Archive link

Full text

Google is now the only search engine that can surface results from Reddit, making one of the web’s most valuable repositories of user generated content exclusive to the internet’s already dominant search engine.

If you use Bing, DuckDuckGo, Mojeek, Qwant or any other alternative search engine that doesn’t rely on Google’s indexing and search Reddit by using “site:reddit.com,” you will not see any results from the last week. DuckDuckGo is currently turning up seven links when searching Reddit, but provides no data on where the links go or why, instead only saying that “We would like to show you a description here but the site won't allow us.” Older results will still show up, but these search engines are no longer able to “crawl” Reddit, meaning that Google is the only search engine that will turn up results from Reddit going forward. Searching for Reddit still works on Kagi, an independent, paid search engine that buys part of its search index from Google.

The news shows how Google’s near monopoly on search is now actively hindering other companies’ ability to compete at a time when Google is facing increasing criticism over the quality of its search results. And while neither Reddit or Google responded to a request for comment, it appears that the exclusion of other search engines is the result of a multi-million dollar deal that gives Google the right to scrape Reddit for data to train its AI products.

“They’re [Reddit] killing everything for search but Google,” Colin Hayhurst, CEO of the search engine Mojeek told me on a call. Hayhurst tried contacting Reddit via email when Mojeek noticed it was blocked from crawling the site in early June, but said he has not heard back.

“It's never happened to us before,” he said. “Because this happens to us, we get blocked, usually because of ignorance or stupidity or whatever, and when we contact the site you certainly can get that resolved, but we've never had no reply from anybody before.”

As Jason wrote yesterday, there’s been a huge increase in the number of websites that are trying to block bots that AI companies use to scrape them for training data by updating their robots.txt file. Robots.txt is a text file which instructs bots whether they are or are not allowed to access a website. Googlebot, for example, is the crawler or “spider” that Google uses to index the web for search results. Websites with a robots.txt file can make an exception to give Googlebot access, and not other bots, so they can appear in search results that can generate a lot of traffic. Recently Google also introduced Google-Extended, a bot which crawls the web specifically to improve its Gemini apps, so websites can allow Googlebot to crawl but block the crawler Google uses to power its generative AI products.

Robots.txt files are just instructions, which crawlers can and have ignored, but according to Hayhurst Reddit is also actively blocking its crawler.

Reddit has been upset about AI companies scraping the site to train large language models, and has taken public and aggressive steps to stop them from continuing to do so. Last year, Reddit broke a lot of third party apps beloved by the Reddit community when it started charging to access its API, making many of those third party apps too expensive to operate. Earlier this year, Reddit announced that it signed a $60 million with Google, allowing it to license Reddit content to train its AI products.

Reddit’s robots.txt used to include a bunch of jokes, like forbidding the robot Bender from Futurama from scraping it (User-Agent: bender, Disallow: /my_shiny_metal_ass) and specific pages that search engines are and are not allowed to access. “/r*.rss/” was allowed, while “/login” was not allowed.

Today, Reddit’s robots.txt is much simpler and more strict. In addition to a few links to Reddit’s new “public content policies,” the file simply includes the following instruction:

User-agent: *
Disallow: /

Which basically means: no user-agent (bot) should scrape any part of the site. “Reddit believes in an open internet, but not the misuse of public content,” the updated robots.txt file says.

“Unfortunately, we’ve seen an uptick in obviously commercial entities who scrape Reddit and argue that they are not bound by our terms or policies,” Reddit said in June. “Worse, they hide behind robots.txt and say that they can use Reddit content for any use case they want. While we will continue to do what we can to find and proactively block these bad actors, we need to do more to protect Redditors’ contributions. In the next few weeks, we’ll be updating our robots.txt instructions to be as clear as possible: if you are using an automated agent to access Reddit, you need to abide by our terms and policies, and you need to talk to us.”

Reddit appears to have updated its robots.txt file around June 25, after Mojeek’s Hayhurst noticed its crawler was getting blocked. That announcement said that “good faith actors – like researchers and organizations such as the Internet Archive – will continue to have access to Reddit content for non-commercial use,” and that “We are selective about who we work with and trust with large-scale access to Reddit content.” It also links to a guide on accessing Reddit data which plainly states Reddit considers “Search or website ads” as a “commercial purpose” and that no one can use Reddit data without permission or paying a fee.

