Further Research - The Rise of AI Visitors and the Evolution of SEO

Tuesday, March 18, 2025

In the last six months, we’ve witnessed a surge in AI-driven searches and scraping. Rather than visiting websites directly, users are increasingly relying on ChatGPT, Perplexity, or Google’s forthcoming “AI Mode” to handle their browsing. As a result, many sites report anywhere from an 11% to 55% drop in organic search traffic, underscoring the shift from human visitors to AI visitors. People are simply reading summarised chatbot responses instead of going straight to your site.

Of course, I’m a stickler for facts and references. While my previous article explored how to prepare for the rise of AI visitors (and the ongoing shift from SEO to “LEO”), it’s vital to see the research behind these predictions. Below is an evidence-based companion piece—warts and all—showing the stats, sources, and trends driving these changes.

More AI “Visitors” Than Ever Before

Websites are now getting “read” by AI nearly as often as by humans. Recent data shows that automated bots – including AI crawlers and content scrapers – account for almost half of all internet traffic (techloot.co.uk & techloot.co.uk). In fact, 49.6% of global web traffic in 2023 was non-human, the highest share on record (techloot.co.uk). A significant driver of this trend is generative AI: as large language models (LLMs) like ChatGPT and others grow in popularity, they rely on ingesting web content via crawlers. The adoption of AI has boosted simple web-scraping bots to 39.6% of all bot traffic in 2023 (up from 33.4% in 2022) (techloot.co.uk) – a direct result of AI models indexing and extracting content for training and answering user queries.

Major AI systems are increasingly indexing and summarising website content on behalf of users. OpenAI, for example, deploys web crawlers (like its GPTBot) to gather internet data for ChatGPT (platform.openai.com). New AI search tools such as Perplexity.ai actively crawl sites in real time to generate answers, and even ChatGPT (via plugins or integrated browsing) can fetch content to respond to prompts. Google has also rolled out its AI Overview (AIO) feature in search results, which uses AI to summarise web content right on the search page. By late 2024, Google’s AI Overviews were appearing for roughly 33% of search queries (across 100+ countries where AIO was enabled) (emarketer.com) – meaning for one in three searches, Google’s AI is reading webpages and displaying an AI-written summary above the links. OpenAI’s ChatGPT-based “SearchGPT” and Perplexity are smaller players but growing; by October 2024, ChatGPT’s search integration was already sending four times more traffic to websites than Perplexity was (emarketer.com). Still, these AI referral traffic levels are tiny compared to Google’s – highlighting that Google’s search (now augmented with AI summaries) remains the dominant gateway (emarketer.com & emarketer.com).

Figure: Google’s AI Overviews (AI-generated summaries in search results) have rapidly become common. By Dec 2024, a BrightEdge study found a large share of queries – especially in specialised domains like healthcare, education, and B2B tech – showed an AI-generated overview at the top (red bars), whereas just months earlier in May 2024 many of those topics had few or no AI summaries (grey bars) (emarketer.com).

(emarketer.com) This illustrates how quickly AI “visitors” (Google’s algorithms) are now consuming website content to produce answers for users.

With AI agents increasingly accessing content, businesses are seeing a shift in their audience: it’s not just humans visiting your site, but also AI models reading and indexing it. In essence, your next website “visitor” might be an algorithm (like ChatGPT, Perplexity, Bing AI, or Google’s SGE) skimming your pages to answer someone’s question. This trend poses a challenge – and an opportunity – to adjust content strategy for this new kind of reader.

Users Rely on AI Summaries Instead of Visiting Websites

There is mounting evidence that internet users are clicking fewer websites and instead getting information from AI-driven summaries or chatbots. Consumer surveys show a clear behavioural shift: among Americans who have used AI chatbots, 35% said they used a chatbot to answer a question instead of using a search engine (bloggingwizard.com & bloggingwizard.com). In other words, a significant chunk of people now skip the traditional Googling-and-clicking process, preferring a bot to fetch answers for them. Another 35% have used chatbots to get something explained to them (bloggingwizard.com & bloggingwizard.com) – a role that might previously have involved reading an article or user manual. Notably, 14% even enlisted AI to summarise longer texts (bloggingwizard.com), letting the bot digest content so they don’t have to read the full source. These statistics underscore a broad trend: users are outsourcing their information-gathering to AI tools. As Consumer Reports observes, the most common uses of chatbots are “to answer a question, in lieu of a search engine, or to have the chatbot explain something” (innovation.consumerreports.org) – essentially replacing direct website visits with AI-curated answers.

