Jessica Jordan

Our team loves to build things, and we’re always thinking about ways to improve how people get their information, while making sure the people responsible for creating that content are fairly compensated. That’s a really difficult challenge, as search traffic collapses and people become more reliant on AI summaries, not publishers. Our big question: What if we could use the agentic web to support both readers and publishers?

Our first step toward that is a small, invite-only beta of Flipboard Edition at edition.flipboard.com.

It does one thing: you tell it what you care about and connect the sites, newsletters, YouTube channels and podcasts you already follow. Once a day, it brings all that together into one Edition of articles, video and audio, laid out to be read.

You can ask about any story, check a claim or use plain words to fine-tune what you see, and the changes will be reflected in your next Edition. Read it on the web, in your inbox or inside Claude.

What matters most: Your Edition is personalized and it has an end. The problem of “keeping up” is often solved by a never-ending feed that's overwhelming and impossible to get through. Your Edition is finishable, and every item traces back to a link you can open, tied to a topic or source you care about.

It is a beta, in the real sense. We're keeping it small on purpose, to find out what breaks and where we can improve. You can send us your feedback at edition-beta@flipboard.com or click the feedback button inside the Edition.

Where We Stand

We do not train AI models on anyone's content, and nothing that passes through this product is used to train a model.

Every story links out to the publisher or creator who did the work. That is how Flipboard has always worked and nothing here changes it.

When AI surfaces a publisher’s or creator’s work in response to a question, the publisher or creator should be rewarded.

Answers or claim checks that include a publisher’s or creator’s content draw credits. A share of every credit goes into a pool, launching post-beta, for the publishers or creators cited in AI answers.

Questions and Answers

What is Flipboard Edition?

It’s a personalized collection of articles, videos and/or podcasts sent daily, built from the sources you choose. This is a continuation of what Flipboard has always been about: creating a personalized way to discover and consume content, now reimagined for what’s possible on the agentic web.

Where does the content come from?

There are three source types. Flipboard Magazines and topics. Sources you've added, like newsletters, podcasts, YouTube channels and Bluesky profiles you follow. And Surf, which aggregates stories from across the open social web. If you connect your Flipboard account, we read your interests once to set up your Edition, then we discard the token.

Do you copy full articles?

No. To build the Edition, we use what the publisher already includes in its own feed: headline, byline, image, topic tags and excerpt. That's what the model works from when deciding what goes where and generating a short line about it. The article body itself is scanned in only two cases: generating a summary or checking a claim. Both of these are triggered by a reader asking about a specific story.

Do links go to the publisher or creator?

Yes. Every story in an Edition links to the original URL on the original domain, with attribution parameters intact, and it opens on the publisher's or creator’s site. Reading the Edition and clicking through is the intended path.

What about paywalled stories?

We do not unlock anything. When a story links to a publisher or creator we know runs a paywall, the card says “Subscription may be required,” so the reader is not surprised. The link goes to the publisher's page, where its business model applies exactly as it would normally.

Are you training AI models on our content?

No. We do not train models, we do not fine-tune models and we do not contribute publisher or creator content to anyone else's training. We use Anthropic's models through its commercial API to compose the Edition and answer questions. Nothing that passes through is training data for us or for them.

Then what is the AI actually doing?

Compiling, not reporting. The ranking is based on how recent a story is, what you have told us you want to see and what you have opened or muted before. The model takes that ranked pool and picks, orders and generates the cover line and a couple of sentences underneath. Its instruction is literal: reference only the stories it was served, never invent content, URLs or titles. It does not write news and it does not paraphrase articles not featured in the edition.

Can it get things wrong?

Yes, and we recommend double-checking responses and reading the links to verify any claims.

How do publishers and creators get compensated?

Asking a question that searches beyond your Edition, summarizing an article or checking a claim costs credits. Half of every credit spent on those goes into a pool for the publishers or creators cited in that answer, split among them and recorded. The other half covers running the search. That accounting runs in production today.

We have yet to settle how or when money leaves that pool, as we're just getting started with the beta. We built the accounting first, on purpose, so that when we do settle it, there's a record reaching back to the first day, rather than a promise starting from whenever we get around to it.

Do publishers and creators get paid for appearing in the edition?

No, and that is deliberate. Appearing in the Edition is how Flipboard.com already works. The publisher/creator gets the placement and the click, on their page, with their ads and subscription prompt in front of their reader. The pool exists for a case that is genuinely new, where AI uses your content to answer a question and the reader may never leave our page.

Why charge readers at all?

Because a searched answer costs real money to produce, and because a product the reader pays for is a product that works for the reader. It's free to create and receive your Edition, tune it and get questions answered about what's in it. Searching, summarizing, checking a claim and rebuilding your edition on demand requires credits. You start with credits and you earn more by reading or taking actions. During the beta, nobody pays for any of it.

Do you crawl our site?

We read your public feed the same way any feed reader reads. We request the article page itself in two situations: to pick up the social share image when your feed does not carry one and when a reader asks for a summary or a claim check of that story. We do not go behind paywalls, we do not use credentials and we do not bulk-download.

What about private posts?

We read public feeds and public APIs. Bluesky is read through the public AppView with no authentication. Surf surfaces what is already public. Sources a reader adds are fetched with no credentials at all, so anything that needs a login simply does not load.

What do you collect about readers?

Account details, which Editions and items you open, email opens and clicks, IP and device data, and your credit balance, used to compose and improve your edition. Analytics run only if you consent. We do not sell personal information. Card details never touch our systems. The full detail, including our subprocessors, is in the privacy policy.

How does this relate to Flipboard.com and Surf?

Flipboard is unchanged. Surf is unchanged. Flipboard Edition is a separate beta on its own domain, drawing on Flipboard's public Topic and Magazine feeds and Surf's aggregation of the open social web. There is no single sign-on with your Flipboard account yet, and the Magazines you made or follow do not come across. Those are questions for later, if this works.

How do I get in?

By invitation for now. If you are a publisher or creator and you want to talk about how your work shows up in it, or you want it not to, write to us at edition-beta@flipboard.com.