Claude Beat ChatGPT 2-to-1 in Our Reader Poll. Here's the Deeper Story You Told Us.

On Sunday, we asked Neuron readers a simple question: what AI tool do you use the most right now? We expected ChatGPT to keep its lead, for obvious reasons. After all, it's the name every non-technical person knows. It's the default answer at dinner parties. It has the memory, the history, the voice mode. It's the bandaid of the AI; the household name brand. The Google of the AI wave of the internet, if you will (sorry, actual Google).

Instead, Claude won by a landslide. And when 200+ of you wrote in to explain why, the story got a lot more interesting than the headline suggests.

First up, the TL;DR

• Final tally (3,143 votes): Claude 1,449 (46%); ChatGPT 790 (25%); Gemini 431 (14%); Microsoft Copilot 180; Perplexity Computer 74; Other 66; Grok 57; LM Studio/OpenCode/Ollama 49; OpenClaw 47.
• Why Claude won the write-in crowd: coding quality (especially Claude Code), writing that "gets me," depth of reasoning on hard work, Cowork for multi-session workflows, and (surprisingly often) ethics and brand alignment. This reason has picked up steam lately.
• Why ChatGPT's 790 stayed put: price, memory/projects (it "knows my projects"), sheer habit, custom personas, and specific sticky use cases like job search, health coaching, and parenting help.
• The Copilot paradox: most Copilot users selected it because their employer forced the issue, not because they preferred it. Many explicitly said they'd rather be using Claude. Microsoft, what y'all doing? Fix this! Maybe buy Tasklet or something and make agents much easier??
• Gemini's base is the Google ecosystem: Android, Workspace, Pixel bundling, plus best-in-class design output via Nano Banana Pro and standout research reports.
• Everyone stacks tools. Claude picked up 34 secondary mentions from people who chose a different primary. Gemini got 36. ChatGPT got 28. Almost nobody uses just one.
• The long tail is alive: 40+ other tools mentioned beyond the poll options, including Cursor (5 mentions), NotebookLM (7), Manus (4), Codex, Jasper, Suno, Descript, Abacus, and one reader running Claude Code + GLM-5.1 via Ollama on a 2012 computer. Hell yeah, brother; this is the way!!

The Headline Finding: Claude by a Mile

In raw numbers, Claude pulled 1,449 votes to ChatGPT's 790. That's a 1.83x lead. Gemini landed a distant third at 431. Microsoft Copilot, despite its enterprise distribution advantage, took fourth at 180.

Quick caveat before anyone screenshots this: this is a self-selected sample of an AI-focused newsletter's most engaged readers. It's not a gauge of the general public. Most people on Earth still default to ChatGPT, because most people on Earth haven't heard of Claude. Although... the latest adoption numbers appear to show that tide is shifting...

But within our audience at The Neuron (AI-curious knowledge workers aged 25-45 who actually open newsletters about AI, plus curious folks of all ages), the shift is real. The write-in comments make it feel less like a fluke and more like a migration in progress.

Why Claude Fans Picked Claude

We sorted the 89 Claude write-ins into themes. Five stood out.

1. Coding. Always coding. "It was ChatGPT until I started using Claude Code" was basically a refrain. One reader wrote about trying Gemini first, hitting a quota wall, and finally giving Claude Code (Sonnet 4.6) a shot: "WHY WASN'T I USING CLAUDE SOONER!" Multiple readers use Claude Code alongside Cursor, Warp, or GitHub Copilot for a layered dev workflow.
2. "It just gets me." Writers love Claude. Novel-writing, essay drafting, trust-document drafting for estate planning, podcast scripts. One reader said Claude Opus produced full-text legal language while ChatGPT kept defaulting to bullet phrases despite clear instructions.
3. Depth of reasoning for "hard work." Several readers explicitly split their stack: Claude for "real meaningful outputs" or "hard work" or "bigger projects," with ChatGPT relegated to quick questions. "Claude for major projects, free Chat for quick Qs" was a common pattern.
4. Cowork is sticking. "I'm using Claude Cowork, running multiple sessions at the same time." "Claude cowork and code are 80-90% of my use." Cowork has become a workflow primitive for power users (not just a novelty).
5. Brand and ethics. This one caught us off guard by how often it showed up. Readers explicitly cited dropping ChatGPT over the Pentagon deal. One mentioned "personal animus against Sam Altman and Elon." Another wrote that they chose Claude because they want "a company that cares to try for good outcomes." For a non-trivial slice of Claude's lead, values, not features, are the loyalty driver.
6. 1. We had a feeling this would help overall AI adoption in what we're calling internally here at The Neuron "The AI player hater's ball of 2026" (a.k.a lead up to the 2026 US midterms), as many folks who lean anti-AI might find this message more appealing and give Claude (and by extension, AI) a more serious shot because of it.
6. The integration compliment: one reader specifically requested coverage of Claude integrations with Canva, MS Office, Asana, and Figma. That ecosystem matters. We got you, we promise. It's coming!

