Forem: Coven AI

The Rise of GEO: What It Is, Why It Matters, and How It Differs from SEO

Coven AI — Thu, 30 Apr 2026 06:27:47 +0000

If your business content doesn't show up when ChatGPT, Claude, or Perplexity answer questions about your industry, you're already losing visibility. Welcome to GEO — Generative Engine Optimisation — and the uncomfortable truth is that most businesses have no idea it exists.

CoveN AI's analysis of 1,003 businesses across 58 countries found that the average GEO readiness score sits at just 36 out of 100. Only 13% of businesses score above 50. That means 87% of companies are poorly positioned for the way people actually find information now.

This isn't a future problem. It's happening right now, and the gap between businesses that understand GEO and those that don't is widening fast.

What Is GEO?

Generative Engine Optimisation (GEO) is the practice of optimising content so it gets cited, referenced, and included in AI-generated responses — not just ranked in traditional search results.

The term was formally defined in the 2023 paper "GEO: Generative Engine Optimization" by researchers from Princeton University, Georgia Tech, the Allen Institute for AI, and IIT Delhi. The paper introduced GEO as a distinct discipline focused on increasing visibility in outputs from large language models like GPT-4, Claude, and Gemini.

Where SEO aims to get your page onto the first page of Google, GEO aims to get your information into the answer itself.

Think about how behaviour has shifted. When someone asks "What's the best CRM for small teams?", they're increasingly likely to ask an AI assistant rather than Google. The AI doesn't return ten blue links — it returns a synthesised answer, often citing two or three sources inline. If your content isn't citation-worthy to the AI, you don't exist in that interaction.

How GEO Differs from SEO

SEO and GEO share some DNA — both care about content quality, structure, and authority — but they're optimising for fundamentally different outcomes.

SEO Optimises for Ranking

Traditional SEO is about earning a position in a ranked list. You compete for keywords, build backlinks, improve page speed, and hope to land in positions 1-10. Success is measured in rankings, click-through rates, and traffic.

The user sees your link, decides whether to click, and then engages with your site. You control the experience once they arrive.

GEO Optimises for Inclusion

GEO is about getting your content selected, synthesised, and cited within an AI-generated answer. You're not competing for position 1 versus position 3 — you're competing to be referenced at all.

The user never clicks through to your site in many cases. The AI extracts your information, reformulates it, and presents it as part of a broader answer. If you're cited, you get attribution (sometimes). If you're not cited, you get nothing.

The Princeton/Georgia Tech GEO paper tested multiple optimisation strategies and found that adding citations, increasing factual density, and using authoritative language increased visibility in generative outputs by up to 40% in some cases.

The Signals Are Different

Google's algorithm weighs backlinks heavily. Generative engines weigh citation-worthiness: is this content factual, well-structured, clearly sourced, and easy to extract?

Here's what matters more in GEO:

Factual density — more verifiable claims per paragraph
Authoritative sourcing — inline citations, references to studies, named sources
Structured content — tables, lists, clearly labelled sections that are easy to parse
Clear attribution — who said this, when, and where
Quotability — sentences that can be extracted cleanly and still make sense

Page speed and mobile-friendliness? Still useful, but not the primary lever. Backlinks? Helpful for authority signals, but not a direct ranking factor in the same way.

Why GEO Matters Now

The usage data is unambiguous. Cloudflare's 2025 Radar report found that GPTBot traffic (OpenAI's web crawler) grew 305% year-on-year. Google's crawl rate, by comparison, grew just 12%.

AI assistants are becoming the default interface for information retrieval, especially for high-intent queries: "What should I do about X?", "Which option is better?", "How do I solve Y?"

These are exactly the queries businesses want to be part of. And right now, most aren't.

CoveN AI's research shows a striking global pattern. The average GEO readiness score across all regions hovers between 33 and 39 out of 100. North America leads slightly at 39, but even that is barely a passing grade. Europe sits at 35. Latin America at 33. The Middle East at 34.

This isn't a problem isolated to developing markets or small businesses. It's a systemic gap.

Industry Breakdown: Who's Ready and Who Isn't

Some industries are adapting faster than others, but even the best performers are barely past the halfway mark.

