AI traffic is up 527%. SEO is being rewritten.

For the past year, we’ve talked about how AI might change search.

That moment is over.

This is no longer a “what if” conversation. We are seeing a measurable shift in web traffic movement.

At Previsible, we analyzed LLM-driven traffic across 19 GA4 properties and found something undeniable: AI platforms like ChatGPT, Perplexity, Claude, Gemini, and Copilot are already influencing how users find and engage with websites.

Not in theory. In actual traffic.

• In just five months, total AI-referred sessions jumped from 17,076 to 107,100.
  • That’s a 527% increase between January and May 2025.
• Some SaaS sites are now seeing over 1% of all sessions coming from LLMs.
• Traffic from ChatGPT, Claude, and others is doubling and tripling across verticals like Legal, Health, and Finance.

If you work in SEO, content, or growth strategy, this moment will feel familiar. Like when mobile-first flipped ranking factors overnight. Or when social transformed from brand garnish into a legitimate acquisition engine.

Every time the rules changed, early adopters won. This time is no different, except it’s moving faster.

So the question isn’t if AI is changing your traffic mix. It’s how much it already has, without you realizing it.

TL;DR: What you need to know about AI search

• AI discovery is up 527%: Comparing the first 5 months of 2025 with the same time frame in 2024 we see how total sessions from LLMs (like ChatGPT, Perplexity, and, Gemini) surged from 17,076 to 107,100 across 19 GA4 properties.
•  LLMs are already part of the user journey: Some sites, especially in SaaS, are seeing over 1% of all traffic initiated by AI results. Primarily to the bottom of funnel users and targeted prospects. 
• High-consultive industries are leading:  Legal, Finance, SMB, Insurance, and Health make up 55% of all LLM-driven sessions, showing that users turn to AI for complex, contextual questions.
• ChatGPT leads, but the field is widening: ChatGPT still dominates, but Perplexity, Copilot, and Gemini are gaining real traction. 
• SEO is splitting and speeding up:  It’s no longer just about ranking in Google. You now need to earn visibility in AI assistants, summaries, and conversational UIs, and they prefer content that’s structured, clear, and genuinely helpful.

AI discovery is up 527% – and it isn’t waiting for you to rank

AI is reshaping web traffic at warp speed.

When we compared January-May 2025 to the same period in 2024, we saw total AI-sourced sessions across 19 GA4 properties jump from 17,076 to 107,100.

That’s a 527% year-over-year increase.

In one standout example, ChatGPT went from just 600 visits/month in early 2024 to over 22,000/month by May 2025.

And when you zoom in by industry, the growth in share is just as dramatic:

• Legal: 0.37% → 0.86% of sessions from LLMs
• Health: 0.17% → 0.56%
• Finance and SaaS are showing similar trajectories, in some cases, exceeding 1% of total traffic

LLMs are becoming a legitimate discovery channel, and they’re doing it fast.

Why it matters:

Most SEO strategies are still stuck in the old timeline:

Optimize → Wait → Crawl → Rank → Convert

That playbook was built for Google’s crawling and indexing cycle – a system that rewards patience, backlinks, and slow iteration.

But LLMs don’t care about that process.

They don’t crawl the same way. They don’t rank in the same order. They don’t wait for your canonical tag to propagate.

They surface content immediately if it’s useful.

The only thing that matters is whether your content helps answer the user’s question in a way the model trusts.

No indexing delay. No competition for blue links. No sandbox.

Just: Is this helpful right now?

That rewires everything.

Content doesn’t need to appear at the top of Google’s SERPs to be found. It needs to be clear, structured, and cited by the model – whether in a blog, a help doc, a case study, or a knowledge base.

And it means the old mindset — publish, wait, and hope Google figures it out — is now dangerously outdated.

We’ve entered the “instant surfacing era” of SEO, where content can be discovered before it even ranks.

If your SEO strategy doesn’t account for that, you’re already behind.

Where LLM traffic is actually going: The real breakdown

Not predictions. Not vibes. Actual traffic.

In the 2025 Previsible AI Data Study, we analyzed LLM-driven sessions across 19 GA4 properties to understand where platforms like ChatGPT, Claude, Gemini, Copilot, and Perplexity are already influencing real user behavior.

