Generative systems like ChatGPT, Gemini, Claude, and Perplexity are quietly taking over the early parts of discovery – the “what should I know?” stage that once sent millions of people to your website.

Visibility now isn’t just about who ranks. It’s about who gets referenced inside the models that guide those decisions.

The metrics we’ve lived by – impressions, sessions, CTR – still matter, but they no longer tell the full story.

Mentions, citations, and structured visibility signals are becoming the new levers of trust and the path to revenue.

This article pulls together data from Siege Media’s two-year content performance study, Grow and Convert’s conversion findings, Seer Interactive’s AI Overview research, and what we’re seeing firsthand inside generative platforms.

Together, they offer a clearer view of where visibility, engagement, and buying intent are actually moving as AI takes over more of the user journey – and has its eye on even more.

Content type popularity and engagement trends

In a robust study, the folks at Siege Media analyzed two years of performance across various industry blogs, covering more than 7.2 million sessions. It’s an impressive dataset, and kudos to them for sharing it publicly.

A disclaimer worth noting: the data focuses on blog content, so these trends may not map directly to other formats such as videos, documentation, or landing pages.

With that in mind, here’s a run-through of what they surfaced.

TL;DR of the Siege Media study

Pricing and cost content saw the strongest growth over the past two years, while top-of-funnel guides and “how-to” posts declined sharply.

They suggest that pricing pages gained ground at the expense of TOFU content. I interpret this differently.

Pricing content didn’t simply replace TOFU because the relationship isn’t zero-sum.

As user patterns evolve, buyers increasingly start with generative research, then move to high-intent queries like pricing or comparisons as they get closer to a decision.

That distinction – correlation vs. causation – matters a lot in understanding what’s really changing.

The data shows major growth in pricing pages, calculators, and comparison content.

Meanwhile, guides and tutorials – the backbone of legacy SEO – took a sharp hit.

Keep that drop in mind. We’ll circle back to it later.

Interestingly, every major content category saw an increase in engagement. That makes sense.

As users complete more of their research inside generative engines, they reach your site later in the journey or for additional details, when they’re already motivated and ready to act.

If you’re a data-driven SEO, this might sound like a green light to focus exclusively on bottom-of-funnel content.

Why bother with top-of-funnel “traffic” that doesn’t convert?

Leave that for the suckers chasing GEO visibility metrics for vanity, right?

But of course, this is SEO, so I have to say it …

Did you expect me to say, “It depends?”

Here’s a question instead: when that high-intent user typed the query that surfaced a case study, pricing page, or comparison page, where did they first learn the brand existed?

Dig deeper: AI agents in SEO: What you need to know

Don’t forget the TOFU!

I can’t believe I’m saying this, but you’ll have to keep making TOFU content.

You might need to make even more of it.

Let’s think about legacy SEO.

If we look back – waaaaay back – to 2023 and a study from Grow and Convert, we see that while there is far more TOFU traffic…

…it converts far worse.

**Note: **They only looked at one client, so take it with a grain of salt. However, the direction still aligns with other studies and our instincts.

This pattern also shows up across channels like PPC, which is why TOFU keywords are generally cheaper than BOFU.

The conversion rate is higher at the bottom of the funnel.

Now we’re seeing this shift carry over to generative engines, except that generative engines cover the TOFU journey almost entirely.

Rather than clicking through a series of low-conversion content pieces as they move through the funnel, users stay inside the generative experience through TOFU and often MOFU, then click through or shift to another channel (search or direct) only when it’s time to convert.

For example, when I asked ChatGPT to help me plan a trip to the Outer Banks:

After a dozen back-and-forths planning a trip and deciding what to eat, I wanted to find out where to stay.

That journey took me through many steps and gave me multiple chances to encounter different brands and filtering or refinement options.

I eventually landed on my BOFU prompt, “Some specific companies would be great.”

From there, I might click the links or search for the company names on Google.

What matters about this journey – apart from the fact that my final query would be practically useless as insight in something like Search Console – is that throughout the TOFU and MOFU stages, I was seeing citations and encountering brands I would rely on later.

Once I switched into conversion mode, I wanted help making decisions. That’s where I’m likely to click through to a few companies to find a rental.

So, when we read statistics like Pew’s finding that AI Overviews reduce CTR by upwards of 50%, and then consider what happens when AI Mode hits the browser, it’s easy to worry about where your traffic goes. Add to that ChatGPT’s 700 million weekly active users (and growing):

And according to their research on how users engage with it:

We can see a clear TOFU hit and very little BOFU usage.

So, on top of the ~50% hit you may be taking from AI Overviews, 700+ million people are going to ChatGPT and other generative platforms for their top-of-funnel needs.

I did exactly that above with my trip planning to the OBX.

Dig deeper: 5 B2B content types AI search engines love

Get the newsletter search marketers rely on.

---

But wait!

The good news is that while that vacation rental company or blue widget manufacturer might not see me on their site when I’m figuring out what to do – or what a blue widget even is – I’m still going to take the same number of holidays and buy the same number of products I would have without AI Overviews or ChatGPT, Claude, Perplexity, etc.

