What OpenAI’s Research Reveals About The Future Of AI Search

The launch of ChatGPT in 2022 didn’t so much cause a shift in the search landscape as trigger a series of seismic events. And, like seismologists, the SEO industry needs data if it’s to predict future tremors and aftershocks – let alone prepare itself for what the landscape might reshape itself into once the ground has finally settled.

So, when OpenAI released a 65-page research paper on Sept. 15, 2025, titled “How People Use ChatGPT,” some of us were understandably excited to finally have some authoritative usage data from inside a major large language model (LLM).

Two key findings leap out:

We’re closer to mass adoption of AI than most probably realize.
How users interact with ChatGPT has fundamentally shifted in the past year.

For SEOs, this isn’t just another adoption study: It’s strategic intelligence about where AI search is heading.

Mass Adoption Is Closer Than You Think

How close is ChatGPT to the tipping point where it will accelerate into mass adoption?

Developed by sociologist Everett Rogers, the diffusion of innovation theory provides us with a useful framework to explain how new technologies spread through society in predictable stages. First, there are the innovators, accounting for 2.5% of the market. Then, the early adopters come along (13.5%), to be followed by the early majority (34%). At this point, ~50% of the potential market has adopted the technology. Anyone jumping on board after this point can safely be described as either the late majority (34%), or laggards (16%).

The tipping point happens at around 20%, when the new technology is no longer confined to innovators or early adopters but is gradually taken up by the early majority. It’s at this point that mainstream adoption accelerates rapidly.

Now, let’s apply this to ChatGPT’s data.

Since launching in late 2022, ChatGPT’s growth has been staggering. The new report reveals that, in the five-month period from February to July 2025, ChatGPT grew from 400 million to 700 million weekly active users (WAU), sending 18 billion messages per week. That represents an average compound growth of roughly 11-12% month-over-month.

700 million WAU is equivalent to around 10% of the global adult population; impressive, but not quite mass adoption. Yet.

(Side note: Back in April, Sam Altman gave a figure of ~800 million weekly active users when speaking at TED 2025. To avoid confusion, we’ll stick with the official figure of 700 million WAU quoted in OpenAI’s report.)

It’s estimated there were approximately 5.65 billion internet users globally at the start of July 2025. This is the total addressable market (TAM) available to ChatGPT.

20% of 5.65 billion = 1.13 billion WAU. That’s the tipping point.

Even if the growth rate slows to a more conservative 5-6% per month, ChatGPT would already have reached at least 770 million WAU as I write this. At that rate of growth, ChatGPT will cross the mass adoption threshold between December 2025 and August 2026, with April 2026 as the most likely midpoint.

Of course, if the rate of growth remains closer to 11-12%, we can expect to tip over into mass adoption even earlier.

For SEOs, this timeline matters. We don’t have years to prepare for mass AI search adoption. We have months.

The window is rapidly closing for any brands not wanting to be left behind.

The Behavioral Revolution Hiding In Plain Sight

Buried within OpenAI’s usage data is perhaps the most significant finding for search marketers: a fundamental shift in how people are using AI tools.

In June 2024, non-work messages accounted for 53% of all ChatGPT interactions. By June 2025, this figure had climbed to 73%. This is a clear signal that ChatGPT is moving from workplace tool to everyday utility.

Things get even more interesting when we look at the intent behind those queries. OpenAI categorizes user interactions into three types:

Asking (seeking information and guidance).
Doing (generating content or completing tasks).
Expressing (sharing thoughts or feelings with no clear intent).

The data reveals that “Asking” now makes up 51.6% of all interactions, compared to 34.6% for “Doing” and 13.8% for “Expressing.”

Let’s be clear: What ChatGPT categorizes as “Asking” is pretty much synonymous with what we think of as AI search. These are the queries that were once the exclusive domain of search engines.

Users are also increasingly satisfied with the quality of responses to “Asking” queries, rating interactions as either Good or Bad at a ratio of 4.45:1. For “Doing” interactions, the ratio of Good to Bad drops to 2.76.

The trend becomes even clearer when we break down interactions by topic. Three topics account for just under 78% of all messages.

Practical Guidance (29%).
Seeking Information (24%).
Writing (24%).

These figures are even more noteworthy when you consider that, in July 2024, “Writing” was easily the most common topic (36%), dropping 12 percentiles in just one year.

And while “Practical Guidance” has remained steady at 29%, “Seeking Information” has shot up 10 percentiles from 14%. What a difference a year makes.

And while “Writing” still accounts for 42% of all work-related messages, the nature of these requests has shifted. Instead of generating content from scratch, two-thirds of writing requests now focus on editing, translating, or summarizing text supplied by the user.

Whichever way you slice it, AI search is now the primary use case for ChatGPT, not content generation. But where does that leave traditional search?