Why ChatGPT's 790 Stayed Loyal

ChatGPT still took second place, and the reasons were clustered around a word you'd recognize from any Econ 101 textbook: switching costs.

Price was the single most-cited reason. "Only because it is cheaper than Claude. I would much rather be using Claude except the price and low limits." "Claude is good (v4.6), but really too expensive!" "I'm not too keen on Claude's pricetag and hype." This came up again and again.

Memory and Projects keep people locked in. "Began with ChatGPT, which now has my history." "It knows the most about my projects." One mother of five built ChatGPT into her parenting stack (meal plans, ADHD learning plans, school-email triage) and doesn't want to rebuild that context elsewhere. Another reader used ChatGPT to revise her CV through a federal-to-state career pivot, and stayed loyal because she could "remain the human in the loop."

Habit is a moat. "Just built a habit for ChatGPT. Haven't taken much time to play with the others yet. It's becoming overwhelming." Relatable.

Then there's the heartwarming stuff.

• One reader credits ChatGPT with saving her hand. After the VA medical center injected the wrong chemical and tried to gaslight her, ChatGPT drafted an email with the right medical terminology to get a specialist referral and emergency surgery.
• Another reader built an alter-ego character named "Aegana" that speaks in a different voice when summoned.
• Someone uses it for mental-health support during a job search.
• Someone else is trying to launch a digital business on TikTok / YouTube / Instagram and wants a cheaper tool that actually saves them time. (Shout to them; we hope they find it... if you do, let us know!!)

The Copilot Paradox: Forced, Not Chosen

Microsoft Copilot took 180 votes overall. In the written responses, 19 readers selected it. And in almost every comment, the same pattern appeared:

Copilot wins because the employer decided it wins. The comments stacked up:

• "Forced by company; at home I like Claude better."
• "I am forced into Copilot which I do not care for at all."
• "Copilot is the easiest to access in a restricted financial services environment."
• "Many of us are in the 365 environment. We need to use Copilot for work even if other AI tools would otherwise be technically more useful."
• "They cram it down your throat on every app."

One reader made the positive case well. In-app micro-interactions (writing an Outlook reply, fixing an Excel formula) beat copy-pasting out to a separate chat window. Microsoft's distribution advantage is real; your data and calendar are already there.

But the most memorable Copilot comment of the bunch came from a user who calls their Copilot "Jason, after Jason Momoa." Jason is about to help her build a meal plan. Godspeed, Jason.

The softer underbelly: multiple enterprise-Copilot users described it as "just as dumb at the start of your 100th chat as your first," unable to share chats, and unable to pick up the same thread on another day. One speculated Microsoft might be measuring number of new chats as a success metric. Concerning if true.

The strategic takeaway here: captive market share is fragile. If employees were allowed to self-select, a meaningful chunk of that Copilot base would migrate to Claude or ChatGPT tomorrow. The recent rollout of Claude-powered features inside Office is going to make that dynamic even more interesting.

Gemini: Riding Google's Rails (And Owning Design)

Gemini's 29 written selections broke down into a few clean buckets.

• Google ecosystem lock-in (the largest). "I have a Google phone." "I have Gemini because I bought the Pixel 10 Pro." "I use Google Workspace at work." "Gemini syncs with my other Google apps on Android." This isn't loyalty so much as default-tool gravity.
• Design work, specifically Nano Banana Pro. "I've been designed heavy in my work so I use Gemini the most often for Nano Pro." "Claude for coding, Gemini for image work, OpenAI API for cheap low-value work." Image generation and editing are Gemini's fortress.
• Research quality. One reader keeps Gemini "even though I don't use it for anything else" because its research reports "are so much better than any other platform." Another pairs Gemini research with Claude writing.
• Video features and ecosystem adjacencies. NotebookLM, Antigravity, Gemini CLI, Google CLI, AI Studio. A reader described using Gemini like a personal AI assistant: recording videos of her grocery haul for recipe ideas, her home gym for workout routines, and dictating bedtime stories for her toddler. That's actually an incredible use case we should write more about.
• Gemini also had the most secondary mentions of any tool: 36. Even people who don't pick it as primary keep it on the bench for something specific (usually images, research, or "second opinions"). That's a durable position. Also, same for us here at The Neuron. Gemini has fluctuated in our stack, but it's remained a solid second even when not first because it's so versatile.

