What is it
Per-Page Pricing is a billing unit where customers are charged per page crawled, parsed, or rendered — the meter for web scraping and document parsing. The page is where web-data infrastructure and document AI converge on the same noun: Firecrawl meters pages scraped into clean markdown, LlamaIndex’s LlamaParse meters PDF pages parsed into structured output, Mistral’s OCR bills $2 per 1,000 pages read, and Exa and You.com both sell full-page content retrieval at $1 per 1,000 pages beside their per-call search APIs.
The unit’s appeal is the same as the document’s: a buyer with a URL list or a PDF folder can count their job before running it. Its weakness is that “a page” abstracts over enormous compute variance — a static blog post and a scanned 600-dpi financial table are the same unit and very different work — which is why most of this cohort prices modes on top of the page count rather than one blended rate.
The SEO tools carry the unit in subscription form: Frase bundles 50 to 1,000 audit pages per month into its tiers, and Scalenut packages page-level optimization quotas the same way — pages as a monthly allowance rather than a metered API line.
How it works
The metered cluster prices pages × mode rate, usually denominated in credits:
| Job | Vendor & rate | Mechanics |
|---|---|---|
| Plain content fetch | Exa Contents $1/1k pages; You.com Livecrawl $1/1k | Known URLs in, text/markdown out |
| Scrape / crawl | Firecrawl 1 credit/page (Scrape, Crawl, Map, Monitor) | Hobby $16/5k credits → Standard $83/100k → Scale $599/1M |
| OCR | Mistral $2/1k pages ($3 with annotations) | Flat rate, no modes |
| Structured parse | LlamaParse 1 / 3 / 10 / 45 credits per page (1k credits = $1.25) | Fast → Cost-effective → Agentic → Agentic Plus |
| Extraction add-ons | LlamaParse +5–15 credits/page; layout +3 | Effort stacks on the base parse |
| Page audits (quota) | Frase 50–1,000 pages/mo by tier | Allowance inside the subscription, no overage rate |
Worked example — the same 100,000 pages, three jobs. Scraping 100k web pages on Firecrawl consumes 100k credits — exactly the Standard tier’s allotment, $83. Running 100k scanned pages through Mistral OCR costs $200. Parsing 100k PDF pages on LlamaIndex’s LlamaParse runs $125 on Fast (1 credit/page) but $5,625 on Agentic Plus (45 credits/page) — a 45x spread on the identical page count, set entirely by how much reasoning the parse applies. The page is the unit; the mode is the price.
Worked example — endpoint drift. Firecrawl’s “1 credit = 1 page” holds for Scrape, Crawl, Map, and Monitor — but Search bills 2 credits per 10 results and Interact bills 2 credits per browser-minute. A workload that mixes endpoints can’t be estimated from page count alone; the usage-metric guide calls this unit drift, and it’s the first thing to audit on any credit-denominated rate card.
Companies using this
7 in-corpus companies meter pages: the API cluster (Firecrawl, LlamaIndex, Mistral, Exa, You.com) billing pages scraped, parsed, OCR’d, or fetched, and the SEO pair (Frase, Scalenut) bundling audit-page quotas into subscription tiers.
Patterns observed
The commodity floor is real and public: plain page retrieval settled at $1 per 1,000 pages at both Exa and You.com, with Mistral OCR only 2x above it — undifferentiated page-touching is priced like a utility. Differentiation, and margin, live in the mode ladder: LlamaIndex spans 1 to 45 credits per page by parse intelligence, and Firecrawl’s historical LLM-extraction endpoint ran 50 credits against the basic scrape’s 1. The pattern mirrors the effort-tier structure that search APIs use on calls — same commodity base, same compute-priced premium ladder.
Credits are the standard wrapper. Firecrawl and LlamaParse both denominate pages in credits with tier-sized monthly pools (and Firecrawl lets buyers nudge the pool size per tier), which buys the vendor repricing flexibility — LlamaParse’s v2 collapsed a 1–90-credit maze into four flat tiers, cutting its top agent mode 50% without touching the dollar price of a credit.
