2026-09-17, Google · SERP API · Tutorials

Google search operators that still work, and how to run them through an API

Google documents six search operators on its Search Help page: quotes for an exact match, site:, the minus sign, before:, after: and filetype:. Other operators that SEO practitioners use every day, such as OR, intitle: and inurl:, are not documented on Google’s current help pages; they may work for your query, and they may stop working without notice.

Documented vs undocumented

Checked on 2026-09-17 against Google’s Refine Google searches help page and Google Search Central’s search operators documentation.

“Not documented” means the operator is not on these Google pages. It does not mean it fails, and “documented” does not mean the result set is complete. Google can change or remove any operator.

The documented operators

Exact match: quotes

Put a word or phrase in double quotes to require it. Use it for brand names that are also common words, product model numbers, and exact sentences.

"acme analytics"
"error 0x80070005"
"the fastest way to reconcile invoices"

site:

Restricts results to a domain, URL or URL prefix. Google Search Central’s site: page gives examples such as site:example.com and a URL prefix combined with a term. Combine with - to exclude a domain:

site:example.com/blog onboarding
site:docs.example.com "rate limit"
"acme analytics" -site:acmeanalytics.com

Exclude: minus

Put - directly before a word you want to leave out: jaguar -car, "acme" -careers. Google’s page documents it for words; prefixing another operator, as in -site:amazon.com, is common practice but not shown there, so check it on your queries.

before: and after:

Google’s help page shows before: and after: with a year or a date, and combining both for a range. It shows dates written with slashes and with hyphens; the examples below use hyphens.

"acme analytics" after:2026-09-01
kubernetes cost optimization after:2025 before:2026

filetype:

Restricts to a file type. Search Central says it matches by the content-type HTTP header or the file extension.

"market sizing" saas filetype:pdf
site:example.gov budget filetype:xlsx

The spacing rule

Google’s help page says not to put spaces between the operator and your search term: site:example.com works as an operator, site: example.com does not.

Undocumented operators

OR, intitle:, inurl:, intext: and variants like allintitle: are widely used in SEO workflows, but they are not on the Google pages checked for this post. If a workflow depends on one:

1. Test it on a handful of queries where you know the right answer.
2. Keep the test set and rerun it on a schedule, so you notice when behavior changes.
3. Prefer a documented form when one exists, such as "exact phrase" instead of intext:, or filter the returned link values in your own code instead of relying on inurl:.

Running operators through this API

Operators are part of the query string, so they go in the query field of POST /v1/monitor/google unchanged. The request also needs country (or gl), and takes optional hl, location or uule, device and pages (1 to 10).

curl -X POST https://api.answerline.dev/v1/monitor/google \
  -H "Authorization: Bearer $ANSWERLINE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "query": "\"acme analytics\" -site:acmeanalytics.com after:2026-09-01", "country": "US", "hl": "en", "pages": 2 }'

Inside JSON, escape the inner double quotes (\"). In Python or TypeScript, the JSON encoder does it for you.

Time filters with the url shape

before: and after: cover most date needs. For Google’s relative time filters (past day, week, month, year), send a complete Google search URL in url instead of query. Only q, gl, hl, uule, num, start, tbs and safe are applied from it; num is read as result depth in whole pages of ten; url cannot be combined with query, location, uule or pages.

{ "url": "https://www.google.com/search?q=%22acme+analytics%22&gl=us&hl=en&tbs=qdr:w&num=20" }

Google does not document tbs values. Third-party references such as Bright Data’s Google URL parameter list (dated February 26, 2026) give qdr:h, qdr:d, qdr:w, qdr:m, qdr:y for the past hour, day, week, month and year, and cdr:1,cd_min:MM/DD/YYYY,cd_max:MM/DD/YYYY for a custom range. See Google search parameters for every field.

URLs with tbm (Google verticals such as Images) are rejected, so imagesize: and src: are not usable here.

Use cases

1. Site audits and indexation spot checks

site: shows a sample of what Google returns for a domain or section. Use it to spot pages that should not be there (staging hosts, parameter URLs, thin tag pages) and sections that return nothing.

site:staging.example.com
site:example.com/tag/
site:example.com "lorem ipsum"

Limits you must respect: Google says site: does not necessarily return every indexed URL, a bare site: query without terms returns results in a relatively random order, and the URL Inspection tool in Search Console is the reliable way to check indexing. This API also has no total-results field, and one request returns at most 10 pages, so treat counts as a sample. Search Central also shows using site: with spam terms to find hacked or injected pages on your own site.

