Google still processes around 8.5 billion searches per day. But something fundamental has changed about what happens after those searches. 68% of them now end without a single click to any website. And a growing portion never go to Google at all.
ChatGPT reached 1 billion monthly active users in June 2026, processing over 2 billion queries daily. Perplexity handles 780 million queries per month. Claude, Gemini, and Bing Copilot are growing alongside them. People are no longer just searching. They are asking, and they are getting answers back instead of a list of links to choose from.
This is not a future trend. It is happening right now, and it has a name: Generative Engine Optimisation, or GEO. And if you have been focused exclusively on traditional SEO, you may be optimising for a channel that is quietly shrinking while the new one grows around you.
What Is SEO and What Is GEO?
Search Engine Optimisation (SEO) is the practice of getting your website to rank highly on Google and Bing. It works by matching keyword strings, earning backlinks, and signalling authority through domain trust. The output is a ranked list of URLs. You click one. You visit a website.
Generative Engine Optimisation (GEO) is the practice of getting your content cited in AI-generated answers. It works by making your content easy for AI systems to extract, understand, and quote. The output is a synthesised answer, sometimes with citations, sometimes without. You may or may not visit a website at all.
The term was formally defined in a peer-reviewed paper published by researchers from Princeton University, IIT Delhi, Georgia Tech, and the Allen Institute for AI in November 2023 (arXiv:2311.09735, presented at KDD 2024). Their work established the first academic framework and benchmark for optimising content for generative search engines, and the findings have significant implications for how businesses should think about online visibility.
SEO vs GEO at a Glance
The Evidence: What Is Actually Happening to Search Traffic
The numbers are stark enough that it is worth spending a moment on them, because they explain why GEO is no longer optional for businesses that rely on organic search.
Zero-click search is the new normal
When Google's AI Overviews appear, 83% of searches end without a click to any external website. In Google's newer AI Mode, that figure rises to 93%. Organic click-through rates for Position 1 results have dropped to just 2.6% when an AI answer is present, down from historical averages of 25-35%.
This is not a small adjustment. Between 2024 and 2026, the share of searches generating at least one click fell by 9.51 percentage points, a relative decline of 22.9% in two years. 5
The HubSpot case study
HubSpot's blog is one of the most documented casualties of AI-driven search. The company went from 24.4 million monthly organic visits in March 2023 to 6.1 million by January 2025, a collapse of 75% in under two years. CEO Yamini Rangan acknowledged the cause directly on an earnings call: "Organic search traffic is declining globally. AI overviews are giving answers, and fewer people are clicking."
HubSpot was not doing anything wrong by SEO standards. Their content was high-quality, well-linked, and regularly updated. The problem was that their content was optimised for a document-ranking system, and they were operating in a passage-extraction system. The rules had changed.
News publishers collectively lost 600 million monthly visits between mid-2024 and May 2025. AI-driven search is not a future disruption. It is a current reality.
The Science Behind GEO: What the Princeton Research Found
The academic paper that defined GEO (Murahari et al., 2023) tested nine specific content strategies against a benchmark of 10,000 diverse search queries spanning nine datasets including MS MARCO, Natural Questions, LIMA, and ELI-5. The results were unambiguous about what works and what does not.
How Each AI Platform Decides What to Cite
Not all AI search engines work the same way, and understanding how each platform retrieves and selects sources matters for where you focus your GEO efforts.
- Two-layer system: static training data plus real-time Bing retrieval
- Bing retrieval activates for commercial-intent queries (terms like "reviews," "comparison," or a year)
- Licensed media partners get preferred access
- 76% of citations come from content published in the last 30 days
- Favours Wikipedia for factual queries (47.9% of top citations in one analysis)
- Real-time web retrieval for every query; new content can appear within hours
- Retrieves 5 to 10 candidate pages, reranks, cites 3 to 4
- Favours Reddit for many queries (46.7% of top citations in one analysis)
- Favours review platforms (G2, Capterra) for commercial queries
- Visitors from Perplexity convert at 10.5% vs 1.76% Google organic average 4
- Draws from Google's own organic index; E-E-A-T signals weighted heavily
- Uniquely favours YouTube (23.3% of citations)
- AI Overview citations from top-10 organic results dropped from 76% to 38% in eight months (Ahrefs, 2025 to 2026)
- Position 1 carries 53% citation probability; Position 10 drops to 36.9%
- Named authors with credentials and clear sourcing strongly rewarded
- Uses Brave Search for web queries
- Cites at sentence level; each cited sentence must stand alone
- Cites user-generated content at 2 to 4 times the rate of competitors
- Favours blog content (43.8% of citations in one analysis)
- Each cited sentence must be independently verifiable
Despite their differences, all four platforms reward the same underlying content properties: a direct answer in the opening sentence, named entities, dated and sourced facts, self-contained paragraphs, and structured data markup.
Backlinks vs Brand Mentions: The Biggest Structural Difference
In traditional SEO, the primary off-page signal is the backlink. A link from a high-authority domain signals that your content is trusted. The link graph is the foundation of Google's PageRank system.
