How AI Search Is Changing Discovery (And Why the Next Few Months Matter for Marketers)

If you are planning SEO and content right now, you have probably felt the ground shift a little. Discovery no longer lives in one place. People still use Google, but they also get answers from AI Overviews and from assistants such as ChatGPT, Gemini and Grok.

I wrote this for marketers who want a clear read on what the data actually says - not another panic headline about the death of search. The classic blue-link journey still matters. It is just no longer the only useful journey, and plans that assume it is will miss a growing share of how people find answers.

At Rise By Digital we work with 150+ clients across the UK, many of them East Midlands businesses who still need classic Google visibility and a plan for AI answers. This piece is written for marketers first (business owners second). I will separate three things clearly:

1. What the current percentages actually say
2. What published forecasts say for about 12 months and a couple of years
3. My own view on why the next few months are the window to get ready - without inventing a fake near-term collapse figure

Start with web sessions, because they are easy to grasp.

Sitefire's analysis of Similarweb daily web sessions for October to December 2025 puts Google Search at about 2.80 billion daily web sessions (94.80%) and all AI search engines combined at about 277 million (5.20%). So traditional Google search still dominates the measured web-session pie. AI search is not "a rounding error" anymore either - it is already a meaningful slice.

Within that AI slice, the same Sitefire piece puts ChatGPT at about 194.73 million daily web sessions (leading, roughly 56% of AI search web sessions globally), AIM at 75 million, Gemini at 47.43 million, and Grok at 7.89 million globally.

Source: Sitefire analysis of Similarweb daily web sessions, October to December 2025.

That is one lens. Graphite uses another, and it is useful for marketers who care about share of discovery, not only "did someone open google.com".

Graphite's research dated 5 March 2026 reports that AI monthly sessions are now 56% the size of search worldwide (34% in the US). Narrower "Asking" prompts that look more like search intent are 28% the size of search worldwide (17% in the US).

On Graphite's adjusted method, Google's share of search-related traffic worldwide fell from 89% in 2023 to 71% in Q4 2025, while ChatGPT sits at 20% worldwide (12% in the US). Search engine usage itself was flat, not collapsing - and the combined search-plus-AI pie grew about 26% worldwide in 2025 versus 2024. ChatGPT is cited at 800 million+ weekly active users via OpenAI in that research. Search Engine Land summarised the Graphite study if you want a shorter read.

Source: Graphite research, 5 March 2026.

For traditional desktop search alone, the Datos / Semrush State of Search Q2 2026 still shows Google at roughly 92-94% in the US and at least about 94% in the EU/UK. Among people who use AI tools, ChatGPT sits in the mid-30%s in the US, Gemini around 19.5% (US, June in that report), and Claude around 11.5% (US). In other words: inside the old "search engine" category Google still looks almost total. Outside that category, assistants are taking real discovery share.

Inside Google, AI Overviews already change what a "search" feels like. Semrush's AI Overviews study found AIO share of keywords was volatile - peaking near 25% in July 2025 and at 15.69% in November 2025. Same-keyword zero-click moved slightly from 33.75% to 31.53% after AIO in that study - so the story is not a simple "every AIO kills every click".

Separately, Semrush's AI SEO statistics round-up cites Bain that around 60% of searches end with no click, and Pew that only 8% click a traditional result when an AI summary appears versus 15% without one. Attribute those Bain and Pew figures to Semrush's citation, not to a Rise By Digital primary study.

Here is the honest line. I have not seen a credible published forecast that says "traditional Google volume will fall X% in the next three to six months" as a measured fact. I will not invent one for the sake of a sharper headline.

What I do think, as someone who lives in client SEO, paid search and content every week, is that the next few months are when habits tip for a lot of marketers' audiences:

• More people ask ChatGPT, Gemini or Grok first for explainers, comparisons and "what should I do" questions
• More Google results include AI Overviews or push people toward AI Mode style experiences
• More of the journey ends without a classic blue-link visit, even when Google is still "in the mix"

Semrush's AI search SEO traffic study (21 July 2025) is careful about timelines, and it notes the shift can happen sooner if Google's default becomes AI Mode. That "sooner" clause is why I am urgent about the next half-year for marketing teams - not because I have a secret near-term percentage, but because defaults and habits move faster than annual planning cycles.

