Google Dorks for web security

A collection of search queries for finding publicly indexed pages, files, and technology footprints during authorized security work. A search result is a lead to investigate, not proof of a vulnerability. Google results can be incomplete, personalized, or change over time.

Start here

1. Choose a topic in the catalog. For a shorter starting set, open best-google-dorks.txt.
2. Limit the query to a domain you own or are authorized to assess. For example, adapt a query to site:example.com filetype:env "DB_PASSWORD" by replacing example.com with your in-scope domain. Check the program's rules before searching related domains or third-party services.
3. Verify any result manually. A matching title, filename, parameter, or login page does not establish exposure or exploitability. If you find sensitive data, do not copy it into an issue or pull request; notify the owner privately.

The all-google-dorks.txt file is a historical, unverified archive. It contains old product fingerprints and queries that may no longer work. Do not treat its size or an entry's presence as evidence of current usefulness.

Find the right list

Goal	Start with
Common file and configuration exposure checks	Short starting list, environment files, backups
Cloud and service configuration	Cloud lists in the catalog
CMS and web server footprints	CMS, web servers
AI applications and agents	Gemini, OpenAI, AI agents, AI guidance
Bug bounty program discovery	Program queries

See the full catalog for every query file and its purpose. Existing filenames are kept stable so external links continue to work; some contain historic spelling or category names.

Search syntax

Google documents site:, filetype:, quoted phrases, exclusion with -, and before:/after: in its search help. Put no space between an operator and its value. For example:

site:example.com filetype:pdf "internal"
site:example.com "login" -inurl:blog

Some older queries use operators or wildcard patterns that are not documented for current Google Search. The former cache: operator no longer works (Google's update). If a query gives surprising results, simplify it and check the current documentation. site: results are not an exhaustive inventory of indexed URLs (Google Search Central).

Interpret results carefully

• A search result for a login page indicates a page, not weak authentication.
• A URL parameter indicates an input to review, not SQL injection, XSS, or open redirect. The filenames for those lists are retained for compatibility.
• A public prompt or agent configuration is not automatically confidential. A key name is not an actual credential.
• Never publish real credentials, private data, or a live target's sensitive URL in contributions.

For site owners

Keep private content behind authentication and authorization, remove accidentally published files, and rotate any exposed credentials. Use noindex only to keep otherwise public content out of search results; it does not restrict access. robots.txt controls crawler access and does not protect secrets: blocked URLs may still appear in results. See Google's guidance and the safe-use guide.

Contribute

Read CONTRIBUTING.md before suggesting a query. Include its purpose, expected signal, likely false positives, and a documentation or public reference when possible. AI-generated suggestions require human review. The repository's small offline check catches duplicate lines and common formatting mistakes; it does not claim to validate live Google results.

License and attribution

The query lists (*.txt) and project documentation (*.md) are licensed under Creative Commons Attribution 4.0 International. You may copy, change, and redistribute them, including commercially. When sharing licensed material, credit Proviesec as the source, link to this repository and the license, and indicate if you made changes. The license applies where copyright or database rights exist; it does not create rights over unprotected facts or short search terms.

The Python checker and GitHub Actions workflow are licensed separately under the MIT License. Reuse of that code requires keeping its copyright and license notice.

Maintained by Proviesec. You can support the project through the links in .github/FUNDING.yml.