The Underdog Corner

Grok (57 votes) had a small but evangelical base. Three themes came up:

• Hallucinates less than ChatGPT and Gemini for high-stakes tasks.
• Distrust of Google and the Sam-Altman-and-Elon vibe check pushes them away from bigger options.
• Tesla-to-laptop continuity: "I talk to it in the Tesla then continue from laptop." This is a smart platform play for Elon.

Also cited: Grok ranks solutions from easiest to hardest to implement, which readers appreciate.

The rest:

• Perplexity (74 votes) is for people who want access to all models through one interface. "With Perplexity, I get all the models." Multiple readers got a free year through school or Perplexity Pro Educator promotions. The criticism: Perplexity Comet (the browser agent) burns credits fast.
• LM Studio / Ollama / OpenCode (49 votes) is the other end of the open spectrum: offline, local, self-hosted. One reader is running Claude Code + Pi Coding Agent + GLM-5.1 via Ollama. Another is running multimodel agents on "an old 2012 busted computer" through Ollama's webUI with Continue and VS-Codium. Respect.
• OpenClaw (47 votes) pulled a motivated crowd who want to move away from centralized megacorp AI. Feel that. One reader is running OpenClaw powered by Sonnet API via OpenRouter ("It's been a huge PITA. I use Claude constantly to fix it. Rotating APIs consumes most of my time"). Open-source purism has a cost. My friend, have you heard of Gemma 4? Watch this.

Everyone Stacks Tools (The Real Story)

Here's the finding that doesn't fit on a bar chart. When we counted which tools got mentioned in addition to people's primary pick, the numbers are striking:

• Gemini: 36 secondary mentions
• Claude: 34 secondary mentions (often from ChatGPT, Gemini, and Copilot users)
• ChatGPT: 28 secondary mentions
• Microsoft Copilot: 21 secondary mentions (mostly "forced at work")
• Perplexity: 18 secondary mentions
• Grok: 8 secondary mentions
• NotebookLM: 7 secondary mentions
• Cursor: 5 secondary mentions
• Manus: 4 secondary mentions

Common stacks readers described:

• "Claude for main work + Gemini for research + ChatGPT for fallback"
• "Copilot at work (forced) + Claude at home"
• "Claude for writing + Gemini for images + OpenAI API for cheap bulk tasks"
• "Claude + Perplexity for research + NotebookLM for mindmapping / podcasts"
• "Gemini for Google sync + ChatGPT for spreadsheets"

The single-tool loyalist is rare. The multi-tool stacker is the median

The Long Tail: 40+ Other Tools You Mentioned

Beyond the nine poll options, readers named a lot of additional tools. Grouped by category, with mention counts:

• Coding agents (beyond Claude Code): Cursor (5), Codex (2), GitHub Copilot (2), Warp (2), Windsurf (1), Pi Coding Agent (1). One reader uses "cursor composer2." Another runs "Claude Code + GPT agents in Cursor."
• Research / all-models aggregators: NotebookLM (7), Abacus LLM (1, "access to 29 models").
• General-purpose AI agents: Manus (4), Hermes Agent (2), Genspark (2), Shortcut.AI (1), Minimax Agent (M-2.7) (1).
• Writing / productivity: Jasper (1), Notion AI (2, including Notion agents).
• Media / creative: Suno (1, music production), Descript (1, "I just started using it for video and it is AMAZING"), Runway (1, video), CapCut (1, video editing), Photoshop generative fill (1, used by a professional retoucher), Canva (1), Figma (1).
• Open-source / privacy-focused: DeepSeek (2), Ollama (2), GLM (1), Proton Lumo (1), Le Chat / Mistral (1).
• Google ecosystem adjacencies: Antigravity (1), Google CLI (1), Gemini CLI (1).
• Curious one-offs we couldn't always pin down: Lexi's protege (1), Repliy (1, possibly Replika?), ChatNow (1), CopyMind (1),"NanoClaw" (1).

