For two years, choosing an AI assistant was barely a decision at all. ChatGPT was the default — the name people used the way they use "Google" for search. That era is over. Comscore's Q2 2026 AI Intelligence Report, published on September 22, shows ChatGPT's share of total AI prompt volume falling from 70 percent in January to 50 percent in June, while Google's Gemini nearly doubled from 17 to 30 percent and Anthropic's Claude surged from 2 to 11 percent. The one-assistant era has given way to a genuine three-way race.
All three shipped new flagship models in 2026: OpenAI released GPT-6 Astra on September 3, Anthropic's Claude Opus 5 arrived in July with Fable 5.1 above it from September 1, and Google's Gemini 3.1 Pro has led its lineup since February. Raw answer quality is no longer where they separate — all three handle everyday work well. What differs is what each is connected to, what it costs at the tier you would actually buy, and which one holds up on the specific thing you do most.
OpenAI's ChatGPT remains the category leader by a wide margin, and its strength is breadth. The free tier offers unlimited text chat on the GPT-5.6 Luna model, with web search, file uploads and image generation included (the image, voice and research features are capped). Paid tiers start with Go at $8 a month, rising to Plus at $20 — which adds the more capable GPT-5.6 Sol model, Agent mode and Deep Research — and on to Pro at $100 to $200 a month for heavy users and earliest access to GPT-6 Astra.
ChatGPT's differentiators are computer use, agentic workflows and tool breadth: it can browse, generate images with GPT Image 2.5, work with files and act across apps. The trade-offs: the free tier in the US shows ads, hitting usage caps drops you to a less capable model automatically, and its context window — about 1.05 million tokens on the flagship — is generous but no longer unique.
Claude: the careful writer's choice
Anthropic's Claude has been 2026's fastest-growing assistant in percentage terms. Web-traffic analysis published in September put Claude at roughly 952.6 million monthly visits — up from just 2 percent share in May 2025 to 9 percent in May 2026 — while Comscore's panel data showed its prompt-volume share rising from 2 to 11 percent in the first half of 2026. Claude's growth is the opposite of Gemini's distribution story: smaller in absolute traffic, but concentrated among users who deliberately sought it out.
Claude's reputation rests on long documents, coding and careful, well-structured answers. The free tier offers the Sonnet 5 and Haiku models with a 200,000-token context window. The catch is pricing structure: Claude has no budget paid tier at all — you go from free straight to Pro at $20 a month ($17 on annual billing), which adds Opus 5 and limited Fable 5.1 access. There is also no image generation; Claude is a text-and-reasoning specialist. For writers, researchers and developers who value precision over breadth, that focus is exactly the appeal.
Gemini: the one that's already everywhere
Google's Gemini grew from 9 to 28 percent of AI chatbot web traffic between May 2025 and May 2026 — roughly 450 percent year-over-year visit growth — largely on distribution. It is built into Google Search and Android by default, reaching people who never went looking for an AI chatbot in the first place. The flagship Gemini 3.1 Pro has led the lineup since February, with the lighter Gemini 3.6 Flash handling everyday queries on the free tier.
Gemini's strengths are search grounding, Google Workspace integration and multimodal input. Pricing is the most aggressive of the three: AI Plus at $4.99 a month doubles free-tier limits, and AI Pro at $19.99 a month unlocks extended 3.1 Pro access with a 1-million-token context window and Deep Research. The quirks: Gemini gates context by plan rather than model — just 32K tokens with no subscription — and its extended-reasoning Deep Think mode is reserved entirely for the $99.99-plus AI Ultra tier, while Anthropic and OpenAI put their best reasoning modes on their standard $20 plans.
Head to head
Plans and pricing
At the standard $20 tier, all three converge: ChatGPT Plus ($20), Claude Pro ($20, or $17 annually) and Gemini AI Pro ($19.99). The differences sit below and above that line. Only ChatGPT and Gemini offer sub-$10 entry tiers — Go at $8 and AI Plus at $4.99 — while Claude jumps from free to $20 with nothing between. At the top end, all three vendors' power tiers land in the same $100–$200 band; the real gap is what the cheaper mid-tier plans include.
What each does best
Published comparisons as of September 2026 converge on a clear division of labour. ChatGPT leads on computer use, agents and tool breadth — the most things an assistant can do. Gemini leads on search grounding, Workspace integration and multimodal input — the assistant most woven into daily digital life. Claude leads on long documents, coding and careful answers — the assistant most trusted with hard thinking. Image generation belongs to ChatGPT (GPT Image 2.5) and Gemini (the Nano Banana family); Claude offers none.
The market momentum story
As Smriti Sharma, Comscore's Chief Analytics and Technology Officer, put it in the September report: "As competition among AI assistants intensifies and prompt volume shifts across platforms, brands and publishers can no longer assume that a single AI experience defines the market." The data backs her up — and the growth is increasingly mobile, with AI assistant adoption up 25 percent on mobile versus 12 percent on desktop since the start of 2025.
Who should choose which?
Choose ChatGPT if: you want the broadest single toolbox — image generation, voice, agents and research in one subscription — or you need the cheapest paid on-ramp at $8.
Choose Claude if: your work is writing, analysis or code; you handle long documents; or you want the most careful, least hype-prone answers. Accept the $20-or-nothing pricing.
Choose Gemini if: you live in Google Workspace or on Android, you want search-grounded answers, or you want the cheapest capable paid tier at $4.99.
And the honest fourth option, recommended by several published guides: use more than one. Gemini for daily lookups, Claude for deep thinking, ChatGPT for image generation — each on its free tier — covers most needs for $0.
Meridian's verdict
There is no single best AI assistant in late 2026, only the best one for your workflow and your budget. ChatGPT remains the default for breadth, Claude is the connoisseur's pick for depth, and Gemini is the value-and-integration play. Given how fast this market moves — GPT-6 Astra is weeks old as we publish — treat any subscription as a month-to-month decision, not a marriage. Revisit the choice quarterly; the leaderboard will keep shifting.
Sources