Here’s what we found:

• Legal topped the chart, with 0.28% of total traffic from LLMs
• Finance followed at 0.24%, showing strong traction in regulated markets
• Health came in at 0.15%, with a mix of ChatGPT, Gemini, and Perplexity sources
• SaaS showed breakout performance but only in some domains, with a selected few getting 1%+ of total sessions from LLMs

But the most important finding?

Legal, Finance, Health, SMB, and Insurance account for 55% of all LLM-sourced sessions across the dataset.

Why these five?

Because people aren’t using LLMs like search engines.

They’re asking contextual, trust-heavy, consultative questions. The kind they’d normally ask a real expert:

• “What should I ask a lawyer before signing this contract?”
• “Is this medication safe with XYZ conditions, XYZ personal information, and XYZ symptoms?”
• “How do I structure payroll as a small business owner that owns a flower shop with 5 employees, 2 part-time and 3 full-time?”

These are high-context moments, and that’s where LLMs are starting to win.

So if your brand plays in a space where trust, clarity, or expertise matters, and your content isn’t optimized for AI, you’re likely missing the exact kinds of requests these models are built to answer.

Model-level insight: Who’s actually driving the traffic?

It’s not just about how much AI traffic you’re getting, but also about who’s sending it. 

Across nearly every vertical, ChatGPT is the dominant contributor, consistently driving 40–60%+ of all LLM traffic.

But this isn’t a single-player story. Other models are gaining share, especially in specific sectors:

• Perplexity is surprisingly strong, contributing over 0.073% of Finance traffic, 0.041% in SMB, and 0.041% in Legal.
• Copilot makes up a meaningful chunk of Legal (0.076%) and Finance (0.036%) sessions. Second only to ChatGPT in both.
• Gemini is emerging in Insurance (0.0075%) and SMB (0.035%).
• Claude is still marginal (<0.001% in most industries), but present across the board.

Bottom line: While ChatGPT leads, LLM discovery is becoming a multi-model landscape, and performance is beginning to vary by vertical and use case.

This has two implications:

1. You can’t just optimize for one model. Visibility across platforms will matter more over time.
2. Different models favor different formats, sources, and structures. Understanding how each one pulls and presents content is your next strategic edge.

How to adapt to the LLM traffic surge starting now

If you’re still treating LLMs like a 2026 conversation, you’re already behind.

The shift is no longer theoretical; it’s happening in your analytics right now. And the teams that move early will stay visible and build a lasting competitive edge.

Here’s how to respond:

1. Start tracking LLM-driven sessions, even if it’s imperfect

You can’t manage what you don’t measure.

Set UTM parameters for AI platforms. Monitor for unexplained spikes in direct traffic. Annotate content that gets surfaced in ChatGPT or Perplexity.

Look for surges in branded search that coincide with AI exposure. Start tracking mentions, not just clicks.

Attribution won’t be perfect – but waiting for standardized reporting is how you miss the wave.

2. Structure your content for AI interfaces, not just human readers

LLMs favor content that’s clean, clear, and scannable. Think bullet points, tight intros, FAQ sections, and strong summaries.

If featured snippets were SEO 2.0, this is 3.0. Answers need to perform inside a model’s response, not just on a results page.

3. Shift your mindset: from ranking to being selected

It’s not about being in position #1 – it’s about being the answer a model chooses to surface.

That means relevance, clarity, and trust signals matter more than ever.

Audit the content already being cited or linked by AI platforms and build a strategy for becoming the go-to source in your space.

If you don’t, your competitor will.

4. Make your content AI-ready across the entire funnel

This isn’t just about blogs.

Product pages, help docs, onboarding flows – every touchpoint is now eligible to be surfaced in an AI conversation.

You need cross-functional alignment between SEO, content, UX, and product teams to ensure your entire site is conversation-ready.

SEO isn’t dying – it’s evolving

SEO is splitting into two tracks: traditional search and LLM-driven discovery.

The second one is growing faster than anyone expected and it’s already rewriting how users find answers and how brands earn visibility.

Move now. Learn fast. Or get left behind.

Read more at Read More