Counterexamples & variants
Frase and Scalenut show the unit defanged: pages appear as audit quotas (50/mo on Frase Starter, 1,000 on Scale) inside flat subscriptions, with no per-page overage rate published — the page counts gate the tier you need rather than metering a bill, which makes them packaging, not pricing. Mistral is the single-rate counterexample to the mode ladder: OCR is $2/1k pages flat (annotations aside), betting that one honest blended rate beats a four-mode menu for a well-bounded task. And Exa keeps the page subordinate: its core products bill per request ($7/1k Search, $12–15/1k Deep Search) with pages only metering the Contents side — a reminder that for search-shaped products the query, not the page, is where the value concentrates.
What this means for buyers vs vendors
For buyers
Price the mode, not the page: get a representative sample of your documents and run them through the cheapest tier that passes quality, because the spread between modes (1–45 credits on LlamaParse) dwarfs the spread between vendors at any single mode. On crawl workloads, ask what bills on failure — redirects, soft-404s, and bot walls are a real fraction of any large URL list — and watch endpoint drift on credit systems where “1 credit = 1 page” carries footnotes. For bundled audit quotas (Frase, Scalenut), translate the tier into an effective per-page price at your actual monthly volume before comparing against the metered APIs.
For vendors
Publish the mode ladder and make each rung’s quality difference demonstrable — the category’s repricing history (LlamaParse v1’s 90-credit maze → v2’s four tiers) shows buyers reward legible effort pricing and punish mode mazes. Keep the commodity rung at the market floor ($1–2/1k) as the funnel, take margin on intelligence above it, and if you wrap pages in credits, hold the page-to-credit exchange rate stable per endpoint — every footnote on “1 credit = 1 page” is forecast error you’re exporting to the buyer, and the credit-model guide covers why that trust erosion compounds.
| Company | Product | Pricing model | Billing units | Free tier | Verified |
|---|---|---|---|---|---|
| Exa | AI web search API for agents — search, contents, deep research, and monitoring endpoints billed per request | Yes | 2026-06-01 | ||
| Firecrawl | Web-scraping and data-extraction API for AI agents — scrape, crawl, map, search, and extract pages into clean markdown/JSON | Yes | 2026-06-02 | ||
| Frase | Agentic SEO and GEO platform that researches, writes, optimizes, and tracks AI-search visibility for content teams. | No | 2026-06-07 | ||
| LlamaIndex | RAG/agent orchestration framework + LlamaCloud document parsing | Yes | 2026-06-10 | ||
| Mistral AI | Open and commercial LLM APIs | Yes | 2026-05-31 | ||
| Scalenut | AI search visibility (GEO) and SEO content platform — tracks brand presence in AI answers and generates ready-to-rank content | No | 2026-06-07 | ||
| You.com | Web search, contents, research, and finance-research APIs for AI systems | Yes | 2026-06-01 |
FAQ
What is per-page pricing?
Per-page pricing is a billing unit where customers are charged per page crawled, parsed, or rendered — the meter for web scraping (Firecrawl), document parsing (LlamaIndex's LlamaParse), OCR (Mistral), and content-fetch APIs (Exa, You.com). One page is one unit, regardless of what it took to process.
How much does it cost to scrape or parse 1,000 pages?
It depends on the job: plain content fetch runs $1 per 1,000 pages (Exa Contents, You.com Livecrawl), OCR runs $2 per 1,000 (Mistral, $3 with annotations), scraping is about $0.83–$3.20 per 1,000 via Firecrawl's credit tiers, and structured parsing spans $1.25 to $56 per 1,000 on LlamaParse depending on mode.
Which companies use per-page pricing?
