ChatGPT Still Leads. The Competition Gives Me More to Think About.

Three anonymous chatbot windows side by side on a desktop monitor.
AI-generated illustration of comparing chatbot tools. The interfaces are conceptual, not real product screenshots.


ChatGPT is still the biggest name in consumer AI, but I find the competition more interesting than the crown. A market with several useful alternatives gives me more room to choose a tool around my work, rather than building everything around whichever service became familiar first.

Similarweb’s latest chatbot traffic update suggests that the web audience is spreading out. It does not, by itself, tell me which subscription deserves a place in my budget.

What the September numbers show

In Similarweb’s September 2026 update, ChatGPT accounts for 56.4% of the measured chatbot website traffic. Gemini follows with 25.0%, and Claude with 9.6%.

ServiceOctober 2025September 2026
ChatGPT71.8%56.4%
Gemini13.6%25.0%
Claude2.3%9.6%
DeepSeek4.1%3.4%
Grok2.8%2.3%
Copilot2.0%1.5%
Perplexity2.5%0.9%

Share of website traffic in Similarweb’s published comparison; rounded figures. These listed services do not exhaust the denominator.

ChatGPT’s share is 15.4 percentage points below October 2025, although it recovered from 53.7% in April 2026. Claude’s September share exceeds DeepSeek, Grok, Copilot and Perplexity combined. The story is competition over time, rather than a one-way collapse.

Visits are not people, and popularity is not performance

I keep three distinctions in mind when looking at this table. A visit is not a unique person. A website is not the whole product. And traffic share is not an accuracy score.

A person who opens a chatbot repeatedly can generate several visits. Someone using a mobile app or an API can be doing substantial work without appearing in this particular website comparison. AI features embedded inside other products add another reason not to call this the share of all AI usage.

Similarweb’s methodology combines several kinds of digital signals and modelling. These are estimates, not a shared audited counter supplied by every AI company. Even the denominator matters: a comparison restricted to seven services can produce different percentages from a wider tracked group.

A declining share also does not necessarily mean fewer visits. If the surrounding market expands faster, a service can grow while taking a smaller slice. I need absolute figures for matching dates before claiming that its audience has shrunk.

My subscription decision starts with actual tasks

For research and blog projects, the useful comparison is much smaller than the AI market. Can the tool find a source I can check? Does it follow a revision without undoing my latest edits? Can I get my work out in a convenient format? How much correction does the result need?

A fair personal comparison could use the same three tasks across services: research a topic from primary sources, revise a piece of writing against specific instructions, and complete a small technical job. I would record the time to a usable result, including the checking and cleanup. That is a proposed way to evaluate them, not a benchmark I have run.

Price belongs in that comparison, but the cheapest headline number is only the start. My look at DeepSeek’s completed-task costs explains why retries, tools and usage patterns can change the calculation.

More choice should mean a better fit

I like seeing more serious competition. It gives readers a reason to revisit assumptions and gives providers a reason to keep improving. It does not give me a reason to subscribe to every service that gains a few points.

For my budget, one tool that reliably fits the work can be worth more than three overlapping subscriptions. This traffic update is an invitation to compare, not an instruction to switch.

Comments

Popular posts from this blog

Obsidian’s New Search: Less Friction for Notes in Two Languages

OPPO Find X9 Ultra vs Find X10 Pro Max: Is the Older Ultra Still Worth Buying?

OPPO Find X10 Technical Preview: Cameras, Design and Missing Specs