In GEO, the primary off-page signal is the brand mention. An Ahrefs study of 75,000 brands found that the correlation between brand mentions and AI visibility was 0.664, compared to just 0.218 for backlinks. Brands in the top quartile by mention volume received 10 times more AI citations than the next quartile.
This means the GEO equivalent of link building is brand mention building: press coverage, podcast appearances, forum discussions, analyst reports, industry roundups, and anywhere else your brand name and area of expertise appear together across the web.
The Technical GEO Checklist: What You Should Actually Do
Based on the Princeton research and the citation pattern data above, here is what GEO optimisation looks like in practice.
44.2% of AI citations come from the first 30% of content (SparkToro, January 2026). Stop writing contextual preamble. Lead with the direct answer in sentence one, then provide supporting detail. Structure: "[Topic] is [definition]. [Key fact]. [Why it matters]."
Statistics Addition was tied for the highest-impact strategy in the Princeton paper, producing 30 to 40% visibility gains. Replace every qualitative claim with a number. "Growing rapidly" becomes "growing 357% year-over-year." "Many businesses" becomes "67% of businesses."
Cite Sources was the single most effective GEO technique. Link to the study, the report, the institution. AI models are more likely to cite a source that itself cites sources. Academic papers, government reports, and .edu/.gov domains carry the highest weight.
Every paragraph should be able to stand alone as a cited answer to a specific question. Use H2 and H3 headers as complete questions. Write each section so it answers one query without requiring context from other sections. AI systems extract passages, not pages.
The llms.txt protocol is to AI crawlers what a sitemap is to Google. Place a curated Markdown file at /llms.txt listing your most important pages with brief summaries. Update robots.txt to explicitly allow GPTBot, ClaudeBot, PerplexityBot, and Google-Extended.
Use Organization, Person, Article, and HowTo Schema.org markup with @id URIs and sameAs references to LinkedIn, Wikidata, and other authority sources. This connects your brand to knowledge graph nodes inside AI models.
Comparison articles drive 32.5% of all AI citations (Previsible, 2025). "X vs Y" and "Best tools for Z" formats are structurally optimised for AI retrieval because they answer specific, high-intent queries with clear, extractable conclusions.
76.4% of ChatGPT citations come from content published in the last 30 days. Content freshness is a stronger GEO signal than an SEO signal. Add a "Last updated" date prominently and update statistics and references regularly.
What SEO Still Does That GEO Cannot Replace
GEO and SEO are not opposites. They are complementary, and strong traditional SEO remains a prerequisite for GEO performance. Here is why.
Between 38% and 54% of Google AI Overview citations still come from top-10 organic results (BrightEdge data). That figure is declining, falling from 76% in mid-2025 to around 38% by early 2026, but it has not reached zero. Domain authority still acts as a citation filter: low-authority domains face a higher bar to be cited by AI systems.
More fundamentally, pages that cannot be crawled by Google or Bing cannot be retrieved by AI Overviews or ChatGPT respectively. SEO's foundational work — crawlability, indexability, site speed, structured data, E-E-A-T signals — directly feeds GEO. You cannot skip it.
Think of it this way: SEO is the foundation. GEO is the layer on top. You need both.
Where to Start
If you have an existing website and you are not sure where your GEO readiness stands, the most useful place to start is a simple audit of three things:
- Can AI bots crawl your site? Check your
robots.txtfile. If it blocksGPTBot,ClaudeBot, orPerplexityBot, you are invisible to those AI systems regardless of your content quality. - Do you have an
llms.txtfile? If not, you are leaving structured context on the table. - Does your content answer first? Take your five most important pages and check whether the first paragraph gives a direct, self-contained answer to the query that page targets. If it starts with context rather than an answer, it needs rewriting for GEO.
At Sonder Digital, GEO and AI search optimisation is baked into every website we build. robots.txt that permits all major AI crawlers, llms.txt, full Schema.org structured data, and content structured for passage extraction are all part of our standard build. If your current website was not built with this in mind, it is worth having a conversation.
Want to know how your website performs for AI search?
Book a free discovery call and we will walk you through where you stand and what would make the biggest difference.
Book a Free CallReferences
- Search Engine Land. Google Zero-Click Searches 2026 Study. searchengineland.com
- DemandSage. ChatGPT Statistics 2026. demandsage.com
- OmniBound. Generative Engine Optimization Statistics. omnibound.ai
- Discovered Labs. How ChatGPT, Claude, Perplexity and Google AI Overviews Cite Sources Differently. discoveredlabs.com
- Click Vision. Zero-Click Search Statistics. click-vision.com
- Murahari, V., et al. (2023). GEO: Generative Engine Optimization. arXiv:2311.09735. Princeton University, IIT Delhi, Georgia Tech, Allen Institute for AI. arxiv.org
- Ahrefs. AI Overview rank overlap analysis, 16 months of data. brightedge.com
- DemandSage. Perplexity AI Statistics 2026. demandsage.com
- Alphametic. HubSpot's SEO Crash and AI Overviews. alphametic.com
- Digital Agency Network. Generative Engine Optimization Statistics. digitalagencynetwork.com
- Yext. How ChatGPT, Perplexity, Gemini, and Claude Decide What to Cite. yext.com
- WPEngine. SEO vs GEO. wpengine.com
