You can now use Claude to search the internet to provide more up-to-date and relevant responses. With web search, Claude has access to the latest events and information, boosting its accuracy on tasks that benefit from the most recent data.
When Claude incorporates information from the web into its responses, it provides direct citations so you can easily fact check sources. Instead of finding search results yourself, Claude processes and delivers relevant sources in a conversational for
48 Comments
morisil
I added this functionality already some time ago in my Claudine agent:
https://github.com/xemantic/claudine/
It costed roughly 30 lines of code:
https://github.com/xemantic/claudine/blob/main/src/commonMai…
msp26
> in feature preview for all paid Claude users in the United States. Support for users on our free plan and more countries is coming soon
US only
ubicomp
Excited to see this. I've really been enjoying Claude. It feels like a different, more creative flavor of experience than GPT. I use Claude a lot for dialogues and exploring ideas, like a conversational partner. Having web access will add an interesting dimension to this.
Brusco_RF
Excited to see how this compares to Perplexity or Gemini. I remember that ChatGPT used to be able to search the web, but last I checked it it couldn't. I wonder why they removed that feature
danirogerc
Couldn't a lot of front-ends using Claude API do this already? What's new?
NBJack
I wonder if it will actually respect the robots.txt this time.
firloop
Any information on what search engine is powering it?
goatmeal
kagi already lets me use claude to search the web. how is this different?
hombre_fatal
Aside, does anyone know of an app like Perplexity for surfing the news in a foreign language (language practice)?
Perplexity's "Explore" tab translates its news to your local language, and its curated news items are all pretty interesting, but the problem is that there are so few of them. I seem to get maybe a dozen stories in a day. I paid their subscription for a month just to listen to the news on my walk, but didn't renew because of this.
A foreign news site like BBC Mundo (Spanish) on the other hand barely has any stories outside of a few niches. Its tech section only has a few stories per week.
Hmm, maybe I want a sort of RSS reader that AI-translates stories for me. But I don't really want to maintain a feed myself either.
Apple News would probably do it since they also have good curation, but afaict they still don't support foreign news sources (why???).
mvieira38
What's up with the geoblocking of Claude features? Not the first time it happens
itpcc
Now I understand why Gitlab was (is?) attacked[0] by those hideous bots.
[0] https://news.ycombinator.com/item?id=43422413
pcj-github
Does not really say /how/ it's performing a web search… Is it tapping into it's "own" corpus of material or calling out to some other web search engine?
punkpeye
There is already a 100 ways of doing it using MCP
https://glama.ai/mcp/servers?searchTerm=search
What's the benefit of bringing native integration?
light_triad
Good news. I integrated Claude with a scrapper to get info from pages and it was not giving hallucinations 99% of the time. Hope this works out of the box now.
blensor
Funny thing is that I have the obsidian-mcp-tools installed and today claude-desktop just starting fetching stuff from the web through that because it exposes a fetch tool to claude.
So this limitation is a bit arbitrary anyway.
ProofHouse
Awesome, but I also do want to say it’s pretty sad it took this long straight up. Literally no excuse. But I’m glad they finally got to a feature that was launched more than a year ago on competitors.
McNutty
I haven't used Claude yet, but heard many good things. So I'm surprised to see that they're so far behind on this feature.
ignoramous
These are interesting times.
It wasn't long ago that a uni senior who worked for a decade+ on Google Search told me that it was hopeless anyone tries to compete with Google not because it sees a tonne of signals that helps with IR but because of its in-house AI/ML.
It turns out that the org that built the ultimate AI/ML that runs rings around anything that came before it for NLP (and thus IR) was a sister team at Google Translate.
It isn't inconceivable that a kid might be able to build a Google-quality web search, scalability aside, on CommonsCrawls data in a weekend. As someone who built re-ranking algorithms for a search engine built atop Yahoo! and Wikipedia (REST/SOAP) APIs back in the late 2000s as a side project (and experienced the launch and subsequent iterations of Echo/Alexa up close at Amazon), the current capabilities (of even the open weight multi-modal models) seem too good to be true.
Google itself though is saved by its enormous distribution advantages afforded by Chrome (3B to 5B users) and Android (3B+), aside from its search deals with Apple and other browser vendors.
fourside
I’ll be interested in trying it. My admittedly limited experience with this on ChatGPT has been disappointing. ChatGPT falls for the SEO content that has taken over the web.
As an example, I recently travelled abroad to a popular vacationing spot and asked ChatGPT for local recommendations on what to do. When it gave me answers directly, they were pretty solid. But when it “searched the web” instead, the answers were awful. Every single result it suggested had terrible ratings. It did this repeatedly. One of those times I asked it to pick something with better ratings and it sort of improved but not by much.
Of course this is another tool and maybe Claude uses better sources or a better algorithm, but in this case where there was a concrete number tied to the results, that while not perfect, aims to rate the quality of a result, it still did not filter out low quality answers. I’m not sure I trust these LLMs to do any better when there aren’t such ratings available. The available input data is just not very good, and now LLMs are being used to feed that low quality, SEO machine.
mediumsmart
thats great – the web he has been trained on or the one from Google?
tttym
It's like a line of platforms waiting for their own agents for web search
rgbrgb
Is there a way to access the new web browsing capability via API?
deadbabe
Is no one concerned about LLMs just feeding people SEO ads as content?