This change in user behaviour is reflected in web traffic patterns. When an AI summarises content at the top of the search results (often called a “zero-click” answer), fewer people feel the need to click through to the source site. For example, Google’s own data (from an experiment by Growth Memo) found that the presence of an AI Overview on a search page led to 8.9% fewer clicks on results (emarketer.com). Over millions of searches, that represents a substantial loss of traffic to publishers. Early analyses of Google’s Search Generative Experience (SGE) predicted that widespread AI answers could cause organic search traffic to drop between 20% and 60% for publishers (searchengineland.com). One digital media network projected that once AI summaries fully roll out, they might see around a 25% average decline in search referral traffic across thousands of sites (searchengineland.com). In line with those predictions, a site management firm (Raptive) warned that overall Google traffic could fall by as much as half for some sites due to AI answered queries (emarketer.com). In short, many in the industry are bracing for a significant decline in human visits as AI answers siphon off clicks.

Real-world examples are already validating this shift. A dramatic case is Stack Overflow, the popular programming Q&A site. In early 2023, as developers flocked to ChatGPT for coding help, Stack Overflow’s page views plummeted – the site saw a 14% drop in traffic in just one month (March to April 2023) (gizmodo.com), and traffic was consistently down each month beyond typical trends. SimilarWeb data showed double-digit percentage declines coinciding with ChatGPT’s surge in popularity (gizmodo.com). Stack Overflow publicly acknowledged that the “surge of interest in ChatGPT” was indeed making a dent in traffic across not just their site but “many other sites” as well (gizmodo.com). In other words, people who used to visit Q&A forums or documentation pages are now often getting answers directly from an AI chatbot – bypassing the website entirely. This example may be extreme (Stack Overflow’s content is exactly the kind chatbots excel at), but it underscores a wider trend affecting publishers in various niches.

Importantly, Google’s own AI summary feature is now widespread, which may further reduce clicks. By November 2024, Google’s AI Overview (AIO) snippets were appearing for a third of all queries, and in certain categories (e.g. health, tech, education) the vast majority of searches now trigger an AI-generated answer at the top (emarketer.com). Publishers initially reported minimal impact, but as AIO expanded, even large media companies started seeing noticeable drops in Google traffic (emarketer.com). In August 2024, many claimed AI summaries hadn’t hurt them much, yet by year’s end Mediavine (which manages many independent publishers) observed a clear decrease in Google referral traffic across its sites (emarketer.com). Simply put, when users get what they need from an AI blurb on Google or an answer from ChatGPT, fewer of them click through to visit the actual source site. This marks a significant change in how audiences engage with online content.

Early Signs of Language Engine Optimisation (LEO)

To adapt to these shifts, marketers and SEO professionals are turning to a new concept: Language Engine Optimisation (LEO). Sometimes also called LLM Engine Optimisation, LEO refers to optimising your content to be understood, indexed, and favoured by AI language models – much like traditional SEO is about appealing to search engine algorithms. In practical terms, LEO means structuring and writing your website content in a way that AI systems (like Google AI, ChatGPT and Perplexity) can easily interpret it and confidently use it to answer questions. Industry experts describe LEO as ensuring that “AI-driven models correctly interpret, rank, and recommend your content,” as opposed to classic SEO which focused on pleasing the old search algorithms (saffronedge.com). The reason behind this shift is clear: AI chatbots and conversational search are replacing many traditional queries, and “AI models now summarise content and display answers without clicks.” (saffronedge.com) If your content isn’t readily usable by these AI systems, it might not be surfaced to users at all.

Over the past 6–12 months, early adopters of LEO strategies have begun to emerge. SEO specialists are blogging about how to optimise for ChatGPT, Perplexity, Google AI, Bing AI, and other LLM-based tools. For example, marketers advise writing in a clear, conversational tone and using explicit Q&A formats, so that an AI can easily extract question-answer pairs from your page (wallaroomedia.com & wallaroomedia.com). They also emphasise adding structured data and schema markup (like FAQ schema, How-To steps, etc.) to make the content machine-friendly (wallaroomedia.com). The goal is to help AI algorithms identify the context and key points of your content. A website that is well-structured, semantically rich, and authority-backed is more likely to be chosen by an AI as a source. On the flip side, if a page is disorganised or unclear, an AI might either ignore it or misinterpret it, resulting in your information being left out of AI-generated answers.

