The Pricing Question Nobody Has Answered Cleanly
Until 2025 there was no observable price for an AI citation. AI labs ingested publisher content, AI assistants cited that content in answers, and no money moved at the per-citation level. The 2025 launch of Cloudflare Pay-Per-Crawl, the maturation of TollBit and ProRata as content marketplaces, and the addition of ScalePost as a publisher-side aggregator changed that. By April 2026 there are four observable price surfaces for the value of an AI citation, and they tell a partly consistent, partly fragmented story.
This page synthesises the observable pricing from those surfaces, augments it with bilateral-licensing implied per-citation rates where those can be inferred, and produces a defensible answer to the question "what is an AI citation actually worth in 2026."
Per-Fetch Prices Across the Four Major Surfaces
The per-fetch price (what a single AI crawler pays to fetch a single URL) is the most directly observable rate. The table below summarises typical April 2026 ranges across the four major surfaces. Ranges are reported because pricing varies by publisher tier, content type, and bot identity.
| Surface | Typical per-fetch (USD) | How it is set |
|---|---|---|
| Cloudflare Pay-Per-Crawl | $0.0005 to $0.05 | Publisher sets; Cloudflare merchant of record |
| TollBit | $0.001 to $0.10 | Publisher sets; per-URL granularity available |
| ProRata | $0.002 to $0.20 | Publisher sets; ProRata recommends bands by vertical |
| ScalePost | $0.001 to $0.05 | Aggregator-mediated; per-section pricing |
The ranges overlap meaningfully. The midpoint of all four surfaces lands at roughly $0.01 per fetch for general content. Premium news, primary research, and proprietary datasets price meaningfully higher. Generic blog content prices at the low end. Below $0.001 per fetch the operational overhead of settlement starts to dominate the revenue, so most publishers do not price below that floor.
Per-Citation Implied Prices
A citation in an AI-generated answer is the downstream outcome of one or many fetches. The implied per-citation price is the per-fetch price divided by the citation rate (the fraction of fetches that translate to an actual citation in a user-facing answer). Citation rates vary widely by platform: Perplexity cites at the highest rate (a meaningful share of fetches translate to visible citations), ChatGPT search lower, ChatGPT main lower still, Claude lower still in published answers (though Claude tends to ground internally without surfacing every source).
Assuming typical citation rates by platform, the implied per-citation price for general content is roughly $0.05 to $0.50 in 2026, with premium news content reaching $1 to $5 per citation through marketplace settlement, and bilateral licensing implying meaningfully higher per-citation rates for high-tier news and primary research (often into the low double digits per citation when total deal value is divided by total cited volume).
The Bilateral Licensing Layer
The known bilateral deals (NYT/OpenAI, Reuters/Meta, Reddit/Google, Axel Springer/OpenAI, News Corp/OpenAI, AP/OpenAI, Vox Media/OpenAI, Le Monde/OpenAI, the various FT and Bloomberg arrangements) imply much higher per-citation rates when reduced to a per-citation basis. The reason is that these deals carry large fixed-fee components covering training-data rights, data-feed access, and other deliverables that go beyond per-citation pay-per-crawl. Comparing bilateral implied rates to marketplace per-citation rates is comparing different products.
For the purpose of pricing your own content, the marketplace rates are the appropriate benchmark unless you can credibly negotiate a bilateral deal. Bilateral deal eligibility starts roughly at major-publisher scale and authoritative-content profile; below that scale, marketplace rates are the realistic ceiling.
Price Variance by Vertical
Vertical-level price variance is significant. Primary research, financial data, and legal precedents price highest, often 5-10x the general content baseline because the content is hard to substitute. News content prices roughly 2-3x general baseline driven by recency and editorial authority. Encyclopedic and reference content prices below general baseline because Wikipedia and similar sources are widely available substitutes. Branded marketing content typically does not transact at all because AI labs prefer neutral sources.
How Citation Value Score Anchors These Prices
Citation Value Score (CVS) is the methodology Presenc AI uses to anchor per-page and per-publisher pricing in defensible signals. CVS combines crawl activity, content quality, authority, and outcomes into a composite score that correlates with marketplace pricing where comparable transactions are observable. Publishers using CVS-anchored pricing tend to price above the marketplace median for content that scores high on the methodology and below median for content that scores low, with the result that CVS-priced inventory tends to attract more marketplace transactions than category-mean pricing.
Methodology
Pricing data is compiled from publicly disclosed marketplace ranges, publisher case-study disclosures, and observable transactions where Presenc AI customers have given permission for inclusion. Bilateral deal implied rates are calculated from publicly disclosed total-deal-value figures divided by published or estimated cited-volume figures. All ranges are reported as USD-equivalent at April 2026 exchange rates. The page is updated quarterly. Last update: April 30, 2026.