Unless you’re a publisher or make money off impressions, you’ll still have the same amount of money to be made.

It just might take fewer website visits to do it.

More about TOFU

Traffic at the bottom of the funnel is holding steady for now (more on that below), but the top of the funnel is being replaced quickly by generative conversations rather than visits.

The question is whether being included in those conversations affects your CTR further down the funnel.

The folks at Seer Interactive found that organic clicks rose from 0.6% to 1.08% when a site was cited in AI Overviews.

And while the traffic was far lower, ChatGPT had a conversion rate of 16% compared with Google organic’s 1.8%.

If we look at the conversion rate for organic traffic at the bottom of the funnel – which we saw above – it was 4.78%.

Users who engage with generative engines clearly get further into their decision-making than users who reach BOFU queries through organic search.

But why?

While I can’t be certain, I agree with Seer’s conclusion that AI-driven users are pre-sold during the TOFU stage.

They’ve already encountered your brand and trust the system to interpret their needs. When it’s time to convert, they’re almost ready with their credit card.

Why bottom-funnel stability won’t last much longer

Above, I noted that “traffic at the bottom of the funnel is holding steady for now.”

It’s only fair to warn you that through 2026 and 2027, we’ll likely see this erode.

The same number of people will still travel and still buy blue widgets.

They just won’t book or buy them themselves. And at best, attribution will be even worse than it is today.

I spoke at SMX Advanced last spring about the rise of AI agents.

I won’t get into all the gory details here, but the Cliff Notes are this:

Agents are AI systems with some autonomy that complete tasks humans otherwise would.

They’re rising quickly – it’s the dominant topic for those of us working in AI – and that growth isn’t slowing anytime soon. You need to be ready.

A few concepts to familiarize yourself with, if you want to understand what’s coming, are:

• AP2 (Agent Payments Protocol): A standard that allows agents to securely execute payments on your behalf. Think of it as a digital letter of credit that ensures the agent can only buy the specific “blue widget” you approved within the price limit you set. Before you say, “But I’d never send a machine to do a human’s job,” let me tell you, you will. And if you somehow prove me wrong individually out of spite, your customers will.
• Gemini Computer Use Model API: A model with reasoning and image understanding that can navigate and engage with user interfaces like websites. While many agentic systems access data via APIs, this model (OpenAI has one too, as do others) lets the agent interact with visual interfaces to access information it normally couldn’t – navigating filters, logins, and more if given the power.
• MCP (Model Context Protocol): An emerging standard acting as a universal USB port for AI apps. It lets agents safely connect to your internal data (like checking your calendar or reading your emails) to make purchasing decisions with full context and to work interactively with other agents. Hat tip to Ahrefs for building an awesome MCP server.

***Dig deeper: ***How Model Context Protocol is shaping the future of AI and search marketing

Why do these protocols matter to a content strategist?

Because once AP2 and Computer Use hit critical mass, the click – that sacred metric we’ve optimized for two decades – changes function.

It stops being a navigation step for a human exploring a website and becomes a transactional step for a machine executing a task.

If an agent uses Computer Use to navigate your pricing page and AP2 to pay for the subscription, the human user never sees your bottom-of-the-funnel content.

So in that world, who – or rather, what – are you optimizing for?

This brings us back to the Siege Media data.

Right now, pricing pages and calculators are winning because humans are using AI to research (TOFU and MOFU) and then manually visiting sites to convert (BOFU).

But as agents take over execution, that manual visit disappears. The “traffic” to your pricing page may be bots verifying costs, not humans persuaded by your copy.

The 2026 strategy

This reality pushes value back up the funnel.

If the agent handles the purchase, the human decision – the “moment of truth” – happens entirely inside the chat interface or agentic system during the research phase.

In this world, you don’t win by having the flashiest pricing page.

You win by being the brand the LLM recommends when the user asks, “Who should I trust?”

Your strategy for 2026 requires a two-pronged approach:

• For the agent (the execution): Ensure your BOFU content is technically flawless. Use clean schema, accessible APIs, and clear data structures so that when an agent arrives via MCP or Computer Use to execute a transaction, it encounters no friction.
• For the human (the selection): Double down on TOFU. Focus on mentions and citations. You need to be the entity referenced in the generative answer so that users – and agents – trust you.

As we move toward 2026 and then 2027 (it’ll be here sooner than you think), the “click” will become a commodity more often handled by machines.

The mention, however, remains the domain of human trust. And in my opinion, that’s where your next battle for visibility will be fought.

Time to start – or hopefully keep – making the TOFU.

Contributing authors are invited to create content for Search Engine Land and are chosen for their expertise and contribution to the search community. Our contributors work under the oversight of the editorial staff and contributions are checked for quality and relevance to our readers. Search Engine Land is owned by Semrush. Contributor was not asked to make any direct or indirect mentions of Semrush. The opinions they express are their own.