Google Recipes API: Recipe SERP Data Without Managing Browsers
If your team tracks recipe rankings, ingredient trends, or cook-time patterns across Google's recipe results, the slow part is rarely the analysis — it's keeping a browser fleet, proxy pool, and anti-bot handling alive.
A hosted scraping API replaces that with one request: you pass a Google recipe SERP URL, the service renders the page in a real browser, rotates the IP, and returns HTML or parsed JSON. This page covers what fields come back, how to request structured JSON, what each option costs in credits, and how to tell the difference between an empty result and a blocked one.
What a recipe SERP request returns
Each Google Recipes result is expected to carry the recipe title, the source domain, the ingredient list, cook time, rating, calories per serving, and the recipe URL. That set covers the three workflows most teams build: ingredient extraction for trend or product analysis, rating analysis for demand forecasting, and cook-time extraction for kitchen scheduling.
A useful way to think about the output is in two layers. The first layer is the page itself — the HTML a real browser renders. The second layer is the parsed record you actually want to store. The API can return either, and the choice changes both your parsing work and your credit cost.
Field map: default fields vs. selector-driven fields
The fields below are the ones the page describes as available on a Google Recipes result. The 'how to get it' column is the practical decision: some fields come back in the parsed record without extra configuration, while others need you to name them explicitly with extract_rules (CSS selectors) or ai_extract_rules (a natural-language spec).
Credit cost: pick the cheapest option that still returns the page
Credits are consumed per successful request, and the amount depends on which options you enable. The base request starts at 1 credit. JavaScript rendering adds 5 credits. Premium proxies cost 10 credits without JS rendering or 25 with it. Stealth proxy costs 75 credits. Credits are valid for one month and do not roll over.
A practical ordering: try the base request first. If the page comes back but the recipe block is missing, add render_js (5 credits). If the page is blocked or shows a CAPTCHA, move to premium_proxy (10 or 25) before jumping to stealth_proxy (75). Use country_code when the recipe results differ by locale — for example, when you need the version of a SERP that a user in a specific market would see.
Getting structured JSON back
There are two routes to parsed JSON. The first is extract_rules: you map CSS selectors to field names, and the response comes back as parsed JSON with those names. The second is ai_extract_rules: you describe the fields you want in natural language, and the AI-driven extraction fills them in. Both are documented as returning parsed JSON.
A useful rule of thumb: use extract_rules when the page structure is stable and you care about exact field names; use ai_extract_rules when the recipe markup varies across sources and you would rather describe the target than maintain selectors. You can also request the page as clean Markdown if the next step is feeding it into an LLM for normalization or trend analysis.
Empty result vs. blocked page vs. wrong locale
Three failure modes look similar in a log but need different fixes.
- Empty or partial JSON with a 200 response: the page rendered, but your selectors did not match. Check the field map and confirm the selector targets the recipe block, not a wrapper that only appears in some layouts.
- A page that returns a challenge or a CAPTCHA instead of recipe content: the request was blocked. Move from the base request to premium_proxy, and add render_js if the challenge is JavaScript-driven. Stealth proxy is the last step, not the first.
- Recipe results that are valid but for the wrong market: the SERP is localized. Set country_code to the market you need before re-running the request.
Keeping these three separate in your pipeline avoids the common mistake of escalating to the most expensive option when the real problem was a selector or a locale.
Worked example: one request, one recipe record
Suppose you want to track how a single recipe query's top result changes over a month. You send the Google recipe SERP URL for that query with render_js enabled, because the recipe block is rendered client-side. The response is parsed JSON containing the recipe title, source domain, ingredient list, cook time, rating, calories per serving, and recipe URL.
You then store one row per day with those fields, plus the date and the country_code you used. Over a month, the row set lets you answer questions like 'did the top result's rating change?' or 'did the ingredient list shift?' without re-parsing HTML. If a day's row is empty, the failure distinction above tells you whether to fix the selector, escalate the proxy, or check the locale.
