Best SERP APIs 2026: Google Search Data Compared

A SERP API turns a search query into structured results without you owning the unblocking problem. On published pricing as of September 5, 2026, Decodo has the cheapest documented entry rate of the products compared here at $0.50 per 1,000 standard requests, Oxylabs bills per successfully delivered result at $1.00 per 1,000 Google results on its entry plans, Bright Data is $1.50 per 1,000 pay-as-you-go with the largest recurring free tier, and Thordata publishes a flat volume ladder from $1.20 down to $0.70 per 1,000. What decides your bill, though, is none of those numbers on its own.

What a SERP API actually does

A SERP API is a scraping API narrowed to one job: you send a search query — plus a location, a language and a device — and get back the results page, usually already parsed into JSON. The provider absorbs proxy rotation, browser fingerprinting, CAPTCHA handling and the parsing of a page whose markup changes without notice.

That last part is the one teams underestimate. Rotating IP addresses is a solved problem you can buy by the gigabyte; keeping a parser alive against a search page that ships layout changes, new result blocks and per-country variants is an ongoing maintenance cost with no end date. Our scraping API vs proxy comparison works through the break-even between the two routes in detail, and the web scraping guide covers where either fits in a full data-collection stack.

If your only goal is checking your own rankings on a handful of keywords, neither is necessary — a rank tracker is. SERP APIs earn their place when you need the raw results at a volume, cadence or geographic spread that a tracker will not sell you, or when the results feed something else: a price-comparison product, an ad-verification pipeline, a retrieval layer for a model.

How we evaluate SERP APIs

ProxyFacts has not run head-to-head SERP benchmarks of these products. Until it does, this comparison is built strictly on documented evidence: vendor pricing pages, product documentation and compliance policies, all fetched and re-verified on September 5, 2026, with each figure traceable to the page it came from. Performance figures such as success rates are labelled as vendor claims throughout. Where a vendor does not publish a number, we say so instead of estimating.

Two consequences of that rule are visible below. A vendor that markets a SERP product without publishing a rate is excluded from the cost comparison rather than given an assumed price, and a vendor whose entry rate differs by target is quoted at its Google rate, because that is the number a reader of this page will actually pay. ProxyFacts is monetised through disclosed affiliate partnerships; those partnerships never change which numbers we report or the order they appear in.

SERP APIs compared

Engineer's take (Hinata): The rate card is the easy half. The number that actually sets your bill is one you compute yourself: keywords × locations × devices × checks per month. A thousand keywords tracked daily on desktop and mobile across three countries is 180,000 SERPs a month — the difference between the cheapest and dearest option on that workload is a real budget line, and it moves by a factor of three depending on which cell of that formula you inflate. My second rule is architectural: never let a vendor's parsed JSON become your database schema. Store the raw payload, parse in a layer you own, and switching provider stays a parser change instead of a migration. And treat JavaScript rendering as a switch with a price tag rather than a feature — on the products here it is either a separate rate or a separate product, and turning it on globally because one target needed it is the most common way a SERP budget doubles quietly.

1. Decodo SERP Scraping API — cheapest documented entry, and the only permanent free plan

Decodo (formerly Smartproxy) publishes the most granular rate card of the products here, and the cheapest way in. Per its SERP scraping API page fetched September 5, 2026, requests are billed by type: $0.50 per 1,000 on standard proxies, $0.75 per 1,000 for standard plus JavaScript rendering, $1.00 per 1,000 on premium proxies, and $1.50 per 1,000 for premium plus JavaScript. All prices exclude VAT, and the page states you only pay for successful requests.

The plan ladder is where the detail matters. The Free plan covers 2,000 standard requests a month at 10 requests per second with no credit card. Starter is $19/mo for 38,000 standard requests — but only 25,000 with JavaScript, on the same plan. Professional at $49/mo covers 163,000 standard or 75,000 JavaScript-rendered requests at 25 requests per second, and Business at $99/mo covers 707,000 or 165,000 at 50 requests per second. Paid plans carry a 14-day money-back guarantee.

