AINEWS
Reddit, Reuters, and USA Today are now weighing a Google AI opt out. Each block could also cost a publisher its place in search results.
Reddit signed a $60 million deal with Google in 2024 to license its posts for AI training. A year later, the company is weighing the opposite move, and in doing so it has named what every publisher is now running the math on: the cost of being findable on the open web.
According to the Wall Street Journal, via a Gizmodo re-report, Reuters, USA Today, People, Politico, and the Economist are all openly weighing what used to be unthinkable: a future in which their content does not feed Google’s AI at all. The lever is the same set of crawler tokens every site already uses in robots.txt. The cost is new.
As one expert vendor, BotDetector, laid out for the Journal, Google’s Google-Extended token tells the company’s AI training pipelines, including the Gemini family, to keep out. It doesn’t stop AI Overviews, the answer boxes that have been absorbing search clicks for two years. The nosnippet tag is more aggressive: it can strip a publisher out of Overviews entirely, but it also hides them from regular search snippets. The noindex tag is the nuclear option. It blocks indexing at all.
The bind, as the Journal’s reporting lays it out, is structural: publishers who want to keep their place in regular search results have, in practice, been letting Google index them for AI Overviews too. The two surfaces are stacked. You can opt out of training, but you can’t easily opt out of answers without giving up the click traffic that pays for the newsroom.
The trade-off publishers are now running in public is real, and it is starting with named executives. Reddit’s 2024 deal was a paid license; an opt-out would be a clean reversal. Reuters President Josh London told the Journal the company is “certainly looking at the economic trade-offs between search and AI summaries.” USA Today CEO Mike Reed went further: “It’s time to take a stand and say enough is enough.” People Inc. CEO Neil Vogel was blunter still: “Turning them off and blocking them entirely is 100% on the table.” Each of those statements was carefully weighted. None of them commit to the cut.
The driver is the search drought. AI Overviews and AI answers keep more clicks inside Google’s own surfaces, and publishers describe a thinned stream of referral traffic as the answer boxes absorb the queries that used to send readers downstream. A Google spokesperson pushed back, telling the Journal that its AI sends “billions of clicks to the web every week” and “highlight[s] links to the web and help[s] creators and publishers.” That line lands as a tension beat, not a rebuttal. The figure is self-reported; no third-party measurement is attached.
What is being born is not a boycott. It is a market. The unit of account for AI-era content was the click; now it is also the training corpus, and the two carry very different prices. Reddit priced the corpus at $60 million a year. Google prices the click at “billions,” with no dollar figure attached. Publishers are testing whether they can sell the same content twice, or whether the second sale costs them the first.
The next move belongs to the publishers who say they’re ready to pull the trigger. If any of them actually does, the contract on the table is no longer a scrape license. It is a re-priced access deal, in which search visibility is something publishers get paid for, not something they assume is free.