If you pull Google results through an API without a date filter, you mostly get evergreen pages: docs, homepages, Wikipedia, and two-year-old blog posts. Fine for rank tracking. Useless when you only want what showed up yesterday or during Q1.
Google already has a built-in parameter for this: tbs. It is the same string Google puts in the browser URL when you click Tools → Any time → Past week. If you need a google search api date filter, passing tbs straight through your search request is the cleanest way to do it.
Here is the cheat sheet I keep around, plus what actually changes in the JSON response across three different tbs values on the same query.
The tbs cheat sheet
| What you want | tbs value | Notes |
|---|---|---|
| Past hour | qdr:h | Incident or breaking-news checks |
| Past 24 hours | qdr:d | Daily monitoring jobs |
| Past week | qdr:w | Weekly digests |
| Past month | qdr:m | Append a number for N units, e.g. qdr:m6 = past 6 months |
| Past year | qdr:y | Annual lookups |
| Custom date range | cdr:1,cd_min:1/1/2026,cd_max:3/31/2026 | Dates must be M/D/YYYY (US order) |
| Sort by date | sbd:1 | Newest first instead of relevance |
| Past week + newest first | qdr:w,sbd:1 | Combine with a comma |
Same query, three tbs settings
Using a single endpoint (POST https://api.serpbase.dev/google/search), here is a small Python snippet that runs the same query ("openai news") three ways and prints the date field from organic:
import os import requests URL = "https://api.serpbase.dev/google/search" API_KEY = os.environ["SERPBASE_API_KEY"] def run_query(query: str, tbs: str | None = None): payload = {"q": query, "gl": "us", "hl": "en"} if tbs: payload["tbs"] = tbs resp = requests.post( URL, headers={ "X-API-Key": API_KEY, "Content-Type": "application/json", }, json=payload, timeout=30, ) data = resp.json() if data.get("status") != 0: raise RuntimeError(f"status={data.get('status')} id={data.get('request_id')}") return data.get("organic", []) cases = [ ("No tbs (default)", None), ("Past 24h (qdr:d)", "qdr:d"), ("Q1 2026 (cdr)", "cdr:1,cd_min:1/1/2026,cd_max:3/31/2026"), ] for label, tbs in cases: items = run_query("openai news", tbs) dated = sum(1 for x in items if x.get("date")) print(f"\n=== {label} ({dated}/{len(items)} have a date) ===") for item in items[:3]: print(f" [{item.get('date') or 'no date'}] {item['title']}") print(f" {item['link']}") What changes between the three runs:
- No
tbs(default): Most of the top organic results have nodatefield at all (often only 1 or 2 out of 7–10). You get the company's own/newshub, Wikipedia, and high-authority landing pages. -
tbs: "qdr:d"(past 24 hours): Nearly every organic item carries a relativedatestring ("3 hours ago","14 hours ago"). Evergreen hubs drop out and fresh coverage takes the top slots. -
tbs: "cdr:1,cd_min:1/1/2026,cd_max:3/31/2026"(custom range): Every dated result falls inside Q1 2026 ("Jan 18, 2026","Mar 12, 2026", etc.). Useful for backfilling historical research or auditing what was published during a specific launch window.
Three details that usually trip people up
- ISO dates fail silently.
2026-01-01will not behave like1/1/2026. Always formatcd_minandcd_maxasM/D/YYYYwithout leading zeros. - Plain English does not work.
tbsis forwarded straight to Google, so"past week"or"last 24 hours"is ignored. Stick toqdr:w/qdr:d. - Google decides the date. If an old article gets edited and its visible "Updated on" timestamp changes, Google can include it inside
qdr:d. For strict deduplication in daily monitors, store seenlinkvalues locally (SQLite or a set) instead of assumingqdr:donly returns brand-new URLs.
If you want more tested examples and edge cases (including combining tbs with raw Google search URLs), we wrote a fuller breakdown here: Google Search API Date Filter: Get Results From the Past Day, Week or a Custom Range. Full disclosure: I work on SerpBase, though the tbs syntax itself works the same anywhere Google's tbs parameter is supported.