If you wait for a perfect industry chart that proves the tipping point after the fact, you will brief 2027 work with 2024 assumptions.

For a clearer published horizon, use Gartner.

Gartner's 19 February 2024 press release predicts that by 2026, traditional search engine volume will drop 25%, with search marketing losing share to AI chatbots and other virtual agents.

That does not mean Google disappears. It is a large shift in volume and in where marketing attention has to sit. If your reporting still treats organic Google sessions as the only proof that content worked, you will under-read demand that now shows up as assistant answers, branded direct visits after an AI recommendation, or shorter assisted paths.

Semrush's AI search SEO traffic study goes further for the topics they modelled. They find AI search visitors may surpass traditional search visitors for digital marketing and SEO topics by early 2028 - and sooner if AI Mode becomes the Google default.

In the same study, the average LLM visitor was worth about 4.4x a traditional organic visitor, and ChatGPT weekly active users grew 8x from October 2023 to April 2025 to 800 million+.

Sources: Adam Coley / Rise By Digital near-term outlook (qualitative); Gartner press release 19 February 2024; Semrush AI search SEO traffic study, 21 July 2025.

Read that as a planning signal for marketers in competitive knowledge categories, not as a promise that every local plumber's traffic flips on the same date. The direction is clear enough to change how you brief content, PR and measurement this year.

I still want strong technical SEO, clean site structure and pages that earn clicks. That has not gone away. What has changed is the job description around it.

1. Optimise for answers as well as rankings. We have written before about SEO vs AEO vs GEO - classic search engine optimisation, answer engine optimisation, and generative engine visibility. The practical version: clear entities, quotable facts, original experience, and pages that an assistant can trust enough to cite.

2. Stop treating AI copy as a finished product. Assistants reward specificity and lived detail. Thin rewrites get ignored in both Google and AI answers. That is why AI content still needs human creativity is not a soft branding line - it is a visibility line.

3. Measure discovery, not only last-click organic. Watch branded search, direct, assisted conversions, and qualitative "how did you hear about us" answers. AI referral reporting is still messy. Build a simple Live Dashboard view that includes the questions your sales team hears, not only sessions from google / organic.

4. Use AI tools inside the agency workflow without losing judgement. We use AI agents heavily ourselves - including Grok Bot for the busywork - so humans can spend time on strategy, creative and client judgement. The same rule applies to your content ops: speed is useful, unchecked publishing is not.

5. Brief campaigns for two surfaces. Paid and organic creative still need a strong landing page. They also need a crisp, factual story that can survive being summarised by Gemini or ChatGPT. Vague taglines travel badly into AI answers.

If you only do five things after this post, do these:

1. Map your top 20 money queries - where do they currently appear in Google, AI Overviews, ChatGPT, Gemini and Grok?
2. Tighten the pages that should be cited: original data, clear authorship, updated dates, plain-English answers near the top.
3. Add an "answer block" to key service pages (what it is, who it is for, proof, next step) without keyword stuffing.
4. Align PR and LinkedIn with the same facts you want assistants to repeat.
5. Agree how you will report AI-influenced demand to leadership, even if the analytics are imperfect.

That is enough to move before the Gartner 2026 horizon gets any closer.

The shift is already visible in the percentages. The forecasts for 2026 and early 2028 are published. My ask is simpler: treat the next few months as preparation time, not panic time - and build a marketing system that still works when the answer is an AI summary instead of a blue link.

If you want a second pair of eyes on how your current methods and processes line up with that picture - what still works in traditional search, where AI answers already matter for your topics, and what to tighten first - we can walk through it with you at Rise By Digital. Call 01509 357412 or get in touch via risebydigital.com.