SEO Is Dying: How Businesses Must Adapt for AI Search and Generative Engines

Summary

Search is being rebuilt around answers, recommendations, and summaries. That changes the commercial problem for brands.

For two decades, digital visibility mostly meant ranking well for the right keywords. The operating model was clear enough: technical SEO, authority links, content volume, page quality, and conversion paths. That model still matters. But it no longer describes the whole market.

During a conversation with Tom Mason at a Canon event, a recurring idea emerged: organisations are no longer simply competing for search ranking, but for how AI systems understand and represent them. That led me to look more closely at the shift.

Tom’s company, AwarenessAI, frames as AI visibility and representation. The point is if customers now ask ChatGPT, Gemini, Claude, Perplexity, Copilot, or Google’s AI Overviews what to buy, who to trust, and which vendor to shortlist, and they get the right responses (based on your brand) on these engines, then your brand contest moves upstream.

AI systems increasingly act as an interpretation layer between the market and the company. They summarise positioning, compare vendors, cite sources, infer trust, and sometimes leave brands out altogether. That creates a new competitive asset: Generative Engine Optimisation, or GEO.

GEO is not a shiny replacement for SEO. It is the discipline of making a business legible, credible, and citable to AI-mediated discovery systems. The firms that treat it as a governance, data, content, and reputation problem will move faster than those that treat it as another keyword channel.

Main Arguments

1. Traditional SEO is no longer enough because AI interfaces compress research journeys into generated answers.
2. Zero-click search was already weakening the website visit as the default outcome; AI Overviews and answer engines make that pattern stronger.
3. AI visibility depends on entity clarity, third-party corroboration, structured data, citations, and semantic authority, not keyword density alone.
4. Representation risk becomes a board-level brand issue when AI systems describe a company incorrectly, use stale information, or blend similar entities.
5. Marketing, PR, product, legal, data, and investor relations now share one problem: the public machine-readable version of the company must match the intended market position.

Why Traditional SEO Is Losing Its Monopoly

Use Google’s own product moves as evidence. AI Overviews reached broad rollout after the May 2024 launch in the US, expanded to more than 100 countries in October 2024, and reached more than 200 countries and territories across more than 40 languages by May 2025. Google also introduced AI Mode, a more conversational search interface.

Then use behavioural data. SparkToro and Datos estimated that 58.5% of US Google searches and 59.7% of EU Google searches ended with zero clicks in 2024. Pew Research Center later found that in March 2025, Google users clicked a traditional result in 8% of visits with an AI summary, compared with 15% without one. Source-link clicks inside AI summaries occurred in only 1% of visits.

A high ranking can still create value, but it may no longer create the visit. Visibility and traffic separate.

What Changes Technically

Embeddings turn words, pages, companies, people, products, and topics into mathematical representations. Search then becomes a similarity problem, not only a keyword match.

Vector retrieval finds content that sits close to a user’s intent, even if the wording differs.

Retrieval augmented generation (RAG) combines a language model with retrieved documents. The model drafts the answer; the retrieval layer decides what evidence enters the room.

Knowledge graphs and entity systems decide whether “Canon” means the camera brand, a legal principle, a religious role, or an event sponsor. The same logic affects SMEs with common names, merged founders, stale subsidiaries, and renamed services.

Structured data, schema markup, consistent profiles, product feeds, case studies, FAQs, documentation, analyst mentions, media coverage, and review ecosystems all become machine-readable evidence.

What GEO Really Means

The paper “GEO: Generative Engine Optimization” formalised generative engines as systems that synthesise responses from multiple sources and studied methods for improving source visibility. The authors reported that GEO methods could improve visibility by up to 40% in their tests.

GEO is not “write for robots.” It is the management of evidence across systems that models can retrieve, trust, compare, and cite.

The practical components:

• Entity clarity: models must know who the business is, what it sells, where it operates, and how it differs from similarly named entities.
• Corroboration: claims need support from credible third-party sources, not only the company website.
• Citation fitness: pages must answer specific questions clearly enough for an AI system to cite them.
• Semantic coverage: content should map to buying questions, comparison queries, objections, integrations, risks, and use cases.
• Freshness: stale pages and old profiles create model drift.
• Governance: someone must own AI representation as a recurring business control.

AI Representation Risk

AI systems can misstate facts in ways that normal analytics tools never show.