Google did not respond to a request for comment, but its announcement of the company’s deal with Reddit points out not only how valuable Reddit is for training AI, but what many of us already know: As Google Search gets increasingly worse in turning up relevant search results, one of the best ways to still get them is to add “Reddit” to your search queries, directing Google to a site where real humans have been writing advice and recommendations for almost two decades. There are a lot of ways to illustrate how useful Reddit can be, but I’m not going to do better than this video:

https://www.youtube.com/watch?v=tcJcw55zIcc

The fact that Google is the only search engine that leads users to that information now, and that it is apparently the result of a $60 million deal around AI training data, is another example of the unintended consequences of the indiscriminate scraping of the entire internet in order to power generative AI tools.

“We've always crawled respectfully and we've done it for 20 years. We're verified on Cloudflare, we don't train AI, we're like genuine, traditional genuine searching, we don't do ‘answer engine’ stuff,” Hayhurst said. “Answer engine” is Perplexity’s name for its AI-powered search engine. “The whole point about Mojeek, our proposition is that we don't do any tracking. But people also use us because we provide a completely different set of results.”

Reddit’s deal with Google, Hayhurst said, makes it harder to offer these alternative ways of searching the web.

“It's part of a wider trend, isn't it?” he said. “It concerns us greatly. The web has been gradually killed and eroded. I don't want to make too much of a generalization, but this didn't help the small guys.”

monke-beepboop

  • glans [it/its]
    ·
    5 months ago

    we are really in need to a viable FLOSS search engine. That can do its own indexing instead of repackaging google results like searx does. Maybe the spidering could be distributed somehow so the small self hosters could benefit from it while also able to apply their own standards, priorities, sorting etc.

    • hypercracker
      ·
      edit-2
      5 months ago

      unfortunately search is expensive in a way that FLOSS does not solve, it requires a lot of hosting infrastructure and boring volunteer labor to fine-tune results to combat spam (and spam might even benefit from looking at the FLOSS rules that filter it)

      • glans [it/its]
        ·
        5 months ago

        (and spam might even benefit from looking at the FLOSS rules that filter it)

        idk i feel like that might be a problem which is created or at least greatly exaggerated by monopolies. if there was a diversity of search engines it would be much more difficult to do shitty SEO on all of them at the same time. You'd need a whole team combing through repo hosting sites and mailing lists to figure it out.

    • Owl [he/him]
      ·
      5 months ago

      Crawling could be distributed and shared, but indexing is a bigger problem.

      All the things you'd want to be different on some sort of federated search platform (standards, priorities, and sorting, as you say) are things that require different indexing. But the index is the big expensive part that would most need to be shared.

      • christian [he/him]
        ·
        5 months ago

        Thanks for this recommendation, I hadn't heard of this. Looks promising.

      • glans [it/its]
        ·
        5 months ago

        Oh that looks cool. Do you run an install of it or are you using the install on their main page? Are there other instances

        The hw requirements aren't prohibitive. I mean it's not nothing, maybe a few hundred upfront and then the connection. Well within reach especially with support of an existing organization who'd be willing to physically house it. I guess SSDs would be the largest part of the cost.

        an x86-64 machine, have at least 16GB of RAM, and at least 4 cores. It is designed to run on physical hardware, and will likely be very expensive to run in the cloud.

        Crawling requires a decent network connection, ideally at least 1 Gbps. 100 Mbps will work, but will be slower.

        Storage requirements are highly dependent on the size of the index, and the number of documents being indexed. For 100,000 documents, you can probably get away with 2 TB of SSD storage, and 4 TB of mechanical storage for the crawl data.

        I don't know how far 100k documents gets you. It doesn't sound like much if you are going for the whole internet but if you are curating a more narrow subset it could be enough.

        This page has their philosophy and towards the bottom of the page, links to similar projects.

        • thethirdgracchi [he/him, they/them]
          ·
          5 months ago

          I just use their search page, not trying to run it on my own. But yeah it's actually like possible for you to run, as opposed to something like Google.