Why rank tracking needs an API
Checking rankings by hand gives you one person's view: their location, their language, their search history. Rank tracking needs a neutral, repeatable view of the result page, for many keywords, in several markets, every day or week.
Fetching result pages yourself is hard to keep running. Search engines limit repeated automated requests, the layout changes often, and the links in results are wrapped in click tracking. The SERP API for Google, Bing and DuckDuckGo handles proxies, retries and parsing and returns organic results, ads, featured snippets, related questions and related searches as JSON. Blocks and timeouts are not billed.
Step 1: run a query for one market
A request names the engine, the query and the market. country sets both the country the engine localises results for and the exit country of the request. language sets the interface and result language. device is desktop or mobile, which matters because mobile and desktop rankings often differ.
curl https://api.siftzo.com/v1/serp \
-H "X-API-Key: $SIFTZO_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"engine": "google",
"q": "trail running shoes",
"country": "us",
"language": "en",
"device": "mobile",
"page": 1
}'In the response, each entry in results.organic has a position, title, url, displayed_url and snippet. Links are unwrapped from the engine's click tracking, so you can match them against your own URLs directly.
Step 2: go beyond the first page
Many pages you care about rank below the top ten. Use page from 1 to 10 to fetch further result pages. Positions continue across pages: on page 2 with ten results per page, the first organic result has position 11. You can store positions as they arrive without recalculating anything.
On Google, num asks for 10–100 results per page, which can save requests, though Google may still return ten. num is refused for the other engines, and DuckDuckGo supports page 1 only. Decide how deep you need to go per keyword: tracking the top 30 for your main terms and the top 10 for the long tail is a common split.
Read parse_status before you trust a position
Every response has results.parse_status. complete means every section of the page was read. partial means there are results, but some could not be read; warnings says why. failed means the page was not a result page Siftzo understands. If your page is missing from a partial or failed result, record "unknown" rather than "not ranking", so a parsing gap never looks like a ranking drop. Pass include_html: true when you want the raw page to fall back on.
Markets, languages and devices
A keyword is not one ranking but several: the same query returns different results in different countries, in different interface languages and on phones versus desktops. Treat each combination of engine, query, country, language and device as its own series, and never mix them in one chart.
Choose the combinations from your own traffic data rather than covering everything. If most visitors from a market arrive on mobile, track mobile there first. If you publish in several languages, track each language version in the market it is meant for. Without language, the language follows the exit country, so set it explicitly whenever a country has more than one common language, to keep runs comparable over time.
Step 3: schedule, store and report
A simple rank tracker has three tables:
- Keywords, each with engine, country, language and device.
- Checks: keyword, date, request ID and parse status.
- Positions: check, URL, position and result type.
Run checks from your own scheduler at a fixed time of day, and keep your concurrency within your plan's limit; a 429 concurrency_limit answer comes with a Retry-After header. Rankings move a little from hour to hour, so a daily or weekly check at a consistent time gives cleaner trends than frequent ones.
Beyond your own position, store the full top results. Seeing which URLs enter and leave the top ten, and when a featured snippet or a block of related questions appears, often explains a traffic change better than one number does. Teams can use separate API keys with their own logs and credit caps, for example one per client or per project.
Doing it responsibly
Track the keywords you need, at the frequency your decisions need. Every query sends real traffic to the engine, and a smaller, well-chosen keyword set is easier to act on anyway. Use results for analysis of search visibility rather than republishing them. The responsible use guidelines apply to SERP requests as they do to scrapes. Credits per SERP request are listed on the pricing page.
Try it on your own data
1,000 free credits every month, no card needed. Failed requests are never billed.
Questions
Which search engines does the SERP API support?
Google, Bing and DuckDuckGo. Set engine to google, bing or duckduckgo. DuckDuckGo supports page 1 only, and num is available for Google only.
Can I track mobile and desktop rankings separately?
Yes. Set device to mobile or desktop. Store the device with each check, because rankings for the same keyword often differ between the two.
How do positions work on page 2 and beyond?
Positions continue across pages. With ten results per page, the first organic result on page 2 has position 11, so you can store them without recalculating.
What should I do when parse_status is partial?
Read warnings to see which section could not be read. Treat a missing URL in a partial or failed result as unknown rather than as a ranking drop, and request include_html when you need the raw page.