Last updated: August 3, 2026
When people ask “does Google AI use Reddit?” the short answer is yes, and there is a signed contract behind it. But the fuller picture is different from the OpenAI licensing story. OpenAI had to go out and buy access to a lot of content because it did not run a web crawler at the scale of Google Search. Google already indexes most of the open web, so it has signed comparatively few explicit publisher licensing deals for Gemini training data. The deals it has signed are notable precisely because they are exceptions.
Table of Contents
This post maps every verified Google and Alphabet content or data deal that feeds its AI products, separates signed agreements from talks, and explains the tension at the center of it all: publishers say AI Overviews are eating the referral traffic they depend on, and some are now suing or seeking regulation instead of signing. For the parallel picture on the other side of the industry, see our OpenAI publisher deals map and the broader AI content licensing hub.
Why Google has fewer licensing deals than OpenAI
The single most important fact about Google’s AI data strategy is that Google is a search engine first. Googlebot has crawled and indexed the open web for more than two decades, and Google’s public position is that content already accessible to Search can inform its AI features unless a publisher opts out. That is a very different starting point from a lab that has to license or scrape a corpus from scratch.
The practical result: Google signs a licensing deal only when it wants something the open crawl cannot easily give it. That tends to be one of two things.
- Structured, real-time access at scale. Reddit’s firehose of fresh community posts is far more useful through a licensed real-time data API than through ordinary crawling, which is throttled and delayed.
- Fresh, authoritative, time-sensitive facts. A wire service like the Associated Press produces breaking news that Google wants inside the Gemini app quickly and with a reliable source.
For most other content, Google’s stance has been that it does not need to pay, and reporting through 2025 described Google as taking a tougher line than OpenAI in licensing conversations with news publishers. That posture is the root of the antitrust and regulatory fights covered later in this post.
Every confirmed Google and Alphabet AI content deal
The table below lists the agreements that are publicly confirmed to feed Google’s AI products or model training. Figures labeled “reported” come from press reporting rather than official disclosure, and “undisclosed” means the parties did not release terms.
| Partner | Announced | Scope | Reported value |
|---|---|---|---|
| Shutterstock | 2023 onward | Licensed images, video, and metadata as AI training data (Alphabet named among Shutterstock’s AI data customers) | Undisclosed (industry deals reported at 25 to 50 million dollars each) |
| February 2024 | Licensed access to Reddit’s real-time Data API for model training and richer Reddit displays in Google products; Reddit adopts Google Cloud Vertex AI | Reported ~60 million dollars per year | |
| Associated Press | January 2025 | Real-time news feed piped into the Gemini app to improve up-to-date answers | Undisclosed |
| Reddit (expanded) | April 2025 | Reddit Answers built on Gemini via Vertex AI; deeper product integration on top of the 2024 license | Undisclosed |
Reddit: the flagship deal (February 2024)
In February 2024, Google and Reddit announced an expanded partnership. Reddit disclosed in its IPO paperwork that the arrangement was worth a reported 60 million dollars a year. The deal gives Google licensed access to Reddit’s Data API, which delivers structured, real-time, human-written content from Reddit’s communities. Reddit’s SEC filings describe AI licensing customers getting “continuous access” to the data API plus periodic bulk transfers of Reddit data over the life of the deal, which matters because Reddit content constantly grows as users post and comment.
Two things flow from this. First, Google can use the licensed content to help train and ground its models. Second, Google surfaces Reddit more prominently across its products. Reddit, in turn, moved onto Google Cloud’s Vertex AI platform. If you have noticed how often Reddit threads appear in Google results and AI answers, licensed access may contribute to availability, but the citation frequency does not prove a single cause: independent analyses through mid-2025 found Reddit was the single most-cited domain in Google’s AI Overviews.
The Reddit deal is also the clearest answer to the keyword question “google reddit deal”: it is real, it is signed, it is reportedly around 60 million dollars a year, and it grants Google licensed real-time data access rather than a one-time content dump.
Shutterstock: the stock media pipeline (2023 onward)
Shutterstock runs a large AI data-licensing business, which it has said generated more than 100 million dollars in 2023, and it has publicly named Alphabet (Google) among the big technology companies buying that training data, alongside Meta, Amazon, and Apple. Individual big-tech Shutterstock deals have been reported in the range of 25 to 50 million dollars, but Google’s specific terms were not disclosed. This is a training-data supply relationship for images and video rather than a real-time editorial feed, so it sits in a different bucket from the Reddit and AP deals, but it is a confirmed licensed content source.
Associated Press: real-time news for Gemini (January 2025)
On January 15, 2025, Google announced a deal with the Associated Press to pipe a feed of real-time information into the Gemini app. Google’s VP of global news partnerships, Jaffer Zaidi, framed it as helping users looking for up-to-date information. Financial terms were not disclosed, and Google did not spell out whether Gemini would attribute AP content or link back to source articles.