A model may describe a cybersecurity consultancy as a generic IT services firm. It may pull an old headquarters address. It may confuse a founder with another person of the same name. It may omit regulated credentials, understate enterprise experience, or recommend competitors because third-party sources make them easier to verify.

Public cases show the risk. In 2024, Air Canada had to compensate a customer after its chatbot gave incorrect bereavement fare information. Google also publicly addressed erroneous AI Overviews after users shared examples of odd outputs during the early US rollout.

Those examples differ from brand discovery, but they prove the same operating reality: customers can act on machine-generated statements, and companies carry the cost when those statements mislead.

Business Implications

Marketing teams lose the comfort of a single channel metric. Rankings, impressions, and organic sessions still matter, but they no longer capture whether AI systems recommend the brand.

SEO agencies must expand into content architecture, entity management, digital PR, structured data, source testing, and AI visibility measurement. Some will rebrand old SEO packages. Better firms will build new test methods across ChatGPT, Gemini, Claude, Perplexity, Copilot, and Google AI features.

Consulting firms get a new advisory lane: AI-mediated market presence. This sits between growth strategy, brand governance, data architecture, and risk. It will matter most in sectors with complex buying journeys: professional services, healthcare, cybersecurity, financial services, B2B software, manufacturing, education, and high-consideration consumer products.

Investor perception also shifts. Analysts, journalists, and potential acquirers increasingly use AI tools to build first-pass views of companies and markets. If AI systems return weak, inconsistent, or outdated company narratives, the problem is not cosmetic.

The AI Visibility Operating Model

Organisations need a working model that treats AI visibility as an operating capability, not a campaign.

1. Baseline AI Visibility Audit

Test the prompts real buyers use. Include category prompts, comparison prompts, problem-led prompts, local prompts, integration prompts, risk prompts, and investor-style prompts.

Track four outputs: whether the brand appears, how it is described, which sources the system uses, and which competitors appear nearby.

2. Entity and Knowledge Graph Clean-Up

Audit structured data, social profiles, business directories, Wikidata or Wikipedia where appropriate, Crunchbase, Companies House, Google Business Profile, product directories, partner pages, review platforms, and industry databases.

The goal: one coherent machine-readable identity.

3. Authoritative Content Ecosystem

Build content around questions AI systems must answer to recommend the company responsibly. That includes:

• “Who is this company for?”
• “What evidence supports its claims?”
• “How does it compare with alternatives?”
• “Which sectors does it serve?”
• “What risks or limits should buyers know?”
• “What proof exists outside its own website?”

4. Citation Engineering

Treat citations as distribution infrastructure. Publish pages with clear factual answers, named authors, dates, schema markup, references, and stable URLs. Earn mentions in sources that AI systems already retrieve: industry reports, credible news, partner sites, software marketplaces, procurement portals, standards bodies, and specialist forums.

This is not link spam. It is evidence design.

5. Multimodal Discoverability

AI search increasingly reads images, video, audio, product data, and PDFs. Optimise transcripts, alt text, product specs, demo videos, slides, and technical documents. Make the business understandable across formats.

6. AI Reputation Governance

Create ownership. Marketing can’t manage this alone.

Set a monthly review cadence. Log harmful outputs. Keep a source register. Track outdated facts. Define escalation paths for legal, regulatory, or investor-sensitive errors. For larger firms, connect this to brand risk and AI governance committees.

7. Measurement

Do not reduce GEO to one score. Use a small measurement set:

• Share of model: how often the brand appears across target prompts.
• Recommendation position: whether it appears first, later, or only with caveats.
• Representation accuracy: whether descriptions match the company’s intended position.
• Citation quality: which sources AI systems use.
• Competitor adjacency: which firms appear in the same answers.
• AI-attributed demand: referral traffic, sales notes, form fields, and customer interviews that mention AI tools.

How Businesses Should Adapt

Start with the buying journey, not the tool list. Identify the questions a buyer, journalist, analyst, investor, or candidate might ask an AI system before they ever visit the website.

Run those prompts across ChatGPT, Gemini, Claude, Perplexity, Copilot, and Google Search with AI features. Save the outputs. Repeat the same prompt set monthly. The first pass will usually expose errors: unclear positioning, missing proof, weak third-party validation, outdated sources, or competitors that own the category language.