Seven in this corpus: Exa, Firecrawl, Frase, LlamaIndex, Mistral, Scalenut, and You.com. The API cluster meters pages directly or through credits; Frase and Scalenut bundle page audits as monthly quotas inside SEO subscriptions.
Why do parse tiers cost up to 45x more per page?
Because the page is an abstraction over wildly different compute: a static HTML page and a scanned table-dense PDF cost very different amounts to process. LlamaParse prices the mode (Fast 1 credit, Cost-effective 3, Agentic 10, Agentic Plus 45 per page) so buyers choose the effort level per document rather than paying a blended rate.
Do failed or empty pages bill?
Policies differ by vendor and endpoint, and it materially affects crawl economics — large crawls hit redirects, soft-404s, and bot walls constantly. Check whether the meter counts attempts or successes, and whether retries re-bill, before estimating from a URL list.
Trivia
-
The plain page-fetch converged to a near-identical price at two unrelated vendors: Exa's Contents API and You.com's Livecrawl add-on both charge exactly $1.00 per 1,000 pages — commodity pricing emerging without a cartel, just a shared cost floor.
-
LlamaParse's v2 repricing collapsed a 1–90-credits-per-page mode-and-model maze into four flat tiers (1, 3, 10, and 45 credits) — and the 45-credit Agentic Plus tier still spans a 45x price range against Fast for the same physical page.
-
On Firecrawl, "1 credit = 1 page" holds only for four of six endpoints: Search bills 2 credits per 10 results and Interact bills 2 credits per browser-minute, so the effective per-page price quietly varies by which endpoint touched the page.
Related billing units
- Credit-Based BillingA billing unit where customers pre-purchase or are allocated a pool of credits that deplete as they use the product, often at variable rates per feature.
- Token-Based PricingA billing unit common in LLM and AI products, where customers are charged per input and output token processed.
- Per-Seat PricingA billing unit where the vendor charges a fixed fee per named user, regardless of how much each user consumes.
- Per-Resolution PricingA billing unit unique to AI customer-support products, where the vendor charges only when an AI agent resolves a customer issue without escalation.
- Bandwidth-Based PricingA billing unit where customers are charged per gigabyte of data transferred out of the platform.
- Per-Function-Invocation PricingA billing unit where customers are charged per serverless function invocation, often combined with a separate compute-time charge.
- CPU-Hour PricingA billing unit where customers are charged for the CPU time their workloads consume, typically measured in vCPU-seconds or vCPU-hours.
- GB-Hour PricingA billing unit where customers are charged for the memory their workloads consume over time, measured in gigabyte-hours.
- GPU-Hour PricingA billing unit where customers are charged for GPU time consumed, typically measured per-second or per-hour by GPU type.
- Per-API-Call PricingA billing unit where customers are charged per API request, regardless of payload size or processing time.
- Per-GB Storage PricingA billing unit where customers are charged per gigabyte of data stored on the platform per month.
- Media-Minute PricingA billing unit where customers are charged per minute of audio or video processed — used by speech, voice, and video AI vendors.
- Per-Request PricingA billing unit where customers are charged per request served — the generic meter for inference endpoints, search, scraping, and browser infrastructure.
- Per-Event PricingA billing unit where customers are charged per event ingested — the native meter of observability and billing-infrastructure platforms.
- Vector Storage PricingA billing unit where customers are charged for vectors stored or indexed — the storage dimension of vector database pricing.
- Per-Character PricingA billing unit where customers are charged per character of text processed — the standard meter for text-to-speech and translation.
- Per-Document PricingA billing unit where customers are charged per document processed or generated — common in AI writing, SEO, and document-intelligence tools.
- Per-Transaction PricingA billing unit where customers are charged per financial or billing transaction processed — the meter of billing and accounting platforms.
- Active-User PricingA billing unit where customers are charged per monthly or daily active user rather than per provisioned seat.
- Per-Task PricingA billing unit where customers are charged per task an automation or agent executes — Zapier's historical unit, now spreading to AI agents.