• Directly accessible to GPTBot (not behind a login, paywall, or bot-blocking rule)
• Structured with clear, direct answers at the top of the page
• Backed by credible signals (E-E-A-T, external citations, brand mentions)

Perplexity AI

Perplexity is a dedicated AI search engine — its entire product is AI-powered search rather than a chat assistant with search added on. This distinction matters: users on Perplexity are almost always in search mode, looking for information, products, or services. The intent is high.

Perplexity crawls the web with PerplexityBot. Like ChatGPT, if PerplexityBot is blocked, your content cannot be cited. Unlike ChatGPT, Perplexity is more aggressive about surfacing real-time, current sources — fresher content has a relative advantage.

Perplexity also displays its sources prominently in the answer interface, making citation visibility more commercially valuable. Users can see exactly where the information came from and click through. Being cited in Perplexity answers therefore combines the brand visibility of an AI mention with the traffic potential of a traditional link.

Google AI Overviews

Google AI Overviews are unique because they sit inside the Google search results page — the same environment where traditional SEO matters. But the optimisation required to appear in an AI Overview is not the same as the optimisation required to rank organically.

AI Overviews prioritise pages with:

• Direct answers to the query question (answer-first structure)
• FAQPage or HowTo schema markup signalling the content is structured Q&A
• Strong domain authority and topical credibility
• Proper speakable schema on key passages

Critically, appearing in an AI Overview does not always correlate with ranking highly organically. You can rank third for a keyword and be cited in the AI Overview. You can rank first and not be cited. Traditional SEO rank and AI Overview inclusion are related but distinct outcomes.

---

What "Being Cited" Means vs "Ranking"

These two outcomes require different strategies and should be measured separately.

Ranking means appearing in a list of results ordered by relevance. It is a position — first, third, tenth. The goal of traditional SEO is to move up this list. Success is measured by position and by the click-through rate your position generates.

Being cited means an AI system has selected your content as a source for a generated answer. There is no "position" in the traditional sense — you are either cited or you are not. The equivalent of ranking first is having your content quoted directly in the AI's answer. The equivalent of ranking tenth is being mentioned as a source link that the user might click.

The citation economy is much more compressed than the ranking economy. Google's first page shows ten results. An AI answer typically cites two to seven sources. This means the difference between being cited and not being cited is enormous — there is no equivalent of the "second page" in AI search. You are either in the answer or you are invisible.

This changes how you should think about content strategy. For traditional SEO, producing the tenth-best article on a topic still generates some traffic. For GEO, producing the tenth-most-citable article on a topic generates nothing. The bar is higher, but so are the rewards — an AI citation combines brand visibility (your business is named in the answer), traffic potential (the source link), and authority building (repeated citation builds recognition over time).

---

The Statistics: Where Things Stand

78% of companies across all sectors are already funding GEO programmes. Only 23% are measuring the results. That gap between action and measurement is significant — it means most businesses are spending time and money on AI search visibility without knowing whether it is working.

The measurement gap exists because GEO is genuinely harder to measure than traditional SEO. There is no equivalent of Google Search Console for AI search. You cannot see your "AI search position" in a dashboard. Measurement requires a combination of: manually testing AI engines for your key queries, monitoring brand mention tools for AI-generated contexts, and tracking referral traffic from AI platforms (ChatGPT, Perplexity, and others show up in GA4 referral data when they drive click-throughs).

At seoandgeo.co.uk, our GEO Score within the Digital Visibility Score attempts to proxy this. It is not a perfect measurement of AI citation frequency — that would require running queries against live AI systems at scale. But it measures the five underlying factors that drive AI citation, which are within your control.

On the traffic quality side, the data is striking. Visitors arriving from AI search platforms convert at 4.4x the rate of visitors from organic search. The reason is intent — people who ask ChatGPT "What is the best accountant for a small business in Manchester?" are much further along the buying journey than people who type "accountant Manchester" into Google. The queries AI search handles tend to be more specific, more commercial, and more ready to act.

---

The Five Components of AI Visibility

When we assess a website's GEO Score, we look at five distinct components. Every one of these can be improved — none of them requires a complete website rebuild.

Component 1: AI Crawler Access

Is your website accessible to the AI crawlers that index content for search responses? The main crawlers are GPTBot (OpenAI/ChatGPT), ClaudeBot (Anthropic), PerplexityBot (Perplexity), and Googlebot-Extended (Google AI features). Each must be explicitly allowed in robots.txt — and many websites accidentally block them through overly restrictive wildcard rules like Disallow: / for all bots.

How to check: Visit yourdomain.com/robots.txt. Look for any Disallow rules that apply to GPTBot, ClaudeBot, or PerplexityBot, or any wildcard Disallow: / rules that would affect all bots. If you see any, these crawlers cannot index your content.

How to fix: Add explicit Allow rules for each AI crawler in your robots.txt file, or modify the wildcard rule to exclude these crawlers from the blocked list.

Component 2: Content Structure for Citation

AI systems extract citations at the passage level — they identify specific sentences or paragraphs to quote, not whole pages. This means the internal structure of your content matters enormously. Content that buries the answer after paragraphs of context is harder for AI to cite accurately. Content that leads with the answer, uses clear headings, and structures information as direct responses to questions is far easier.

The answer-first principle: State your main point in the first sentence of each section. Do not save the punchline for the end of a long explanation. AI engines are looking for extractable facts, not narrative build-ups.

Question-based headings: Structure your headings as questions ("How does X work?" "What does Y mean?") so that AI systems can match them to user queries more directly.

Citation capsules: Short, self-contained paragraphs that make a specific factual claim with enough context to stand alone. A good citation capsule can be extracted verbatim and still make complete sense.