Some tangible examples of LEO in action: Companies are starting to monitor how their brand or content appears in chatbot responses, then tweaking content accordingly. One recommended tactic is to literally ask ChatGPT or Perplexity, “What’s the best X in [my domain]?” and see if your brand or page is mentioned; if not, that’s a signal you need to build more authoritative content on that topic (wallaroomedia.com). SEOs suggest using these AI responses as feedback – if the AI doesn’t cite you, figure out why and fill that gap. Another emerging practice is feeding AI with your content proactively (for instance, offering structured data or submitting to AI indices) so that the models learn about your site. We’re also seeing toolmakers coin new acronyms (LLMO, GEO – Generative Engine Optimisation – etc.) but all boil down to the same idea: making content AI-ready. While it’s still early days, businesses that depend on search traffic are wisely experimenting with LEO techniques now. Ensuring that chatbots “know” your brand and can accurately summarise your pages could be crucial for staying visible as AI-driven search grows.

SEO Tactics Evolving, Not Completely New

It’s important to note that these LEO concepts aren’t entirely new inventions – they’re really an evolution of modern SEO best practices. In many ways, what’s good for AI models overlaps with what’s good for human users and traditional search engines. High-quality, well-organised content has always been the cornerstone of SEO, and that remains true in the AI era. The referenced article’s ideas are not revolutionary novelties; they build on the trajectory SEO has already been on. For years, Google’s algorithm updates have been rewarding content that is relevant, authoritative, and user-friendly, while punishing spammy tactics. So, when we talk about “optimising for AI,” it largely means continuing to focus on clarity, depth, and structured information – just with an eye towards how an algorithm will parse it.

Crucially, outdated SEO tricks like keyword stuffing are long dead. Search engines have become highly sophisticated at detecting and penalising such manipulative practices (betterlinks.io & betterlinks.io). In 2024, “keyword stuffing is not only ineffective but can also damage your SEO efforts.” (pickageek.com) Google’s algorithms (e.g. Panda and subsequent updates) actively demote sites that cram in keywords unnaturally (betterlinks.io). This push away from cheap gimmicks has forced SEO to mature towards more semantic, user-intent-focused approaches – which directly aligns with what AI chatbots need as well. An AI like ChatGPT isn’t counting your keyword density; it’s trying to understand the meaning and facts in your content. Thus, the same old-school hack that fails SEO will certainly fail LEO. There’s no “gaming” a chatbot with hidden text or repetitive phrases – if anything, those will just confuse the AI or be filtered out.

Think of LEO as SEO for a new primary audience: AI readers. Rather than solely writing for human eyeballs and hoping Google’s bot likes it, you now must assume both humans and algorithms will read your page directly. In practical terms, that means your content should be explicit, well-structured, and rich in context, so that an AI can grasp it without misinterpretation. One SEO specialist calls this “AI readability optimisation” – structuring content for AI to process, extract, and summarise effectively (gravitatedesign.com). As they bluntly put it, “AI search is exploding… If your content isn’t optimised, it won’t get cited, seen, or clicked.” (gravitatedesign.com)

In other words, if an AI can’t easily pull useful information from your site, it will simply use someone else’s content in its answer. By contrast, content that is clear, conversational, and authoritative is more likely to be selected by AI systems and presented to users (often with a citation link). So the mindset shift for businesses is to consider how an algorithmic reader will digest your page. Are your headlines and subheadings descriptive? Are you directly answering common questions in your text? Is important information easy to find (for example, in bullet points or summary boxes)? These factors all improve human UX and also make it straightforward for an AI to identify the key points of your page.

In summary, SEO is evolving – not ending. The rise of AI chatbots and generative search is forcing a natural progression in optimisation tactics. The fundamental goal is still the same: deliver valuable, well-structured content that meets users’ needs. The twist is that now the delivery mechanism includes AI intermediaries. Businesses should embrace this by optimising content for dual consumption: appealing to human readers while also being machine-friendly for AI interpretation. Strategies like LEO simply build upon established SEO principles (high-quality content, semantic richness, technical clarity) and extend them to the context of large language models. The era of “write for the user, and for the algorithm” now means writing for an AI agent as an intermediary for the user. Companies that recognise this shift – and adjust their content approach accordingly – will be well positioned to maintain visibility in an AI-driven search landscape, whereas those clinging to old tactics (or ignoring AI altogether) risk seeing their organic traffic dwindle as the web’s traffic patterns continue to transform.

Sources: Recent industry reports, surveys, and expert commentary have been used to support these insights, including BrightEdge and eMarketer data on AI-driven search results (emarketer.com & emarketer.com), Consumer Reports’ 2023 chatbot usage survey (bloggingwizard.com & bloggingwizard.com), the Imperva 2024 Bot Traffic report (techloot.co.uk & techloot.co.uk), and SEO thought leadership on LLM optimisation best practices (saffronedge.com & gravitatedesign.com). These demonstrate the trend of AI taking a larger role in content consumption and the corresponding need for evolved SEO tactics.