Gyms score highest for GEO readiness at 42 out of 100 — possibly because fitness content is naturally list-heavy, FAQ-driven, and citation-friendly. Plumbers also score 42, likely benefiting from how-to content and structured service descriptions.

At the other end, offices score just 29 out of 100 for GEO readiness. Dentists sit at 32. Restaurants at 34. Solicitors at 34.

Even high-performing sectors like healthcare (39) and accountancy (38) are only marginally better. Web agencies — the businesses you'd expect to be ahead of the curve — score just 37.

The takeaway? No industry has solved this yet. Early movers have a genuine opportunity to establish authority while competitors are still figuring out what GEO even is.

The Key Signals: What Makes Content GEO-Friendly?

The Princeton/Georgia Tech research tested nine optimisation methods across multiple generative engines (GPT-4, Claude, and others). The strategies that worked best shared common characteristics.

1. Authoritative Sourcing

Content that cites credible sources gets cited more often itself. This creates a reinforcing loop: authoritative content is more likely to be referenced, which increases its authority further.

Inline citations, references to studies, and named experts all signal to a generative model that this content is reliable. Vague claims like "studies show" or "experts believe" do the opposite — they signal uncertainty.

2. Factual Density

Generative engines favour content that packs verifiable information into concise sections. Fluffy intros, repetitive phrasing, and vague generalisations dilute factual density.

A paragraph with three specific claims ("X costs £200", "Y takes 48 hours", "Z requires certification") is more citation-worthy than three paragraphs of marketing copy.

3. Structured Formatting

Tables, bullet lists, clearly labelled headings, and step-by-step instructions are easier for models to extract and reformulate. Unstructured prose is harder to parse and less likely to be cited accurately.

If your content answers a question but buries the answer in paragraph seven, it's not GEO-optimised.

4. Citation-Friendly Language

Sentences that can be extracted cleanly and still make sense out of context perform better. This is the "quotability" test.

Compare:

"As we mentioned earlier, this approach can sometimes help in certain situations depending on various factors."

Versus:

"Structured data markup increases AI citation rates by making content machine-readable."

The second sentence is self-contained, specific, and factual. It can be cited directly without additional context.

5. Recency and Maintenance

Generative models favour recently updated content, especially for time-sensitive queries. A 2019 blog post about "best practices" is far less likely to be cited than a 2026 post, even if the advice is similar.

Regular content updates signal that information is current and maintained.

SEO and GEO Are Complementary, Not Competing

Here's the critical point: GEO doesn't replace SEO. The two work in tandem.

Traditional search isn't disappearing. Google still processes billions of queries daily, and ranking well in search results still drives traffic. But the information landscape now has two layers:

Layer 1 (SEO): Get found in traditional search engines and earn clicks to your site
Layer 2 (GEO): Get cited in AI-generated answers and influence how your brand and expertise are represented

Businesses that optimise for both layers win twice. Those that ignore GEO are ceding control over how their expertise is synthesised and presented.

Some of the tactics overlap. High-quality content benefits both SEO and GEO. Structured data markup (schema.org) helps search engines and AI models parse your content. Authoritative backlinks boost domain trust across the board.

But some tactics diverge. Keyword density matters less in GEO. Backlink velocity matters less. What matters more is whether your content is extraction-friendly — can a model confidently pull facts from it and cite you accurately?

What This Means for Your Business

If you're running a business website in 2026 and you haven't thought about GEO, you're likely already behind.

The data is clear: 39% of businesses score below 30 out of 100 on GEO readiness. That's not "room for improvement" — that's functionally invisible to AI systems.

The businesses that adapt early will establish themselves as default sources in their industries. The ones that wait will find themselves excluded from an increasingly large share of customer research and decision-making.

This doesn't require a complete content overhaul. Start with your highest-value pages:

Service descriptions
Product pages
FAQ sections
Industry expertise or thought leadership content

Ask yourself:

Is this content factually dense, or is it mostly fluff?
Can key claims be extracted cleanly and cited out of context?
Are there inline citations or references to credible sources?
Is the content structured (lists, tables, headings) or a wall of text?
Has it been updated recently?

Small changes compound quickly. Adding a table to a pricing page, restructuring an FAQ, citing a source inline — these are low-effort, high-impact adjustments.

