Birbla

Join the discussion

Write your take first — we'll ask for email only when you're ready to publish.

Sign in
  • Hacker News
  • Alternate headline:

    “After three years of unprecedented investment including governments, Facebook, and Google dumping money into the promise of AI, ChatGPT still has over 50% of the traffic.”

    Also, grok isn’t been on that list, further proving that no serious people use it.

    by Forgeties79
  • Grok is there, just a little bit too small to label. Dark brown on the chart.

    4.6 is a good model, by the way. Legitimately frontier-level and a good communicator.

    by A_D_E_P_T
  • Given this anyone still thinking Google lost needs to update their belief system
    by AndrewKemendo
  • Google has lost $302.8B while only making $25B on AI.

    That's not only losing, but record-setting losing.

    by VCFundedGenYer
  • Lost at what, though? Waymo is gonna pan out but they sold their domain business to squarespace despite having GCP, and Google workspace deletes data when changing billing which is absolutely unacceptable.
    by fragmede
  • What's the methodology? Is Gemini actual gemini.google.com or mixed with google.com and Android searches?
    by inerte
  • 54% to go, good work team.
    by thot_experiment
  • Why? Who is the replacement? Some state backed Chinese thing that is running loss leaders for a rug pull in the near future like Deepseek just did?
    by dyauspitr
  • Meme: "I'm doing my part" Starship Troopers.
    by razster
  • This source-ier chart is much better: https://cdn.prod.website-files.com/6213ddd7bd3eb8fb93bf1da4/...

    Stacked charts might be 2nd only to pie charts in badness.

    by xnx
  • I used to use Gemini (the product) because it's free. But its responses are so incredibly bad, and the interface has so many dumb bugs, that I stopped using it and switched to OpenCode Go models + Web UI. Gemini models can recite deep knowledge, but they can't do constructive things, can't do extended web research, can't provide more than a couple options for an answer. Either their system prompt is nerfed like crazy or the models just suck at agentic work. Even ChatGPT from 2025 was better than Gemini today.
    by 0xbadcafebee
  • Gemini's Deep Research is pretty good. In contrast, strangely enough, ChatGPT nerfed its Deep Research offering, although a custom GPT or Work mode can trivially make up for it. I admit that Google's offerings generally have too many serious bugs to be usable.
    by OutOfHere
  • I think it'd be more informative to see total users per provider, rather than % shares.
    by Retr0id
  • Worldwide AI-chatbot web visits: ChatGPT 76% → 54%. Gemini 6% → 28%. Claude 1% → 9%. This is not monthly users and not revenue.
    by echohive42
  • I understand Claude being niche (until they started running Superbowl ads?), but why'd it take Google this long to start catching up?
    by andai
  • Gemini is actually good enough for pretty much everyone I know who is not in tech and uses Android. These people are downright pleased with Google's AI summaries.

    The fact that Google can push Gemini out in so many ways with their existing marketshare is mind boggling. The fact that ChatGPT hasn't lost 50-100% of web share shows it's Google's game to lose at this point.

    by xyzzy_plugh
  • > These people are downright pleased with Google's AI summaries.

    Consider me part of "those people", I guess, but are you guys not, at this point? There was a period near the start where they were bad, but I find them perfectly usable now. I'll often open a tab, Google something, and just read the summary because it's basically instant and I don't need any more than just that, a summary.

    I'm genuinely confused why people take so much issue with them. Sure they're still wrong sometimes, but it's blatantly obvious when that's the case, and I can just scroll past.

    by ipdashc
  • But are they paying for Gemini?
    by dyauspitr
  • > Google's AI summaries

    i really doubt the AI summaries use Gemini or even Gemma

    by TiredOfLife
  • It’s good. Returns a bespoke response is some ridiculous 0.3s or something like that which is mind blowing and I have no idea how they are pulling that off at all in the first place let alone the sheer scale of it.
    by dyauspitr
  • If only they could make it be wrong less? I know they need to have these results appear in record time, but I wonder the collective damage being done by just how horribly wrong the answer often is.
    by stingraycharles
  • I've been using Gemini as an advance search engine. Their main search with AI isn't that useful but Gemini is. They're the only AI online I will use and just for searching/compiling. For work/development it is Qwen3.8. Then again now that I found out Pi has a web search function I've been using Gemini less and less.
    by razster