Jun 13, 2026 · 5 min read
How much is your content worth to AI?
A practical guide to what AI companies actually pay for newsletters, podcasts, video archives, and blogs — with real comparables, the math, and a free tool to get a price range for your own work in under a minute.
If you write a newsletter, run a podcast, post videos on YouTube, or publish independently on the web, this is the question worth answering right now: what is your work actually worth to an AI lab?
Most creators have no idea. The deals are private. The numbers in the press are rounded to the nearest absurd-sounding billion. And the only "valuation" most creators have ever seen is the ad-revenue split their platform pays them — which has nothing to do with the price an AI buyer would pay for the back catalogue their model needs.
So here's the actual answer, with the math, the comparables, and a tool you can run in under a minute that gives you a real number for your own work.
The short answer
For an independent creator with a real, owned archive — somewhere between $1,200 and $245,000 per licence. The range is that wide because the inputs vary by an order of magnitude. The four factors that move it:
- Size — items in the archive (articles, episodes, videos). More items, more raw material for training, higher price.
- Age — how many years of consistent output. A two-decade archive is rarer than a two-year one, and rarity is a multiplier.
- Content type — long-form text and transcribed audio command the highest per-item rates. Short-form social posts the lowest. Video sits in between, depending on whether the buyer wants the visuals or the transcripts.
- Whether AI has already taken it. This is the counter-intuitive one. If your work is already in Common Crawl (the giant public web crawl most major models train on), the marginal value to a buyer goes down — they can already get it for free. If you've kept it behind a paywall or never indexed it, your work is scarce training data and the price goes up. We have a free check that tells you which side you're on.
Combine those four numbers and you get a price range. That's what our valuation tool does in 30 seconds.
Real comparables — what AI buyers are actually paying
The deals you've read about — Reddit's $60M / year with Google, News Corp's $250M / 5 years with OpenAI, Stack Overflow's reported low-eight-figures with Google — are at one end of the spectrum. Those are mega-licences for top-tier institutional corpora.
What we care about for an independent creator is the per-item, per-year implied rate. Strip out the press-release theatre and the rates start to look like this:
| Archive shape | Implied price | What it looks like |
|---|---|---|
| One year of a small newsletter | ~$1,200 | 50 long-form posts |
| Five years of a niche podcast | ~$4,500 | 200–300 episodes with decent transcripts |
| A focused independent tech blog with traction | ~$9,200 | 600–800 posts, half a million words |
| Five years of a video creator | ~$18,500 | A few hundred long-form videos, transcripts included |
| A long-running essay site | ~$32,000 | A decade of writing on one beat |
| A regional reporting archive | ~$48,000 | Local investigative work that doesn't show up in generic corpora |
| A specialist publication (trade, academic) | ~$75,000 | Domain-specific content, the kind of dataset model labs actively hunt for |
| Ten years of investigative reporting | ~$124,000 | The thing newsroom-style models genuinely need |
| A podcast network's full back catalogue | ~$182,000 | Multi-show, decade-plus of broadcast-quality transcripts |
| Two decades of independent publishing | ~$245,000 | The whole canon, single owner, clean rights |
These are mid-points of a range. The actual price for a given archive can sit anywhere within a wider band. The point is — none of these numbers are zero, and none of them are the rounding-error figures platforms have trained creators to accept.
How the math works (in one paragraph)
Every per-item rate looks something like this: start with a baseline price per item by content type (long-form text: highest; podcast transcripts: high; short video: medium; image: low; tweet-length text: tiny). Multiply by an age factor — older, recency-balanced archives are more valuable than a single year's output. Multiply by an exclusivity factor — non-exclusive licences are the floor; exclusive (only one AI buyer ever) doubles or triples it. Subtract a Common-Crawl discount if your content is already public-web-indexed. Add a domain premium if your subject matter is hard to find (specialist, regional, or paywalled).
Our /estimate tool runs that calculation for you with real, recent comp data baked in. It also runs the Common Crawl check automatically if you give it a domain.
Why "I don't have a wallet, I don't sell to AI" isn't an answer
The whole reason this market is forming right now — and prices are still high — is that AI buyers have a regulator-shaped urgency to prove the data is licensed. The EU AI Act, the New York Times copyright case, the Anthropic class action, the OpenAI inquiries: they all turn on one question — can the lab show that the data it trained on was acquired lawfully? For everything they scraped before 2023, the answer is increasingly "no." For everything they license now, the answer is "yes."
So the buyer demand isn't going away. The window where independent creators have leverage — small enough catalogues that big-platform deals don't satisfy them — is now. In two years it may close.
Three things you can do in the next ten minutes
- Run the free valuation. Plug in your archive size, the years you've been making it, and the type. You'll get a price range and the reasons behind it. No sign-up, no email.
- Run the scrape check on your domain. If your work is already in Common Crawl, you know which lever to pull (move to mainnet-licensed access, raise the price for non-Common-Crawl content, etc.). If it isn't, your archive is scarcer than you thought.
- If both numbers look interesting, list it. Connect a Substack / Beehiiv / Ghost / YouTube / your own site; we verify ownership, you set the price, and AI buyers see the listing.
The single most expensive mistake an independent creator can make in 2026 is to assume the back catalogue is worth nothing because no one ever offered. The market only knows what you list.
Get your number. Run the free valuation — takes 30 seconds, no sign-up, real comp data.