Measuring GEO Performance

Unlike SEO, where you can check rankings and traffic in Google Analytics, GEO is harder to measure directly. Generative engines don't provide dashboards showing how often you've been cited.

But proxy metrics exist:

Increase in GPTBot or Claude-Bot crawl activity (check your server logs)
Branded search volume — if more people search for your business after encountering you via AI, that's a signal
Direct traffic spikes correlated with AI-generated content that cited you
Qualitative feedback — customers mentioning they found you via ChatGPT or Perplexity

CoveN AI's Agent Analytics tool provides a GEO readiness score by analysing your site's structure, factual density, schema markup, and citation-worthiness. It's one of the few tools purpose-built to assess AI visibility rather than just traditional SEO.

The Bottom Line

GEO is not a buzzword. It's a formal discipline with peer-reviewed research behind it, and it's already reshaping how businesses get discovered.

The academic foundation laid by the Princeton/Georgia Tech team has given the industry a framework. Now it's up to businesses to act on it.

The gap between the 13% of businesses scoring above 50 on GEO readiness and the 87% lagging behind will define competitive advantage over the next few years. Early movers will own mindshare in AI-driven discovery. Late movers will be citing the early movers.

Want to know where you stand? Run a free scan at aa.covenai.io and get your GEO readiness score in under 60 seconds.

Ready to see where your site stands? Run a free ACO scan at aa.covenai.io — no sign-up required.

The Real Cost of Being Invisible to AI Agents (Spoiler: It's Already Affecting Revenue)

Coven AI — Thu, 23 Apr 2026 13:17:02 +0000

Your competitors are being recommended by Claude, Perplexity, and ChatGPT right now. Your business isn't.

It's not because their products are better. It's because AI agent visibility has become the new frontier of digital discovery, and most businesses don't even know they're already losing.

When a potential customer asks an AI agent to "find the best accountants in Manchester" or "compare project management tools for remote teams," that agent is making recommendations based on what it can see, understand, and verify about your business. If your website isn't structured for AI comprehension, you simply don't exist in that conversation.

The cost isn't theoretical. It's measurable, it's happening now, and according to Coven AI's analysis of 1,003 businesses across 58 countries, 39% of companies score below 30/100 on AI readiness—effectively invisible to the agents that are increasingly mediating purchase decisions.

How AI Agents Actually Make Recommendations

When someone asks Claude to help them find a service provider or Perplexy to research software options, the AI agent doesn't conduct a traditional Google search. It synthesises information from multiple sources, evaluates credibility signals, and makes determinations about which businesses best match the query intent.

This process relies on structured data, clear semantic markup, and machine-readable information about what you offer, who you serve, and why you're credible. Traditional SEO focused on keywords and backlinks. Agentic commerce requires something more fundamental: your website must be comprehensible to non-human readers.

Here's what that looks like in practice. A CFO asks ChatGPT: "I need a solicitor experienced in Series A fundraising in London." The agent evaluates dozens of law firms, but only recommends three. The others? Their websites mentioned fundraising in blog posts, but didn't structure that information as a service offering with verifiable credentials.

The firms that got recommended had proper schema markup identifying their practice areas, attorney credentials in machine-readable formats, and clear signals about their geographic service area. They weren't better lawyers. They were just visible to the agent making the introduction.

The Revenue Impact You're Already Feeling

If you've noticed qualified leads declining despite steady website traffic, agent traffic might explain the gap. Potential customers are still researching. They're just not finding you during that research.

Consider three specific revenue impacts happening right now:

1. Missed Recommendations at Point of Intent

A prospect with budget and timeline asks an AI agent for vendor recommendations. Your competitor gets cited with a description of their capabilities and a direct link. You don't appear at all. The prospect never visits your website because they never learn you exist.

This isn't a failure of your marketing content. It's a failure of AI discovery infrastructure. The agent couldn't parse your service offerings, couldn't verify your credentials, couldn't match your capabilities to the query. Your competitor's website structure made all three trivial.

2. Lost Comparison Traffic

Even when prospects know about your brand, AI agents are increasingly mediating the comparison phase. Someone asks: "Compare [Competitor A] and [Competitor B] for enterprise project management." If your product isn't included in that comparison, you've lost a sales opportunity you might have previously captured through direct search.