ConanRus
finally
jetrink
> With web search, Claude has access to the latest events and information, boosting its accuracy on tasks that benefit from the most recent data.
I'm surprised that they only expect performance to improve for tasks involving recent information. I thought it was widely accepted that using an LLM to extract information from a document is much more reliable than asking it to recall information it was trained on. In particular, it is supposed to lead to fewer instances of inventing facts out of thin air. Is my understanding out of date?
tantalor
It says a lot about their product vision and intended market that the example query is typescript migration question.
Do they not care about typical search users? Only developers?
beezle
At this point it is probably easier to poison web pages for AI crawlers in a way that does not taint the human experience.
ilaksh
I've been using Tavily's search API for my MindRoot agents. Seems to work fairly well and much easier to set up than Google's search API.
Anyone know if there is something better? I was thinking of trying Perplexity maybe.
tgtweak
Feels like a catch-up feature to chatgpt… honestly the biggest holdback for me on anthropic is the output token limit on sonnet… 8000 tokens max output is really limited (and 200k tokens in) compared to other offerings – especially considering that I suspect most sonnet users are not chat-users but api users.
bfeynman
why does perplexity exist anymore? They were one of anthropic's biggest customers and had been finetuning claude models for search for a while.
NewJazz
Bizarre that they choose to publish this right as a thread criticizing AI crawlers gets bumped off the front page.
https://news.ycombinator.com/item?id=43422413
tcdent
Searching the web is a great feature in theory, but every implementation I've used so far looks at the top X hits and then interprets it to be the correct answer.
When you're talking to an LLM about popular topics or common errors, the top results are often just blogspam or unresolved forum posts, so the you never get an answer to your problem.
More of an indicator that web search is more unusable than ever, but interesting that it affects the performance of generative systems, nonetheless.
CalChris
I find myself Googling less often these days. Frustrated with both the poor search results and impressed with the quality of AI to do the same thing and more, I think search's days are numbered. AOL lasted as an email address for quite some time after America Online ceased to be a relevant portal. Maybe Gmail will as well.
joshstrange
Massive props to Anthropic for announcing a feature _and_ making it available for everyone right away.
OpenAI is so annoying in this aspect. They will regularly give timelines for rollout that not met or simply wrong.
Edit: "Everyone" = Everyone who pays. Sorry if this sounds mean but I don't care about what the free tier gets or when. As a paying user for both Anthropic and OpenAI I was just pointing out the rollout differences.
Edit2: My US-bias is showing, sorry I didn't even parse that in the message.
g42gregory
Anybody knows if this search will works inside Cursor? Do we need to do anything to make 3.7-thinking to search the web?
nimish
So what's perplexity's raison detre at this point?
scudsworth
"claude can now ddos random websites . . . more so"
douglee650
You know what Claude can't do? Successfully vibe me a solution to:
"""
i need a bashrc command that will map the alias "logg" to open macvim to the file at ~/log.txt, then execute the macro defined by "<leader>z"
"""
Note <leader>z ends with user in insert mode, Claude provides solution below but puts me in edit mode. (I still have to press "i")
alias loggg='mvim ~/log.txt -c "normal <leader>z"'
hansmayer
So, referring specifically to the example they show on the front-page, what value does this bring actually? The best example they could come up with is Typescript migration ? Really? Weren't the LLMs supposed to be a superior alternative to searching the web? Why do we need to produce more CO2 to do the same we could have done at the fraction of the cost, of course at the time when the google search was still working?
wewewedxfgdf
I'm waiting for Claude's API to support projects with file uploads like its web UI.
ralusek
Best models for search, in order:
OpenAI Deep Research
Grok Deep Search
Gemini Deep Research
Grok + Search
Gemini + Search
ChatGPT + Search
These are just my opinions, but I do use this feature all the time. Haven't used Claude enough to get a sense of where it would fit in.
gcanyon
Funny, I literally just two days ago asked Claude to provide an outline of the functionality of a product, giving it the web site. It of course refused. So I downloaded the text of the site and passed that in, and got mediocre results.
The results based on giving the source URL directly were better. Still a bit generic and high-level and vague, as LLMs tend to be, but better than the text-download version a couple days ago. And of course much easier to generate!
ForTheKidz
Great, now we just need a decent search engine.
artembugara
Search the web is apparently using SERP.
It’s just breaks my head. We’ve build LLMs that can process millions of pages at a time. But what we give them is a search engine that is optimized for humans.
It’s like giving a humanoid robot access to a keyboard with a mouse to chat with another humanoid robot.
Disclaimer: I might be biased as we’re kind of building the fact search engine for LLMs.
jsight
I really want these to be able to find and even redisplay images. "Search all the hotels within 5 miles of this address and show me detailed pictures of the rooms and restrooms"
Hotels would much rather show you the outside, the lobby, and a conference room, so finding what the actual living space will look like is often surprisingly difficult.
sylware
What I really would like to know: do they use a web crawler with an AI strapped to the mouse and keyboard of a javascript-ed web engine?
Taters91
so can I
greatNespresso
Woaw, so excited about this! Has anyone tried it out already?