The AI Wars: Battling For The Future Of Search

ChatGPT may be reshaping the landscape, but Google hasn’t been sitting idle.

Currently rolling out to 180 countries worldwide, AI Mode is Google’s biggest response yet to ChatGPT’s encroachment on its territory. Setting the scene for what is likely to become a competitive struggle between Google and OpenAI to define and dominate AI search.

ChatGPT has an advantage in having largely established the conversational search behaviors we’re now seeing. Instead of piecing together information by clicking back and forth on links in the SERPs, ChatGPT provides users with complete answers in a fraction of the time.

Meanwhile, Google’s advantage is that AI Mode grounds responses against a highly sophisticated search infrastructure, drawing on decades of web indexing expertise, contextual authority, and myriad other signals.

The stakes are high. If Google doesn’t transition aggressively enough to seize ground in AI search and protect its overall search dominance, it risks becoming the next Ask Jeeves.

That’s why I wouldn’t be surprised at all to see AI Mode become their primary search interface sooner rather than later.

Naturally, this would be a massive disruption to the traditional Google Ads model. Google’s recent launch of a new payment protocol suggests it is already hedging against the risk of falling ad revenue from traditional search.

With everything still so fluid, it’s virtually impossible to predict what the search landscape will eventually look like once the dust has settled and new business models have emerged.

Whichever platform ultimately dominates, it’s all but certain that AI search will be the victor.

Instead of focusing on what we don’t know and waiting for answers, brands can use what they do know about AI search to seize a strategic advantage.

Rethinking Traffic Value

With most websites only seeing ~1-2% of traffic coming from LLMs like ChatGPT, it would be tempting to dismiss AI search as insignificant, a distraction – at least for now.

But with ChatGPT about to hit mass adoption in months, this picture could change very rapidly.

Plus, AI search isn’t primarily about clicks. Users will often get the information they need from AI search without clicking on a single link. AI search is about influence, awareness, and decision support.

However, analyzing traffic from AI sources does reveal some interesting patterns.

Our own research indicates that, in some industries at least, LLM-referred visitors convert at a higher rate than traditional search traffic.

This makes sense. If someone has already engaged with your brand through one or more AI interactions and still chooses to visit your site, they’re doing so with more intent than someone clicking through in search of basic information. Perhaps they’re highly engaged in the topic and want to go deeper. Or perhaps the AI responses have answered their product queries, and they’re now ready to buy.

Even if it results in fewer clicks, this indirect form of brand exposure could become increasingly valuable as AI adoption reaches mass market levels.

If 1-2% of traffic currently comes from AI sources at 10% market adoption, what happens when we reach 20% or 30% adoption? AI-mediated traffic – with its higher conversion rate – could easily grow to 5-10% of total website visits within two years.

For many businesses, that’s enough to warrant strategic attention now.

Strategic Implications For Search Marketers

Traditional keyword optimization hasn’t been cutting it for a while. And things aren’t about to get any simpler for anyone hoping to capture the intent-driven queries dominating AI interactions.

Digital marketers and SEOs need to think beyond algorithms, considering aspects that aren’t always so easily captured in a spreadsheet, such as user goals and decision-making processes.

This doesn’t mean we should abandon those SEO fundamentals essential to healthy, scalable growth. And technical SEO remains as important as ever, including proper site structure, fast loading times, and crawlable content.

However, when it comes to the content itself, the emphasis needs to shift toward providing greater depth, expertise, and user value. AI systems are far more likely to reward original, comprehensive, and authoritative information over keyword-optimized but otherwise thin content.

In short, your content needs to be built for “Asking.”

Focus on the underlying needs of the user: information gathering, interpretation, or decision support. And plan your content around “answer objects.” These are modular content components designed to be reused and repurposed by AI when generating responses to specific queries.

Instead of traditional articles targeting specific keywords, build decision frameworks that include goals, options, criteria, trade-offs, and guardrails. Each of these components can provide useful material for AI to cite in responses, whichever AI system that might be.

Preparing for AI search isn’t about looking for ways to game an algorithm. It’s about creating genuinely useful content that helps users make decisions.

For many brands, this will mean moving away from individually optimized pages to entire content ecosystems.

The Way Ahead

OpenAI’s research gives us the most authoritative picture yet of AI search adoption and user behavior. The data shows that we’re approaching a tipping point where AI-mediated search will become mainstream, while user behavior has shifted dramatically toward information seeking over content generation.

Meanwhile, the competitive landscape remains extremely fluid.

The message is clear, for now at least: Build for “Asking.”

Start planning strategies around intent-driven, decision-supporting content now, while the landscape is still evolving.

The businesses that can establish their authority in AI responses now will be in the best position when AI search does reach mass adoption – regardless of which platforms ultimately dominate.

More Resources:

Featured Image: Collagery/Shutterstock