An important nuance: AP has supplied Google with news for years, so this is not the first time the two have worked together. What is new is routing an AP real-time feed directly into the Gemini chatbot experience. It is a narrower, feed-specific arrangement, not a sweeping training-rights license like some of the OpenAI news deals. For how the OpenAI news partnerships compare, see our roundup of AI content deals beyond OpenAI.
The 2025 Reddit expansion and renegotiation
At Google Cloud Next in April 2025, Reddit and Google announced that Reddit Answers, Reddit’s own conversational search feature, would run on Gemini via Vertex AI, with summaries citing existing posts and comments. That is a product integration layered on top of the original license.
Separately, reporting from September 2025 said Reddit was renegotiating its Google deal to move away from a flat annual fee toward “dynamic pricing” that rises as Reddit’s content becomes more central to AI answers, and to tie the arrangement to sending traffic back to Reddit communities. As of July 2026 those renegotiated terms have not been publicly confirmed, so treat the dynamic-pricing structure as talks, not a signed change.
Deals that were only talks, and content that is not licensed
Being honest about what is not signed is as important as the confirmed list.
- News publisher licensing talks (July 2025). Google reportedly opened exploratory AI content-licensing discussions with around 20 national news publishers in mid-2025. Reporting described Google as taking a hard line, and no broad slate of signed news licenses has been announced. Treat this as talks.
- Reddit dynamic-pricing renegotiation (2025). Discussed above: proposed, not confirmed as signed.
- The open web generally. The vast majority of what informs Gemini, Google AI Mode, and AI Overviews is crawled and indexed content, not licensed content. This is the core reason the licensed list is short. Google’s argument is that indexed content is fair to summarize; many publishers disagree.
- Automattic (WordPress and Tumblr) data. This frequently comes up in AI-data discussions, but the reported Automattic data deal was with other AI companies, not Google, so it does not belong on Google’s list. We flag it only to prevent confusion.
AI Overviews and the publisher traffic problem
The reason Google’s licensing list stays short while tension keeps rising is AI Overviews. When Google answers a query directly at the top of the results page, users click through to sources far less often, and that hits the ad and subscription revenue publishers depend on.
The Pew Research Center studied roughly 68,879 Google searches from about 900 US adults in March 2025. Its findings, published in July 2025:
- Users clicked a result 8 percent of the time when an AI Overview appeared, versus 15 percent when it did not.
- They clicked a source link inside an AI Overview only about 1 percent of the time.
- Around 18 percent of searches produced an AI Overview, and the most-cited sources were Wikipedia, YouTube, and Reddit.
Other datasets point the same direction. Digital Content Next reported meaningful year-over-year declines in Google referral traffic to news and non-news brands, and an Ahrefs analysis published in early 2026 found AI Overviews correlated with a large reduction in click-through rates for top-ranking pages. The exact percentages vary by methodology, but the trend is consistent: AI Overviews reduce the clicks that used to flow to publishers.
This is the crux of the conflict. Because Google already crawls the web, it does not need to license most content to build AI Overviews. Publishers argue that using their journalism to answer queries in place of a click, without paying, is different from traditional search that sent them traffic.
Lawsuits and regulation instead of licenses
With few licenses on offer, some publishers turned to courts and regulators.
- Penske Media Corporation v. Google (September 2025). Penske, the owner of Rolling Stone, Billboard, and The Hollywood Reporter, sued Google in federal court in Washington, DC. Rather than a copyright claim, Penske used an antitrust theory: it alleges Google effectively forces publishers to allow their content in AI Overviews or accept lower Search rankings, and that the AI summaries unfairly compete by answering queries so users do not click through. It is widely described as the first time a major US publisher took Google to court over AI search. Google moved to dismiss the case in early 2026 and says AI Overviews send traffic to a wider range of sites.
- US antitrust remedies (September 2025). In the separate DOJ search-monopoly case, Judge Amit Mehta issued remedies in September 2025 that stopped short of a breakup but barred certain exclusive distribution deals and required Google to share portions of its search index and user-interaction data with qualified competitors. Google and the DOJ both filed appeals in early 2026, so the outcome is not final.
- UK CMA opt-out (June 2026). In what it called a world-first, the UK’s Competition and Markets Authority secured a requirement that Google let publishers opt out of having their content used to power AI features in Search, and barred Google from penalizing opted-out publishers in rankings. Google was given a window to implement the controls.
Put together, the picture is a company that prefers indexing to licensing, a set of publishers that increasingly want to be paid or to opt out, and regulators starting to force the question. For the wider legal picture across the industry, see our overview of AI copyright lawsuits.