Component 3: Brand Entity Signals

AI systems do not just read your website — they cross-reference your brand across the web. Consistent, accurate information about your business across multiple authoritative sources builds the "brand entity" that AI systems use when deciding how to describe you.

This includes: LinkedIn company page, Google Business Profile, Companies House listing (for UK businesses), industry directories, press mentions, and consistent NAP (Name, Address, Phone) data across citations. A business with strong, consistent brand entity signals is easier for AI to describe confidently. A business with inconsistent or sparse signals gets described cautiously or not cited at all.

Component 4: Schema Markup

Structured data in JSON-LD format helps AI systems understand the type and context of your content. The most important schemas for GEO are:

• Organisation with sameAs properties linking to your social profiles and key directory listings — this directly builds brand entity recognition
• FAQPage — makes individual Q&A pairs directly extractable by AI systems
• LocalBusiness — essential for geographic citation in local queries
• Article/BlogPosting — signals that content is authoritative reference material
• Speakable — specifically designed for voice and AI reading, marks the most citable passages

Component 5: llms.txt

The llms.txt file is new and underused. Placed at yourdomain.com/llms.txt, it is a plain-text or markdown file that describes your business in AI-readable format: what you do, your key services, your authoritative content, and how you want to be described. Think of it as a README file for AI systems.

The format is simple but powerful. A well-structured llms.txt file includes: a brief description of your business, links to your most important pages, a description of your target customers, your key products or services with brief descriptions, and any specific instructions about how to describe or cite you accurately.

As of early 2026, fewer than 3% of websites have an llms.txt file. This is both a gap and an opportunity — being an early adopter in your industry means your llms.txt content gets absorbed into AI knowledge before your competitors have one.

---

How seoandgeo.co.uk Measures and Improves This

Our GEO audit covers all five components above plus platform-specific AI search optimisation. Here is what we check and what we recommend when we find gaps.

Crawler access check: We test each AI crawler against your robots.txt in real-time during the audit. If any are blocked, this is flagged as a critical finding with the exact robots.txt rule causing the block and the precise fix needed.

Content structure analysis: Our AI Citability agent reads your key pages and scores them on: answer-first structure, heading quality, citation capsule presence, and question-answer completeness. Pages that score below threshold get specific rewrites recommended — not generic advice like "improve your content", but specific structural changes with examples.

Brand entity assessment: We check the presence and consistency of your brand entity signals across Google Business Profile, LinkedIn, key UK directories (Yell, Thomson Local, FreeIndex), Companies House, and your own site's social profile links. The output is a list of gaps and inconsistencies.

Schema audit: We check every schema type present on your site, validate it against Schema.org specifications, and identify missing schema types that would improve AI citability. For Comprehensive tier audits, we generate the missing JSON-LD and provide it ready to paste into your site.

llms.txt generation: Comprehensive tier audits include a custom llms.txt file built from your audit data — designed to give AI systems an accurate, favourable description of your business and guide them to your most important content.

---

Key Takeaways

• AI search now accounts for a significant and growing share of all search queries. The shift from ten blue links to AI-generated answers is already happening at scale.
• ChatGPT, Perplexity, and Google AI Overviews work differently and require different optimisation approaches. One-size-fits-all GEO does not work.
• The citation economy is brutal: 2–7 sources cited per answer versus 10 per Google page. Being cited requires meeting a higher bar than ranking tenth.
• 78% of companies fund GEO. Only 23% measure it. The measurement gap is where the opportunity sits.
• Visitors from AI search convert at 4.4x the rate of organic search visitors. The audience is more valuable — but you have to be visible to reach it.
• The five components of AI visibility (crawler access, content structure, brand entity, schema, llms.txt) are all improvable without rebuilding your website.

---

Frequently Asked Questions

Does AI search really affect small businesses or just big brands?

AI search affects small businesses significantly, in some ways more than large brands. When someone asks ChatGPT "Who is the best accountant in Leeds?" or "Which plumber in Bristol is reliable?", the AI draws on local and niche sources — not just the biggest brands. The citation economy of 2–7 sources per answer means there is genuine opportunity for well-optimised small businesses to appear in answers that large brands do not dominate.

How does ChatGPT decide what to cite?

ChatGPT uses a combination of its training data and real-time web browsing. For web search queries, it prioritises sites that AI crawlers can access without being blocked, content structured with clear headings and direct answers, pages with strong E-E-A-T signals, and sources that appear consistently across multiple authoritative references. There is no single ranking factor — it is a weighted combination.

Is GEO the same as AI SEO?

GEO (Generative Engine Optimisation) is the most commonly used term for the practice of optimising for AI search. You will also hear "AI SEO", "AI search optimisation", and "LLM optimisation". They all refer to the same goal: making your content visible to and citable by AI-powered search engines. The techniques overlap with traditional SEO in some areas but diverge significantly in others.

What is the single most important GEO action for a small business?

Check whether AI crawlers can access your site. Specifically, check your robots.txt file to see if GPTBot, ClaudeBot, or PerplexityBot are blocked. This is the most common GEO problem we find — businesses completely invisible to AI search not because of bad content but because a robots.txt rule is blocking the crawlers. Check this by visiting yourdomain.com/robots.txt in your browser.

How long does it take for GEO changes to show results?

Unlike traditional SEO where you wait for Google to recrawl and rerank, GEO improvements can show results much faster — sometimes within days of AI crawlers indexing your updated content. Brand entity signals build more slowly, over weeks to months. The most impactful quick wins (unblocking AI crawlers, adding llms.txt, restructuring content for answer-first format) typically show measurable results within 2–4 weeks.

---

Want to check your AI visibility score? Get your audit and see exactly where you stand across all five GEO components.