Coven AI's research found that only 13% of businesses score above 50/100 on AI readiness. That means in most competitive sets, the businesses getting included in AI-mediated comparisons aren't necessarily the market leaders. They're the ones whose websites structure product information in ways agents can reliably extract and compare.

3. Competitor Citations in Your Category

Perhaps most frustrating: prospects who should be perfect fits for your offering are being steered elsewhere. Not because the AI agent evaluated both options and preferred your competitor, but because it only found your competitor in the first place.

When an agent can't confidently determine what you offer, who you serve, or why you're credible, it defaults to businesses where those signals are clear. Your competitor gets the referral not through superior marketing, but through superior machine readability.

Who's Being Left Behind (The Data)

Coven AI's analysis of 1,003 businesses reveals stark patterns in who's visible to AI agents and who isn't.

Globally, the average GEO/AI readiness score sits at just 36/100. That's not a passing grade by any measure. But the distribution matters more than the average: 39% of businesses score below 30/100, whilst only 13% score above 50/100.

Geographic disparities tell part of the story. North American businesses average 39/100 on AI readiness—still poor, but notably ahead of Latin America (33/100) and the Middle East (34/100). Europe, despite its digital infrastructure, scores just 35/100 on average.

Industry patterns are even more revealing. Gyms lead AI readiness at 42/100, likely because fitness businesses adopted structured local SEO early and that foundation translated well to agent visibility. Plumbers also score 42/100, benefiting from clear service taxonomies and strong local signals.

At the other extreme, office businesses score just 29/100 on AI readiness, and restaurants—despite being natural fits for AI recommendations—average only 34/100. Dentists (32/100) and estate agents (34/100) are similarly poorly positioned for the shift to agentic commerce.

These aren't small gaps. A dental practice scoring 32/100 is functionally invisible when a family asks Claude to "find a gentle dentist near us who works well with anxious children." A restaurant at 34/100 won't appear when someone asks Perplexity to "recommend romantic restaurants in Bristol with good vegetarian options."

Meanwhile, their competitors with even marginally better AI readiness—say, 45/100 instead of 34/100—capture those referrals by default.

Real-World Agentic Commerce Scenarios

Abstract statistics matter less than concrete examples. Here's what agent invisibility looks like in actual purchase journeys happening today.

The Claude Shopping Session

A startup founder asks Claude: "I need to set up company infrastructure. Recommend accounting software, a business bank, and a solicitor for UK company formation."

Claude provides specific recommendations with reasoning. It suggests Xero or FreeAgent for accounting, Tide or Starling for banking, and names two solicitor firms with relevant experience. Each recommendation includes a brief explanation of why that option fits the founder's context.

Five other accounting packages, three other business banks, and dozens of solicitor firms serve the exact same market. They didn't get mentioned. Not because they're inferior options, but because Claude couldn't confidently extract and evaluate the information needed to recommend them.

The businesses that got recommended had clear service descriptions, transparent pricing signals, and structured information about their ideal customers. The agent could verify they were legitimate, understand what they offered, and match them to the query. Everyone else was invisible.

The Perplexity Research Project

A marketing director asks Perplexity: "Compare marketing automation platforms for B2B SaaS companies with 20-50 person teams."

Perplexity generates a detailed comparison table covering six platforms. It includes pricing information, key features, integration capabilities, and specific recommendations based on use case. The research feels comprehensive.

But twelve other platforms serve exactly that market segment. Some are better fits for the director's specific needs. They didn't make the comparison because Perplexity couldn't reliably extract the information needed to include them.

The platforms that got compared had structured product pages, clear pricing information, and explicit declarations of their target customer. The platforms that got excluded had the same capabilities described in blog posts, case studies, and marketing copy that AI agents struggle to parse definitively.

The ChatGPT Local Discovery

A family relocating to Manchester asks ChatGPT: "Find gyms in South Manchester with childcare, swimming pools, and yoga classes."

ChatGPT recommends three specific facilities with descriptions of their amenities. One of those gyms gains a new family membership worth £150/month for an average of 18 months. Lifetime value: £2,700.

Four other gyms in South Manchester offer exactly those amenities. They lost a qualified prospect who was actively looking to join. The difference wasn't their facilities or pricing. It was whether ChatGPT could confidently determine they met the specified criteria.