Then fix the evidence base.

Update the company site so core pages answer machine-readable questions directly. Add organisation, product, FAQ, article, person, and review schema where relevant. Keep founder, leadership, address, product, sector, and credential data consistent across the web. Publish comparison pages carefully; buyers ask AI systems to compare, so refusing to discuss alternatives leaves the answer to someone else.

Strengthen third-party sources. AI systems lean on corroboration. That means partner pages, industry directories, earned media, credible guest commentary, case studies, standards references, marketplace listings, customer reviews, conference pages, and analyst databases. A brand’s own site can say what it wants. The wider web teaches models whether to believe it.

Rewrite content for retrieval. Pages should have clear titles, direct answers, dates, named authors, source links, and plain product facts. Avoid vague category claims. If the company serves mid-market UK healthcare providers with secure patient messaging software, say that. Machines need specificity; so do buyers.

Build an internal AI representation register. Record the most common AI descriptions of the company, harmful errors, missing proof points, recurring competitor mentions, and source dependencies. Assign an owner. Review it with marketing, communications, legal, product, and sales.

Finally, connect GEO to revenue operations. Ask inbound leads how they researched the company. Add “AI assistant” or “AI search” as an attribution option. Train sales teams to ask what prospects saw in ChatGPT, Perplexity, Gemini, or Google AI answers. The market signal will show up in conversations before dashboards catch it.

Counterarguments

It is only fair to note that SEO is not dead. Search engines still crawl pages, rankings still shape discovery, and Google still sends huge volumes of traffic. For many transactional searches, local searches, and branded queries, classic SEO remains profitable.

The stronger claim is narrower: SEO no longer controls digital discovery by itself. The ranking page has become one input into a larger answer ecosystem.

Another risk is that GEO becomes a new manipulation industry. The academic literature already points to adversarial optimisation risks. If brands can push models toward preferred answers through repetition, citation theatre, or synthetic authority, users may get worse information. That is why governance matters. GEO must focus on accuracy, evidence, and clarity, not trickery.

Supporting Research Sources

• Google launched AI Overviews broadly in the US in May 2024 and described the system as integrated with its core ranking systems, not a standalone chatbot. Source: Google Search, May 2024.
• Google expanded AI Overviews to more than 100 countries in October 2024, reaching more than 1 billion monthly users. Source: Google Search, October 2024.
• Google said AI Overviews reached more than 200 countries and territories and more than 40 languages by May 2025. Source: Google Search, May 2025.
• Google introduced AI Mode as a more conversational Search interface for complex and follow-up queries. Source: Google Search, March 2025.
• Pew Research Center found lower click-through behaviour when AI summaries appeared in Google Search. Source: Pew Research Center, July 2025.
• SparkToro and Datos estimated high zero-click rates in both the US and EU. Source: SparkToro, July 2024.
• The original GEO paper studied optimisation methods for generative engines. Source: GEO: Generative Engine Optimization.
• The RAG paper explains why retrieved evidence changes generated answers. Source: Lewis et al., 2020.
• Air Canada’s chatbot case shows how incorrect AI-mediated information can create liability. Source: The Guardian, February 2024.
• AwarenessAI provides a market example of AI visibility and representation moving into specialist services. Source: AwarenessAI.

Statistics With Citations

• 58.5% of US Google searches and 59.7% of EU Google searches ended with zero clicks in the 2024 SparkToro and Datos study. Source: SparkToro.
• For every 1,000 Google searches, SparkToro estimated 360 US clicks and 374 EU clicks went to the open web. Source: SparkToro.
• Pew found that 18% of Google searches in its March 2025 dataset generated an AI summary. Source: Pew Research Center.
• Pew found that users clicked a traditional result in 8% of visits with an AI summary, compared with 15% without one. Source: Pew Research Center.
• Pew found that users clicked a source link inside an AI summary in 1% of visits. Source: Pew Research Center.
• Google said AI Overviews were used by more than a billion people by March 2025. Source: Google Search.
• Google said AI Overviews covered more than 200 countries and territories and more than 40 languages by May 2025. Source: Google Search.
• The GEO paper reported visibility gains of up to 40% from its tested methods. Source: GEO: Generative Engine Optimization.

