LLM-powered search looks less like a passing fashion and more like a structural change. ChatGPT, Perplexity, Gemini and Meta AI keep adding users at a pace traditional engines have not seen in years, while visits to Google, Yahoo and DuckDuckGo have flattened or declined. The open question is timing, not direction. For anyone who publishes online, that makes Generative Engine Optimization (GEO) a present concern rather than a future one.
Generative search is information retrieval through large language models that answer a question directly with a synthesized response, instead of returning a list of links to visit. If you want the mechanics behind it, start with our explainer on how large language models work.
How fast are LLM search tools growing?
The headline platforms report user numbers that would have sounded implausible three years ago.
- ChatGPT passed 600 million monthly active users in early 2025, up from 300 million weekly users in late 2024, according to an OpenAI stakeholder quoted by Reuters.
- Google Gemini claimed 350 million monthly active users as of March 2025, a figure revealed in a court hearing and reported by TechCrunch.
- Meta AI counted roughly 700 million monthly active users in January 2025, close to a 20% jump on the previous month, according to a Meta statement reported by CNBC.
- A March 2025 research note from Aitools found that total chatbot search queries grew more than 80% between April 2023 and March 2025.
Is traditional search actually shrinking?
Traditional engines still dominate in absolute terms. In October 2024, Google still handled around 83.5% of all search-type queries, according to SparkToro. Growth, though, has stalled.
Flat is a new experience for an industry that expanded for two decades. When the default engine stops growing while the challengers double, the trend line matters more than the market share.
As the Aitools analysis and Statista's quantitative research show:
- Total annual visits to Google fell 1.41% year over year.
- Visits to Yahoo and DuckDuckGo dropped 22.50% and 8.77% respectively.
- Visits across all engines slipped 0.51% year over year.
User satisfaction shows cracks as well. John Kiernan, a search expert at WalletHub, put it plainly:
The quality of Google search results in the personal finance vertical has been declining for years, according to WalletHub data.
The decline in Google search quality could be the result of poor execution or honest mistakes [...]
When could LLMs overtake traditional search?
Nobody knows the crossover date, and forecasts should be read as directions rather than deadlines. Three projections frame the range:
- Marcus Tober of Semrush modelled that if ChatGPT sustains its roughly 13% monthly growth, it could rival Google's search volume in under five years.
- Gartner predicted in 2024 that AI chat interfaces would displace 25% of traditional search queries by 2026.
- Morningscore reported that AI search tools could handle 14% of global queries by 2028, up from around 6% in 2024.
The estimates differ widely, which is itself informative: analysts disagree on speed, not on direction. From what the platforms have shared publicly, none of the curves point backwards.
Source: Morningscore, based on data from Similarweb, Statista and Datos
What should publishers do about it?
As behaviour moves toward conversational search, SEO alone no longer covers the whole game. GEO extends it: a strategic approach that keeps content structured, authoritative and machine-readable so AI models can interpret it and cite it. GEO does not replace SEO. It is the next stage of the same discipline, part of the longer arc we traced in the evolution from search engines to generative engines.
Three moves cover most of the ground.
- Make your content machine-readable. LLMs extract content differently from ranking crawlers. Structured data, clean HTML and canonical URLs raise the odds that your pages are represented accurately inside AI answers. An llms.txt file helps models find your key pages.
- Put authority above keywords. LLMs favour high-quality content they can attribute and verify. Build E-E-A-T: experience, expertise, authoritativeness and trustworthiness.
- Keep publishing educational articles. Blog posts account for more than 77% of LLM referral traffic. Factual, contextual writing is what gets cited.
A shift worth preparing for
Search is fragmenting. Google and the other incumbents now share attention with LLM platforms that keep growing in daily utility and trust. Chatbots already rival smaller search engines in usage, their growth rates outpace traditional search by a wide margin, and within a few years they could plausibly become primary entry points for information.
Content that is invisible to LLMs risks becoming invisible in the next generation of search. Content built for them gets quoted. That is the whole case for starting GEO now, while most competitors have not.
We track exactly this at RankWit: which brands the AI engines cite, and why. If you want to measure your own visibility, start with us.
FAQ
What is the difference between LLM search and traditional search?
A traditional engine matches your query against an index and returns ranked links; you do the reading and choosing. An LLM search tool reads the sources for you and answers directly, often citing a handful of pages. The unit of visibility changes from a ranking position to a citation inside the answer.
Will Google disappear?
Almost certainly not, at least on any horizon worth planning for. Google holds the large majority of queries and is adding generative features of its own, such as AI Overviews. The realistic outcome is coexistence: a growing share of questions answered conversationally, with traditional results serving the rest. Both channels reward the same fundamentals.
How is GEO different from SEO?
SEO optimises pages to rank in a list of results. GEO optimises content to be understood, trusted and cited by generative models. The overlap is large: clean structure, authority and useful content serve both. The difference is the target: a position on a page versus a mention inside a generated answer.
How can a business measure its visibility in LLM answers?
Ask the major models the questions your customers ask, in several variations, and record whether your brand appears, how it is described and which sources get cited. Repeat the exercise regularly, because answers shift as models update. Monitoring tools, RankWit among them, automate this tracking across engines and over time.