Read that JavaScript column carefully, because it is the trap: a plan's headline request count is not the count a JavaScript-rendered workload receives. On Starter the allowance drops by a third; on Business it drops by more than three quarters. If your target list needs rendering, the plan you are actually buying is the smaller number.

On compliance, Decodo screens every registration with automated fraud checks and KYC verification, and blocks high-risk target categories outright — banking, government portals, streaming, app stores and ticketing. That is the posture a legitimate rank-tracking or market-research team should want from its supplier.

2. Oxylabs SERP Scraper API — per-result billing and the deepest volume discount

Oxylabs runs SERP through the same Web Scraper API as the rest of its scraping catalogue, priced per successfully delivered result with a rate that varies by source. Per the pricing page fetched September 5, 2026, Google results cost $1.00 per 1,000 on the Free Trial and Micro plans, $0.90 on Starter, $0.60 on Business and $0.50 on Custom+. With JavaScript rendering, the same ladder reads $1.35, $1.35, $1.30, $1.00 and $0.95.

The plans behind those rates are Micro at $49/mo (up to 98,000 results), Starter at $99/mo (up to 220,000), Business at $999/mo (up to 3,330,000), and a Custom+ band. The free trial covers up to 2,000 results without a credit card — a one-off, not a monthly allowance.

One number on that page is worth isolating, because it is the easiest to misread in a vendor's favour: the plan headline rates of $0.50 per 1,000 and below belong to cheaper targets such as Amazon. Google costs roughly twice that on the entry plans. Any comparison quoting the headline against a rival's Google rate is off by a factor of two, which is why this page quotes the Google row.

Billing per delivered result rather than per request is a genuine difference at scale: failed attempts do not consume budget, so an unstable target degrades your latency rather than your invoice. Oxylabs also documents its compliance posture in unusual detail — a KYC form for every customer at signup, risk-based escalation to ID verification, and membership of the Ethical Web Data Collection Initiative — which matters when a procurement reviewer asks where the traffic behind your rank tracking comes from.

3. Bright Data SERP API — largest free tier, and the one to test on hostile targets

Bright Data sells SERP as a distinct product on a pay-as-you-go rate with no commitment: $1.50 per 1,000 requests, per its SERP API page fetched September 5, 2026, with browser rendering bundled rather than billed separately. The Scale plan is $499/mo including 380,000 requests, with additional requests at $1.30 per 1,000; Enterprise is custom. Billing is success-based — the page states you pay only for what is successfully delivered.

Its free tier is the largest recurring allowance found here: 5,000 records a month with no credit card. That is enough to run a real evaluation — a few hundred keywords across a couple of locations, repeated weekly — without a purchase order, which is the honest way to test whether any of these products handles your specific target list.

Two promotions were displayed on the fetch date: a site-wide 50% discount banner and a first-deposit match of up to $500. Neither is recorded as pricing here, because a promotion is not a rate you can plan a budget around.

Bright Data also runs the strictest vetting of any provider in our dataset: network access is limited to verified companies, with corporate email verification, a use-case description and potentially an introductory call plus government-issued ID. For a company with a compliance function that is a feature; for a solo developer it is a barrier, and worth knowing before you invest time in an integration.

4. Thordata SERP API — the most legible volume ladder

Thordata publishes a plain tier table for SERP, which makes it the easiest of these to forecast. Per its SERP API pricing page fetched September 5, 2026 and unchanged from our August capture: 15,000 responses for $18 ($1.20 per 1,000), 50,000 for $55 ($1.10), 150,000 for $135 ($0.90), 500,000 for $400 ($0.80) and 1,000,000 for $700 ($0.70), with custom terms above that. The free trial is 5,000 responses over 7 days — one-off, not monthly. The product covers Google and Bing; DuckDuckGo and Yandex are sold as separate APIs, so check which engines your workload actually needs before comparing on price alone.

Thordata publishes a customer-vetting policy describing multi-level review and video identity verification where needed. What it does not publish is how the residential IP addresses behind the network are acquired, and when we asked directly the vendor declined to provide that documentation, citing confidentiality. We record that as unverified rather than as evidence of wrongdoing — and it is why this page profiles Thordata on its published numbers without recommending it over vendors whose sourcing is documented. Our Thordata review sets out the full picture, and ethical proxy sourcing explains what a good answer to that question looks like.