Academic References

• Aggarwal, P., Murahari, V., Rajpurohit, T., Kalyan, A., Narasimhan, K. R., & Deshpande, A. (2023). “GEO: Generative Engine Optimization.” arXiv:2311.09735.
• Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., et al. (2020). “Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks.” arXiv:2005.11401.
• Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021). “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?” FAccT ’21.
• Nestaas, F., Debenedetti, E., & Tramèr, F. (2024). “Adversarial Search Engine Optimization for Large Language Models.” arXiv:2406.18382.
• Chen, X., Wu, H., Bao, J., Chen, Z., et al. (2025). “Role-Augmented Intent-Driven Generative Search Engine Optimization.” arXiv:2508.11158.

Industry Reports and Market Signals

Google’s product direction points toward AI search as a core interface, not an experiment. AI Overviews, AI Mode, multimodal search, shopping assistants, and generated comparison answers all make the search page more like an operating layer for decisions.

Publisher data points in the same direction. The debate over AI Overviews centres on whether AI summaries reduce site visits, weaken publisher economics, and concentrate attention inside platforms. Even when Google argues that AI features improve satisfaction and send higher-quality visits, the commercial question remains: fewer but higher-intent visits still require a new measurement model.

Agency movement matters too. AwarenessAI, Brand Echo, Presenc AI, Storyzee, and other new providers now sell AI visibility monitoring, AI representation audits, and GEO services. Some are early, some are noisy, and the category still needs maturity. But agency formation usually follows budget formation. That is a market signal.

Relevant Quotes

• Google said AI Overviews are “not simply generating an output based on training data.” Source: Google.
• Google said AI Mode lets people “ask whatever’s on your mind.” Source: Google.
• Pew wrote that users “very rarely clicked on the sources cited.” Source: Pew Research Center.
• The Air Canada tribunal rejected the idea that a chatbot could sit apart from the company’s responsibility. Source: The Guardian.

Future Predictions

Websites will not disappear. They will lose their role as the first stop for many discovery journeys.

The company website becomes an evidence hub: the source AI systems can retrieve, cite, and verify. Brands will still need pages, but the pages must serve both humans and machines.

Brands will compete to become AI-cited entities. In some categories, appearing in an answer will matter more than ranking fourth in search results. In B2B, the first AI-generated shortlist may shape the rest of the buying process.

Large enterprises will have an advantage because they already produce public proof: analyst coverage, partner pages, press, certifications, reviews, annual reports, and procurement documents. SMEs will need sharper positioning and better third-party validation to avoid invisibility.

Consulting firms will package AI visibility into growth, brand, and risk advisory. The best work will combine SEO, knowledge graphs, data governance, digital PR, customer research, and commercial strategy. The weaker work will sell dashboard scores without fixing the source problem.

Search becomes recommendation infrastructure. That is the real shift. The interface does not merely retrieve information; it decides what the user should consider next.

Suggested Charts and Visuals

• Traffic flow chart: classic SEO funnel versus AI-mediated discovery funnel.
• Matrix: SEO ranking factors compared with GEO visibility factors.
• AI representation risk map: inaccurate, outdated, omitted, merged, overclaimed, or mispositioned.
• Operating model diagram: audit, entity clean-up, content evidence, citation engineering, monitoring, governance.
• Metric dashboard mock-up: share of model, citation quality, representation accuracy, competitor adjacency, AI-attributed demand.
• Buyer journey diagram: user prompt, AI answer, cited sources, brand shortlist, website visit, sales conversation.

Conclusion

SEO trained businesses to think in pages, rankings, and clicks.

AI search trains the market to think in answers, entities, evidence, and recommendations.

That does not make websites irrelevant. It makes them part of a wider credibility system. The companies that win will not be the ones that chase every new AI tool. They will be the ones that make themselves easier to understand, verify, cite, and recommend.

The next search battle is not only for attention. It is for accurate representation.

References

• Google: AI Overviews, about last week
• Google: AI Overviews in Search are coming to more places around the world
• Google: AI Overviews expand to over 200 countries and territories, more than 40 languages
• Google: Expanding AI Overviews and introducing AI Mode
• Pew Research Center: Google users are less likely to click on links when an AI summary appears
• SparkToro: 2024 Zero-Click Search Study
• GEO: Generative Engine Optimization
• Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
• The Guardian: Air Canada ordered to pay customer who was misled by airline's chatbot
• AwarenessAI