AI's Next Battle Isn't Chatbot Share. It's Who Owns the Action

01. ChatGPT is still growing

Comscore counted ChatGPT reaching roughly 69% of unique visitors in the AI assistant category as of June 2026.

02. Google's weapon isn't just the model. It's distribution

The Gemini app has passed 1 billion monthly active users, and Google AI Overviews now reaches more than 2.5 billion users a month.

03. The next contest is about action, not just answers

As AI starts booking reservations, making purchases, drafting documents, writing code and handling customer support directly, distribution, permissions, payments, reliability and cost structure start to matter more than raw model quality.

Market share isn't the point. The market is going multi-platform

Framing the AI market as "ChatGPT versus Gemini" only shows half the picture. What Comscore's public data actually shows isn't that ChatGPT usage is shrinking. It's that rival services are growing fast at the same time.

In March 2026, ChatGPT's desktop conversation volume was up 55% year over year. Over the same period, Claude's desktop conversations jumped 1,858% from October 2025, reaching 22 million. Users aren't abandoning one AI for another. They're running several in parallel, depending on the task.

69% Share of unique visitors in the AI assistant category reached by ChatGPT, per Comscore, June 2026

1 billion+ Monthly active users of the Gemini app, disclosed by Google in August 2026

2.5 billion+ Monthly active users of Google AI Overviews

These figures come from different methodologies and time frames, so they can't be compared directly as market share. But the direction is clear. AI isn't heading toward a winner-take-most market. It's heading toward a market where one person uses several different AI tools depending on the purpose.

That means the metric to watch going forward isn't prompt share alone. It's weekly and monthly active users, time spent, paid conversion rates, enterprise seat counts, API revenue, agent execution volume, and actual task-completion rates, taken together.

Distribution beats the model

Once model performance flattens out above a certain threshold, the center of competition shifts to distribution. Google has Search, Android, Chrome, Gmail, YouTube, Maps, Docs and Drive. Users encounter AI inside Google's existing products without ever seeking out the Gemini app on purpose.

Microsoft looks similar. It has Windows, Microsoft 365, Teams, Outlook, GitHub, Azure, and an enterprise security and management stack built around them.

According to figures Microsoft disclosed for its fiscal 2026 fourth quarter, GitHub Copilot users reached 50 million. Copilot revenue grew more than 60% quarter over quarter, and the number of enterprise customers deploying Copilot to a majority of their information workers rose roughly 75% from the prior quarter. AI is turning from software you buy separately into a layer of intelligence added on top of software you already use.

AI looks less like a search substitute and more like a new internet interface

The old path through the internet ran: type a query, pick a link, compare several pages, decide, act. AI compresses this into: ask, compare, get recommended, decide, act.

What matters isn't just that fewer people click through to websites. It's that more decisions are ending inside the AI itself. Google says AI Overviews now has more than 2.5 billion monthly active users, and AI Mode has passed 1 billion.

That changes what publishers and brands need to track. It's no longer enough to watch where you rank in search results. Brands need to track which questions get them mentioned by AI, whether they're cited as a source, whether they make it into the shortlist when compared against competitors, and whether that recommendation actually converts into a purchase, signup or inquiry.

SEO isn't disappearing. AEO and GEO are being layered on top of it.

Content value splits into three layers

Comscore itself has started separately measuring what content AI references, whether that leads to actual site visits, and how AI exposure affects subsequent consumer behavior.

In this environment, content that merely repackages other sites' material may lose economic value. Conversely, proprietary data, original research, quantitative comparisons, hands-on experience, first-person interviews and verified expertise could become more valuable.

Companies need to produce content whose credibility drops noticeably if you strip out the source, even when AI can summarize it.

Advertising moves from above the link into the answer itself

As AI becomes the gateway for search and recommendations, advertising follows. Comscore analyzed hotel-related ChatGPT prompts where a source link was identifiable, and found that the share showing sponsored ads rose from 6% in March 2026 to 14% in April and 24% in May.

OpenAI said in August 2026 that ChatGPT Ads had reached an annualized revenue run rate of $1 billion. Google is also running Search and Shopping ads inside AI Overviews.

In the old model, brands competed to buy the keyword "running shoes." In the AI era, they need to get into the shortlist for a question like "recommend running shoes under $150 with a wide toe box for someone who runs 5K three times a week and wants less knee strain."

Structured information like price, inventory, shipping, return policy, reviews and location is becoming more important. Spending heavily on ads alone isn't enough. Companies need to become the kind of business AI wants to recommend.

For Big Tech investors, the question is who owns the action layer

The most important distinction here is that having a lot of AI users and making good money from AI are two different problems. Even if AI usage explodes, if inference costs, GPUs, data centers and power costs grow faster, that becomes a drag on valuation.

In the end, investors need to watch user growth, usage growth, monetization and per-unit inference cost together, not any single one of them in isolation.

Insight Times Editorial Desk