The Wildest Things You Told Us About Your Workflows

These are the patterns and stories that didn't fit anywhere else, but we couldn't stop thinking about:

• Geography and compliance shape tool stacks more than specs do. One reader uses Manus specifically because they work in Hong Kong under "strict rules against VPN use within the company." For a non-trivial slice of the global workforce, the question isn't "which model is best?" It's "which model is accessible?"
• The "second opinion" workflow is quietly spreading. A Codex Desktop user pipes broader decisions through Gemini "for second opinions to help inform broader decisions." Treating one AI as a reviewer of another's output is an emerging pattern we'll be watching.
• Field-sales training delivered via NotebookLM podcasts. One reader generates podcasts for sales reps so they can "get training on the road or hear about their next visit." Audio-first AI for frontline workers might be one of the most underrated enterprise plays right now.
• The MLB Draft was the tipping point. A loyal Gemini user told us they "just started using Claude for a MLB Draft that Gemini could not handle." That's the specific moment they stopped being a Gemini maximalist. Niche use cases flip brand loyalty faster than benchmark comparisons do.
• Video-calling the AI. Another Gemini user: "Gemini has video option like video call with ai and can ask questions about what you're looking at." Camera-on-you, AI-on-screen has moved from demo-reel gimmick to actual daily workflow for at least some readers. GPT has this as well, but Claude does not... last we checked, anyway.
• Stack math, treated like portfolio allocation. "Claude Cowork and Code are 80-90% of my use. I use OpenAI 10-15%, primarily research. Gemini for an alternative perspective, visual infographics, NotebookLM." Some readers know their tool split down to the percentage point.
• Hardware purchases now ship with AI subscriptions attached. One Claude fan admitted they only tried Gemini first because they got a "free year of Google AI Pro because I bought the Pixel 10 Pro at launch day." Phone bundles are doing the work Google's marketing can't. More reason for Microsoft to launch Clawbooks!! Although arguably Linux should publish a Clawbook. An Openbook, perhaps?
• Team-account dissolution wiped the history. One reader's boss cancelled their team ChatGPT subscription. Every chat, every saved context, every thread: deleted, no export, no retrieval. They migrated to Grok with a new rule: save everything locally. Enterprise data portability just became a real concern.

Oh yeah, and there's that one creative use case... one reader makes, let's call it romantic AI video content for a niche creative vertical and edits it locally to keep the files off any cloud service. Whatever the genre, the workflow choice (local-only editing to keep creative IP off corporate servers) is a pattern more people are going to care about as models get more capable. Offline and local tooling matters.

What This All Tells Us

Four takeaways from reading every comment:

1. Claude's lead is real within the power-user cohort, but it's also fragile at the edges. Price was the single biggest reason people haven't switched. Anthropic's cheaper, smarter Sonnet and Haiku 4.7 models, plus the new Claude-in-Office integrations are likely to convert more ChatGPT holdouts over the next quarter.
2. Microsoft's Copilot base is captive, not committed. When Claude shows up inside Word and Excel natively, a real migration could begin, because many Copilot users are already using Claude at home on their own dime. The plan here can't just be charge agents to access Microsoft apps, right guys? I say go hard into making your own agentic computers for local inference to own the stack (serve high intelligence cloud models + local inference models for regular folks; strip down Microsoft OS to bare bones and build the agents into the CLI and sell the equivalent of an agent Chrome book... call it a Clawbook or something!).
3. Tool choice is becoming identity. Ethical defection from OpenAI and xAI is real. It's not the majority driver, but it's loud enough that it moved some votes. Brand messaging now shapes tool loyalty the way it shapes sneaker loyalty. Narrative capital is alive and well...
4. The single-tool question is dead. How are you standing out in the stack? Nobody uses one AI. Everyone has 3-5. Kinda like streaming services?? The interesting angle from here on out is how any given AI model / tool fits neatly inside the stack of other tools. In 2024-2025, every AI company raced to copy each other's winning features. Now it's time to differentiate.

One open question we don't have an answer to yet: as Claude becomes the underlying model for more third-party products (Perplexity, Copilot, various agents), does Anthropic benefit from brand affinity even when the Claude name is hidden? Or does it lose the narrative to whichever company owns the interface? Also, Claude is the more expensive option here. Can Anthropic keep the favor of the power user and the power normie while keeping their place as the "premium" AI offering? We'll be watching.

Thanks again to everyone who answered the poll and especially the 200+ of you who wrote in. We read every single one. We learned much more from your write-ins than from the bar chart.

(Also: if Jason-the-Copilot ever makes a good meal plan, tell us. We want to know what Jason is cooking.)