Languages and setup
The API is callable from any language. Published SDKs cover Python, Node.js, Go, PHP, Ruby, and Java, and plain cURL works too. The recommended setup path is to test in the dashboard request builder first, then drop the generated snippet into your codebase. That keeps the first request short and makes the option choices visible before they reach production.
Account and legal checks before you deploy
You do not need a Google account to use the API. You sign up for an API key, which comes with 1,000 free credits and no card required, then send the URL. On the legal side, public-facing pages can usually be scraped, but review the target site's terms of service before deploying and avoid post-login content. That is a general caution, not a legal opinion — your own counsel and the target's terms govern.
| Field | What it holds | How to get it |
|---|---|---|
| Recipe title | Name of the recipe as shown in the SERP | Returned in the parsed record |
| Source domain | Site that published the recipe | Returned in the parsed record |
| Ingredient list | Ingredients listed for the recipe | Returned in the parsed record; use extract_rules if you need a specific shape |
| Cook time | Preparation or cooking duration | Returned in the parsed record |
| Rating | Recipe rating value | Returned in the parsed record |
| Calories per serving | Calorie figure per serving | Returned in the parsed record |
| Recipe URL | Link to the full recipe | Returned in the parsed record |
| Custom fields | Any other value rendered on the public page | extract_rules (CSS selectors) or ai_extract_rules (natural-language spec) |
Field availability depends on what the public page renders. Custom fields require an extraction rule; the default record covers the seven fields listed above.
| Option | Credits per successful request | When to use it |
|---|---|---|
| Base request | 1 | First attempt; page returns recipe content without JS rendering |
| JavaScript rendering (render_js) | 5 | Recipe block is rendered client-side and missing from the base response |
| Premium proxy, no JS rendering | 10 | Page is blocked but does not need JS rendering |
| Premium proxy with JS rendering | 25 | Page is blocked and the recipe block needs JS rendering |
| Stealth proxy (stealth_proxy) | 75 | Last step when premium proxy still returns a challenge |
Credits are valid for one month and do not roll over. Costs apply to successful requests. The base request starts at 1 credit.
Frequently asked questions
Do I need a Google account to scrape Google Recipes?
No Google account is required. You sign up for an API key, which includes 1,000 free credits with no card, then send the recipe SERP URL and the service handles the rest.
Is scraping Google Recipes legal?
Public-facing pages can usually be scraped, but review the target site's terms of service before deploying and avoid post-login content. This is a general caution, not legal advice for your situation.
What data can I extract from Google Recipes?
In principle, any value that appears on the public page. To capture a specific value, either map a CSS selector to a field name with extract_rules, or describe the field in plain language and let ai_extract_rules handle the matching — both return structured JSON. Without either, the default record already covers recipe title, source domain, ingredient list, cook time, rating, calories per serving, and recipe URL.
How do I get structured JSON from Google Recipes?
Choose one of two inputs. With extract_rules, you supply CSS selectors paired with the field names you want, and the response is JSON keyed by those names. With ai_extract_rules, you write a plain-language description of the fields instead of selectors, and the AI-driven extraction returns the same JSON shape. Either way, the response body is parsed JSON rather than raw HTML.
Which programming languages can I use to scrape Google Recipes?
Any language. Published SDKs cover Python, Node.js, Go, PHP, Ruby, and Java, and plain cURL works too. Pick whatever your stack already uses.
How much does it cost to scrape Google Recipes?
1,000 free credits on signup, no card required. Paid plans start at $19/month for higher volume and concurrency. Credit cost per request depends on options: 5 credits for JS rendering, 25 for premium_proxy with JS, 75 for stealth_proxy.
Why did my request return empty JSON instead of an error?
An empty or partial JSON response with a 200 status usually means the page rendered but your selectors did not match the recipe block. Check the field map and confirm the selector targets the recipe content. If the page instead returns a challenge or CAPTCHA, the request was blocked and you should escalate to premium_proxy or stealth_proxy.