What this means for your AI visibility
The strategic takeaway for brands and publishers is simple: which sources Google’s AI trusts and cites is now a distribution channel in its own right. Licensed access can make partner content available, but does not guarantee citations, and and the open-web sources that AI Overviews and Gemini choose to cite capture the attention that used to arrive as an organic click. If your brand is not among the cited sources, an AI answer can satisfy the user without ever sending them to you.
That is exactly what optimizing for Google AI Overviews is about, and it is why measuring your presence in AI answers matters. This is what LLM Pulse is built for. LLM Pulse tracks your brand’s visibility and the exact sources cited across Gemini, Google AI Mode, and Google AI Overviews (plus ChatGPT and Perplexity), so you can see whether AI is mentioning you, which competitors it favors, and which URLs it pulls into its answers. You get share of voice, citation tracking, and sentiment on the same prompts week over week, which turns the abstract “AI Overviews changed my traffic” worry into a concrete list of prompts, mentions, and sources you can act on.
Because so much of what Google’s AI surfaces comes from the crawled open web rather than paid licenses, earning citations is still very much within reach: it comes down to being the clear, trustworthy, well-structured source that AI wants to quote. Tracking where you stand today is the first step. To understand the raw materials feeding these systems, our explainer on where ChatGPT gets its data is a useful companion read.
Summary
Google’s licensed content list is short by design. The confirmed deals feeding its AI products are Reddit (reported around 60 million dollars a year, February 2024), Shutterstock stock media (2023 onward, terms undisclosed), and the Associated Press real-time news feed for Gemini (January 2025, terms undisclosed), plus the 2025 Reddit Answers integration. Broad news-publisher licensing was still in talks as of mid-2025, and the defining dynamic is the fight over AI Overviews reducing publisher referral traffic, now playing out in the Penske antitrust suit, the DOJ remedies, and the UK CMA opt-out. Because Google leans on crawling rather than licensing, whether AI cites your brand is something you can influence and should measure.
FAQ
Does Google AI use Reddit?
Yes. Google signed a content-licensing deal with Reddit in February 2024, reported to be worth around 60 million dollars a year, that gives Google licensed access to Reddit’s real-time Data API for model training and for surfacing Reddit content in its products. Reddit is consistently one of the most-cited domains in Google’s AI Overviews.
How much did the Google Reddit deal cost?
Reddit’s IPO disclosures and press reporting put the arrangement at a reported 60 million dollars per year. Google did not officially publish the figure, so it is best described as reported rather than confirmed by Google. Reddit was reportedly renegotiating toward a usage-based “dynamic pricing” model in 2025, but revised terms have not been confirmed.
What data does Gemini train on?
Gemini draws on Google’s own large-scale web crawl and index plus licensed sources where Google has deals, most notably Reddit’s real-time data and Shutterstock media. The Associated Press supplies a real-time news feed into the Gemini app. Most of what informs Gemini comes from indexed open-web content rather than paid licenses, which is why Google has far fewer publisher licensing deals than OpenAI.
Does Google have a deal with the Associated Press?
Yes. In January 2025 Google announced a deal to bring an Associated Press real-time news feed into the Gemini app to improve up-to-date answers. Financial terms were not disclosed. AP had already supplied Google with news for years, so this specifically extends the relationship into the Gemini chatbot.
Why does Google have fewer AI licensing deals than OpenAI?
Because Google already crawls and indexes the open web with Googlebot, it does not need to license most content to feed its AI features. It licenses only when it wants something crawling cannot easily provide, such as Reddit’s structured real-time firehose or a wire service’s breaking news. OpenAI, without a comparable crawler, signed many more content deals to build its corpus.
Do AI Overviews reduce traffic to publishers?
The evidence points that way. A Pew Research study of March 2025 searches found users clicked a result 8 percent of the time when an AI Overview appeared versus 15 percent without one, and clicked a source link inside an Overview only about 1 percent of the time. Publishers including Penske Media have cited this dynamic in litigation, and the UK CMA now requires Google to let publishers opt out of AI features.
Is Google being sued over AI Overviews?
Yes. In September 2025 Penske Media, the owner of Rolling Stone, Billboard, and The Hollywood Reporter, sued Google in federal court in Washington, DC, using an antitrust theory that Google forces publishers to allow their content in AI Overviews or lose ranking. Google moved to dismiss the suit in early 2026 and argues AI Overviews send traffic to a wider variety of sites.
How can I tell if Google’s AI is citing my brand?
You need a tool that runs real prompts against Google’s AI surfaces and records the answers and their sources. LLM Pulse tracks brand mentions, share of voice, exact citations, and sentiment across Gemini, Google AI Mode, and Google AI Overviews, so you can see whether you are being cited, which competitors are, and which pages the AI is pulling from, then work to improve it.
