scrape.land compared to Bright Data, Zyte and Apify
If you are evaluating web scraping APIs, the useful comparison is structural, not a feature checklist. All four products here will get you a page from a site that does not want to give you one. Where they differ is in how the meter runs — per successful request, per gigabyte, or per second of compute — because that determines whether you can predict a bill before you run the job, and it is the thing that does not change when someone reprices next quarter.
Everything attributed to another company on this page is taken from their own public pricing pages and documentation, with the date it was checked. Prices move. If you find something here that is out of date, tell us and we will fix it.
The short version
| Feature | scrape.land | Bright Data | Zyte | Apify |
|---|---|---|---|---|
| What the meter counts | Delivered requests | Requests (Web Unlocker) or GB (proxies) | Successful responses | Compute time, proxy GB, storage |
| Does the rate vary by target site? | No — one rate | Not on Web Unlocker | Yes — 5 difficulty tiers | Indirectly, via runtime |
| Cost per 1,000 at the $499/mo tier | $0.147 | $1.30 | Varies by tier | Not per-request |
| Free tier | 1,000 requests | 5,000 requests/mo | Trial | $5 monthly credits |
| Can you predict a bill before running? | Yes | Yes on Web Unlocker | Only with the tier mix | Not without measuring |
Bright Data and Zyte figures checked 2026-08-07 against their published pricing pages. Apify figures checked 2026-08-07 against apify.com/pricing. Sources are linked in each section below.
Bright Data
Bright Data is the largest vendor in this space and sells two quite different things, which is the source of most of the confusion when people compare it to anything. Its proxy networks — residential, ISP, datacenter, mobile — are billed per gigabyte of traffic. Its Web Unlocker API is billed per successful request. If you are comparing a scraping API to a scraping API, Web Unlocker is the honest comparison, and it is priced for success the same way we are.
So the difference is not the billing philosophy. It is the rate. As published on Bright Data's Web Unlocker pricing page and checked on 2026-08-07: pay-as-you-go is $1.50 per 1,000 requests, and the Scale plan is $499 per month including 383,000 requests, with additional requests at $1.30 per 1,000. Our Scale plan is also $499 per month, and includes 3,400,000 requests — an effective $0.147 per 1,000.
That is the whole argument, at the same price point, in figures either of us can check. What their higher rate buys is a genuinely larger and more residential IP footprint, and if your targets need residential exits specifically, that is a real reason to pay it. Our pool today is rotating datacenter and ISP-grade addresses, self-validated continuously; residential and mobile are on the roadmap and not something we sell today. Anyone who tells you the cheaper option is strictly better is selling you something.
Worth knowing, because we would rather you hear it from us: Bright Data's Web Unlocker free tier is 5,000 requests a month, larger than our 1,000. If a bigger free allowance is what decides this for you, that is theirs, not ours.
Source: brightdata.com/pricing/web-unlocker, checked 2026-08-07
Zyte
Zyte API is the closest product to ours in shape: send a URL, get a response, pay per successful request. Their documentation is explicit that unsuccessful responses and rate-limiting are free, so the "only pay for what lands" property is not something that distinguishes us from Zyte, and we are not going to claim it does.
The structural difference is difficulty tiers. Zyte classifies each target site into one of five tiers, separately for plain HTTP and browser requests, and the rate you pay is determined by which tier your target lands in — assigned automatically. It is a defensible model: hard sites genuinely cost more to unblock, and you are not subsidising someone else's harder targets.
It also means you cannot compute your bill from your request count alone. You need to know the tier of every domain in your list, and a site that gets harder — or gets reclassified — changes your unit cost without you changing anything. If your targets are all easy, that model is probably cheaper than ours. If your targets are mixed, or if you cannot afford a cost line that moves on its own, ours is one number: $0.147 per 1,000 on Scale, whatever you point it at.
Two further pieces of their model, from the same documentation: volume discounts of 25% to 52% attach to monthly commitments in the $100–$500 range, and several capabilities that we fold into the request price are separately metered — screenshots at $0.002 each, automatic data extraction at $0.0004–$0.0016 per data type, and custom LLM attributes billed per token. Ours bills as request-units on the same meter: a screenshot render is 10 units, AI extraction is 2 or 3 on top of the fetch.
If you are already on Zyte, our Python SDK ships a drop-in adapter: change the import and the key. url, httpResponseBody, httpResponseHeaders, customHttpRequestHeaders, geolocation, sessionContext and the async iterator all map onto our equivalents.
Source: docs.zyte.com/zyte-api/pricing.html, checked 2026-08-07
Apify
Apify is a different category of thing, and choosing between it and a scraping API is usually not a close call once you know which problem you have. It is a platform for running scrapers — a marketplace of pre-built Actors plus hosting for your own — not an endpoint you send a URL to.
Its meter reflects that. As published on apify.com/pricing and checked on 2026-08-07, usage is measured in compute units, where one CU is one gigabyte of RAM running for one hour, charged at $0.20 per CU on Free and Starter down to $0.13 on Business. Proxies are billed separately — residential at $7–$8 per GB, datacenter as included IPs then per additional IP, SERP proxy per thousand queries — and storage and data transfer are metered on top. Plans are prepaid credit buckets: Free $0 with $5 of credits, Starter $29, Scale $199, Business $999.
The consequence is that cost per page is not a number you can look up. It depends on how long your Actor runs, how much memory you gave it, how many proxy gigabytes it pulled and how much it stored. A slow scraper and a fast one, fetching the same pages, cost different amounts. That is the correct model for arbitrary compute — you are renting a machine, and a machine does not know what a "page" is — but it means forecasting requires measurement rather than arithmetic.
The honest split: if you want a pre-built scraper for a specific site that someone else maintains, or you need to run genuinely custom multi-step logic close to the proxies, Apify's marketplace is something we do not offer and have no plans to. If you want to send a URL from your own code and get structured JSON back at a rate you can multiply by your page count, that is what we are.
Where we are the weaker option
A comparison page where the author wins every row is not a comparison. Four cases where you should buy something else, and we would rather say so than have you discover them after integrating:
- You need residential or mobile IPs specifically. We sell rotating datacenter and ISP-grade addresses. Residential and mobile are on the roadmap and not available today. Bright Data's footprint here is the largest in the industry.
- You need deep search-engine result pages or rank tracking. Our search endpoint returns one page of ten organic results and does not paginate. Products built for SERP monitoring will serve you better.
- You want someone else's scraper for a specific site. We give you a primitive, not a maintained per-site parser. Apify's marketplace is a real advantage for that.
- You need a contractual uptime or success-rate SLA. We do not publish one, and we are not going to imply one by quoting an availability figure we have not committed to. The enterprise tiers of the vendors above do offer this.
Where the case for us is strongest
The mirror image. You are the customer this is built for if: your volume is in the hundreds of thousands to tens of millions of pages a month, where a per-1,000 rate difference compounds into real money; your targets are a mixed bag rather than uniformly easy, so a single flat rate beats a tier lottery; you want to forecast a bill by multiplying two numbers; and you would rather create a key and send a request in the next five minutes than book a call. Sign-up to first request needs no card and no conversation.