Who is missing from this comparison, and why

Several providers in our tracked dataset are absent from the table above, each for a reason worth stating rather than hiding. SOAX markets a SERP API within its Web Data API line but publishes no per-request or per-result rate on the product page fetched September 5, 2026 — the page routes pricing to a sales conversation, so there is nothing to compare. IPRoyal publishes no SERP product at all: its Web Scraping API remains an early-access waitlist with no search endpoint described, and search-results work is positioned as a proxy use case. DataImpulse sells proxy access only, with no scraping or SERP product on any page we have fetched.

An unpriced product is excluded, never estimated. A comparison that fills gaps with plausible-looking guesses is worse than one with fewer columns, because the guess is indistinguishable from the measured figures beside it.

Specialist SERP-only vendors — the companies that sell nothing but search-results APIs — are also absent, because they are not yet in our tracked dataset. Inclusion there is free, never for sale, and gated on the same due diligence every provider passes: a registered legal entity, documented IP sourcing, a customer-vetting policy, and evidence for headline numbers. We would rather publish the vendors we can account for than pad a table with ones we cannot.

The four costs a SERP rate card does not show you

1. Rendering. On every product here, JavaScript rendering is either a higher rate or a separate product. Enabling it globally because a handful of targets needed it is the most common way a forecast doubles.
2. The engine and target mix. A "SERP API" is rarely one price. Oxylabs charges by source, Thordata sells DuckDuckGo and Yandex separately, and Decodo distinguishes standard from premium proxies. Price your actual engine mix, not the cheapest cell in the table.
3. Parsing lock-in. Structured JSON is the feature you will most regret coupling to. If your warehouse schema mirrors one vendor's response shape, changing vendor becomes a migration rather than a swap.
4. Cadence. Rank tracking is a scheduling problem before it is a scraping problem. Doubling your check frequency doubles the bill exactly, while halving it usually costs nothing you can measure. Our web scraping cost estimation guide walks through modelling that properly, and proxies for SEO monitoring covers the proxy-side route to the same data.

Success rates and pool sizes are vendor claims

Every performance figure on these vendors' pages — success rate, uptime, pool size, response time — is a marketing number until independently measured, and none of them has been independently verified by ProxyFacts. Success-based billing softens the risk, since a failed request does not bill on any of the products here, but it does not remove it: failures still cost you latency and schedule slip.

The practical response is to run the free allowances against your own ugliest target before committing to a plan — the single hardest query on your list decides feasibility, not the average one. For the techniques that reduce failure rates regardless of vendor, see how to scrape without getting blocked.

Compliance and legality

Every provider here gates access behind identity checks, and each publishes an acceptable-use policy prohibiting the obvious abuses. Search-results collection for rank tracking, ad verification, market research and price intelligence is the mainstream, defensible use of these products. Bypassing logins or paywalls, harvesting personal data, and automating account creation are not, and are prohibited by all of them.

The legal position on collecting public search results is genuinely nuanced: it turns on what you collect, where you operate and what contractual terms you have accepted. Search engines' terms of service generally prohibit automated querying, which is a contract question rather than a criminal one in most jurisdictions — but "generally" is doing real work in that sentence. We walk through the landscape in is web scraping legal? This article is not legal advice; get advice for your specific situation.

Which SERP API should you choose?

• Choose Decodo if price per request is the constraint and your targets do not all need rendering: it has the cheapest documented entry rate here, the only permanent free plan, and the only published rate limits.
• Choose Oxylabs if you want parsed Google results billed per delivered result, and your volume is large enough to reach the tiers where its rate falls below the field.
• Choose Bright Data if your targets are heavily defended or you want the largest free allowance to evaluate with, and your company can clear a strict verification process.
• Consider Thordata if you are forecasting high volume and want a published ladder rather than a negotiation — with the caveat that its IP sourcing is undocumented, which is a question a compliance reviewer will ask before you are asked to answer it.