2. Competitor content discovery

Find what a competitor has published on a topic, and when:

site:competitor.com "ai overview"
site:competitor.com/blog after:2026-06-01
site:competitor.com filetype:pdf

Collect organicResults[].link, title and date, and diff monthly. For broader competitor tracking, see competitor analysis.

3. PR and brand mentions

Mentions of your brand outside your own properties:

"acme analytics" -site:acmeanalytics.com -site:linkedin.com after:2026-09-01
"acme analytics" "series b"

Google News results have a separate endpoint (POST /v1/monitor/google/news); see Google News API.

4. Content theft and duplication

Search a distinctive sentence from your page, excluding your domain:

"a distinctive sentence copied from your article" -site:example.com

5. Research documents

"benchmark report" "time to first byte" filetype:pdf after:2025

Building an operator query library

Operator queries are most useful as a maintained library that runs on a schedule, not as one-off searches. A workable structure:

Steps to run it:

1. Render templates per brand, domain and date window.
2. Submit as async tasks, with an idempotency key made from the query id, market and run date, so a rerun of the same schedule is not charged twice. See idempotency keys.
3. Store every organic result with the query id, run date, position, page, link, title and date.
4. Diff against the previous run: new links are the alert; disappeared links are context.
5. Check known_answer queries first; if one fails, pause alerts from queries using the same operators until you have looked.

For monitoring what AI assistants say about a brand rather than which pages mention it, see brand mentions in ChatGPT.

A batch example

Many operator queries at once cost less as async tasks: 3 credits each instead of 5, submitted in batches of up to 500. Python with requests:

import os
import time

import requests

API = "https://api.answerline.dev"
HEADERS = {"Authorization": f"Bearer {os.environ['ANSWERLINE_API_KEY']}"}
BRAND, OWN = "acme analytics", "acmeanalytics.com"

queries = [
    f'"{BRAND}" -site:{OWN} after:2026-09-01',
    f'"{BRAND}" review -site:{OWN}',
    f'"{BRAND}" alternative -site:{OWN}',
]
tasks = [{"taskType": "GOOGLE", "payload": {"query": q, "country": "US", "hl": "en", "pages": 2}} for q in queries]

r = requests.post(f"{API}/v1/async/task/batch", json=tasks, headers=HEADERS, timeout=60)
r.raise_for_status()
ids = [x["task"]["id"] for x in r.json()["results"] if x["success"]]

mentions = {}
while ids:
    time.sleep(10)
    for tid in list(ids):
        d = requests.get(f"{API}/v1/async/task/{tid}", headers=HEADERS, timeout=30).json()
        if d["task"]["status"] not in ("COMPLETED", "FAILED"):
            continue
        ids.remove(tid)
        if d["task"]["status"] == "COMPLETED":
            for o in d["response"]["result"].get("organicResults", []):
                mentions.setdefault(o["link"], o)

for link, o in mentions.items():
    print(o.get("date", ""), o["title"], link)

For large or scheduled runs, add a webhook to each task instead of polling; see webhooks.

TypeScript with fetch, for a single synchronous query:

const query = `"acme analytics" -site:acmeanalytics.com after:2026-09-01`;
const res = await fetch("https://api.answerline.dev/v1/monitor/google", {
  method: "POST",
  headers: { Authorization: `Bearer ${process.env.ANSWERLINE_API_KEY}`, "Content-Type": "application/json" },
  body: JSON.stringify({ query, country: "US", hl: "en" }),
});
if (!res.ok) throw new Error(`${res.status}: ${await res.text()}`);
const { result } = await res.json();
console.log(result.organicResults?.map((o: { link: string }) => o.link));

Cost

Operators do not change the price. A Google Search async task costs 3 credits, a synchronous call 5, and each page after the first adds 2. The three-query batch above with pages: 2 is 3 x (3 + 2) = 15 credits. The free tier includes 500 credits per month; see /docs/credits.

Pitfalls

• Spaces after the colon. site: example.com is not an operator.
• Unescaped quotes. A raw " inside a JSON string breaks the body; escape it or let your JSON library encode.
• Using site: as an index count. Google says it is incomplete; use Search Console for indexing questions.
• Mixing query and url. They are mutually exclusive; sending both returns 400 with the field named in details[].
• Relying on undocumented operators without tests. Keep a small regression set of queries with known answers.
• Forgetting market targeting. Operator results still depend on country, hl and location; a mention check for Germany needs country: "DE".
• Excluding by operator only. Filter organicResults[].link against your own domains in code as well, so a changed operator behavior cannot leak your own pages into a mention report.

Related: what is a SERP, SERP feature change tracking, brand protection and is there an official Google Search API.

Request fields and examples are on the Google Search engine page.

Keep reading