The recommended gyms had structured data marking their amenities, clear location information, and machine-readable service descriptions. The invisible gyms mentioned childcare in a paragraph of body text, swimming in an image caption, and yoga in a class schedule PDF.

Why Traditional SEO Isn't Enough

Many businesses assume their existing SEO investment protects them. It doesn't.

Traditional search engine optimisation focused on keyword targeting, link building, and content creation. Those tactics helped humans find you through Google. They don't help AI agents understand and recommend you through conversational interfaces.

Coven AI's data shows the disconnect clearly. The average SEO score across 1,003 businesses is 56/100—mediocre but functional. The average GEO/AI readiness score is just 36/100. Businesses have invested in being findable but not in being comprehensible.

An AI agent doesn't care that you rank #3 for "commercial solicitors Edinburgh." It cares whether your website clearly declares your practice areas, service locations, and credentials in structured, verifiable formats. A blog post titled "Our expertise in commercial law" doesn't help. A schema-marked service page declaring "Commercial Law" as a practice area with attorneys credentialed in specific subspecialties does.

This isn't about abandoning SEO. It's about recognising that AI discovery requires a different technical foundation. You need both. But if you've optimised for human search engines and ignored machine comprehension, you're visible to Google and invisible to Claude.

Monitoring Your Agent Traffic Share

Most analytics platforms don't distinguish agent traffic from human traffic yet. That gap creates a dangerous blind spot.

When ChatGPT visits your website to evaluate whether to recommend you, it appears in your analytics as a single pageview. When Claude extracts your pricing information to include in a comparison, you see a brief visit from an unidentified user agent. When Perplexity indexes your service pages, it looks like bot traffic you probably filter out.

You can't optimise what you don't measure. Understanding your agent traffic share relative to competitors requires purpose-built monitoring that identifies AI agent visits, tracks which information they extract, and reveals when they recommend competitors instead of you.

The businesses winning in agentic commerce aren't necessarily spending more on marketing. They're measuring different metrics. They track:

Which AI agents visit their website and how frequently
Which pages and data points agents access during evaluation
Whether agents can successfully extract key business information
How often they appear in agent-generated recommendations versus competitors
Which queries lead to competitor citations in their category

This intelligence reveals exactly where agent invisibility is costing revenue and which technical gaps to prioritise fixing.

The Window Is Closing

First-mover advantage in AI agent visibility is substantial and compounding.

Every time an AI agent successfully extracts and verifies information from your website, it builds confidence in citing you. Every time it struggles to parse your offerings or can't verify your credentials, it deprioritises you in future recommendations. These patterns reinforce over time.

The businesses that establish strong agent visibility now will benefit from momentum effects as agentic commerce grows. The businesses that remain invisible will find themselves increasingly excluded from purchase conversations they don't even know are happening.

This isn't speculation about future trends. According to Coven AI's analysis, 39% of businesses are already effectively invisible to AI agents. Those businesses are losing revenue today to competitors whose only advantage is machine readability.

The question isn't whether to optimise for AI agent visibility. The question is whether you'll do it whilst there's still competitive advantage in being early, or wait until it's table stakes and you're fighting to recover lost ground.

What to Do Next

Start by understanding where you currently stand. You can't fix agent invisibility without first measuring it.

Run a free scan at aa.covenai.io to see exactly how AI agents perceive your website. You'll get a detailed AI readiness score plus specific technical gaps affecting your visibility to Claude, ChatGPT, Perplexity, and other agents mediating purchase decisions.

The scan identifies issues like missing structured data, unclear service declarations, unverifiable credentials, and semantic markup problems that make you invisible to agents even when you're perfectly visible to human visitors.

More importantly, you'll see how your AI readiness compares to competitors. If you're a restaurant scoring 34/100 whilst a competitor down the street scores 48/100, you now understand why they're getting recommended and you're not. That intelligence turns agent invisibility from a vague concern into a specific, fixable technical problem.

The businesses that win in agentic commerce won't be those with the biggest marketing budgets. They'll be the ones that made themselves comprehensible to the AI agents already recommending products, services, and vendors to your potential customers.

Your competitors are being cited right now. Make sure you're in the conversation.

Ready to see where your site stands? Run a free ACO scan at aa.covenai.io — no sign-up required.